xxq250
/
homework-jianmu
1
0
Fork 0
Code Issues Pull Requests 1 Actions Packages Projects Releases Wiki Activity

Merge branch '3.0' into test/abs_tag_compute

This commit is contained in:
wenzhouwww@live.cn 2022-06-11 10:53:08 +08:00
parent 8b5269795b 6aaf3476a1
commit d4b8e4afda
154 changed files with 6982 additions and 1054 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

16
docs-cn/20-third-party/01-grafana.mdx vendored
View File

@ -3,6 +3,9 @@ sidebar_label: Grafana
title: Grafana
---
import Tabs from "@theme/Tabs";
import TabItem from "@theme/TabItem";
TDengine 能够与开源数据可视化系统 [Grafana](https://www.grafana.com/) 快速集成搭建数据监测报警系统，整个过程无需任何代码开发，TDengine 中数据表的内容可以在仪表盘(DashBoard)上进行可视化展现。关于 TDengine 插件的使用您可以在[GitHub](https://github.com/taosdata/grafanaplugin/blob/master/README.md)中了解更多。
## 前置条件
@ -23,9 +26,12 @@ TDengine 能够与开源数据可视化系统 [Grafana](https://www.grafana.com/
## 配置 Grafana
### 安装方式一：安装脚本
### 安装 Grafana Plugin 并配置数据源
将集群信息设置为环境变量（也可以使用 [`.env`(dotenv) 文件](https://hexdocs.pm/dotenvy/dotenv-file-format.html) ）：
<Tabs defaultValue="script">
<TabItem value="script" label="使用安装脚本">
将集群信息设置为环境变量；也可以使用 `.env` 文件，请参考 [dotenv](https://hexdocs.pm/dotenvy/dotenv-file-format.html)：
```sh
export TDENGINE_API=http://tdengine.local:6041
@ -42,7 +48,8 @@ bash -c "$(curl -fsSL https://raw.githubusercontent.com/taosdata/grafanaplugin/m
该脚本将自动安装 Grafana 插件并配置数据源。安装完毕后，需要重启 Grafana 服务后生效。
### 安装方式二：手动安装 TDengine 数据源插件
</TabItem>
<TabItem value="manual" label="手动安装并配置">
使用 [`grafana-cli` 命令行工具](https://grafana.com/docs/grafana/latest/administration/cli/) 进行插件[安装](https://grafana.com/grafana/plugins/tdengine-datasource/?tab=installation)。
@ -94,6 +101,9 @@ GF_INSTALL_PLUGINS=tdengine-datasource
![TDengine Database Grafana plugin add data source](./add_datasource4.webp)
</TabItem>
</Tabs>
### 创建 Dashboard
回到主界面创建 Dashboard，点击 Add Query 进入面板查询页面：

14
docs-en/20-third-party/01-grafana.mdx vendored
View File

@ -3,6 +3,9 @@ sidebar_label: Grafana
title: Grafana
---
import Tabs from "@theme/Tabs";
import TabItem from "@theme/TabItem";
TDengine can be quickly integrated with the open-source data visualization system [Grafana](https://www.grafana.com/) to build a data monitoring and alerting system. The whole process does not require any code development. And you can visualize the contents of the data tables in TDengine on a dashboard.
You can learn more about using the TDengine plugin on [GitHub](https://github.com/taosdata/grafanaplugin/blob/master/README.md).
@ -25,7 +28,10 @@ TDengine currently supports Grafana versions 7.5 and above. Users can go to the
## Configuring Grafana
### Option 1: Install with `install.sh`
### Install Grafana Plugin and Configure Data Source
<Tabs defaultValue="script">
<TabItem value="script" label="Using Script">
Set the url and authorization environment variables by `export` or a [`.env`(dotenv) file](https://hexdocs.pm/dotenvy/dotenv-file-format.html):
@ -46,7 +52,8 @@ With this script, TDengine data source plugin and the Grafana data source will b
And then, restart Grafana service and open Grafana in web-browser, usually <http://localhost:3000>.
### Option 2: Install Plugin Manually
</TabItem>
<TabItem value="manual" label="Install & Configure Manually">
Follow the installation steps in [Grafana](https://grafana.com/grafana/plugins/tdengine-datasource/?tab=installation) with the [``grafana-cli`` command-line tool](https://grafana.com/docs/grafana/latest/administration/cli/) for plugin installation.
@ -98,6 +105,9 @@ Click `Save & Test` to test. You should see a success message if the test worked
![TDengine Database TDinsight plugin add database 4](./grafana/add_datasource4.webp)
</TabItem>
</Tabs>
### Create Dashboard
Go back to the main interface to create a dashboard and click Add Query to enter the panel query page:

7
include/common/tmsg.h
View File

@ -1205,6 +1205,7 @@ typedef struct {
int8_t completed; // all results are returned to client
int8_t precision;
int8_t compressed;
int8_t streamBlockType;
int32_t compLen;
int32_t numOfRows;
int32_t numOfCols;
@ -2510,7 +2511,7 @@ typedef struct {
int32_t tSerializeSTableIndexRsp(void* buf, int32_t bufLen, const STableIndexRsp* pRsp);
int32_t tDeserializeSTableIndexRsp(void* buf, int32_t bufLen, STableIndexRsp* pRsp);
void tFreeSTableIndexInfo(void *pInfo);
typedef struct {
int8_t mqMsgType;
@ -2752,8 +2753,8 @@ typedef struct {
char* msg;
} SVDeleteReq;
int32_t tSerializeSVDeleteReq(void *buf, int32_t bufLen, SVDeleteReq *pReq);
int32_t tDeserializeSVDeleteReq(void *buf, int32_t bufLen, SVDeleteReq *pReq);
int32_t tSerializeSVDeleteReq(void* buf, int32_t bufLen, SVDeleteReq* pReq);
int32_t tDeserializeSVDeleteReq(void* buf, int32_t bufLen, SVDeleteReq* pReq);
typedef struct {
int64_t affectedRows;

3
include/common/tmsgdef.h
View File

@ -194,6 +194,7 @@ enum {
TD_DEF_MSG_TYPE(TDMT_VND_ALTER_CONFIG, "alter-config", NULL, NULL)
TD_DEF_MSG_TYPE(TDMT_VND_ALTER_REPLICA, "alter-replica", NULL, NULL)
TD_DEF_MSG_TYPE(TDMT_VND_ALTER_CONFIRM, "alter-confirm", NULL, NULL)
TD_DEF_MSG_TYPE(TDMT_VND_ALTER_HASHRANGE, "alter-hashrange", NULL, NULL)
TD_DEF_MSG_TYPE(TDMT_VND_COMPACT, "compact", NULL, NULL)
TD_NEW_MSG_SEG(TDMT_QND_MSG)
@ -233,6 +234,8 @@ enum {
TD_DEF_MSG_TYPE(TDMT_SYNC_COMMON_RESPONSE, "sync-common-response", NULL, NULL)
TD_DEF_MSG_TYPE(TDMT_SYNC_APPLY_MSG, "sync-apply-msg", NULL, NULL)
TD_DEF_MSG_TYPE(TDMT_SYNC_CONFIG_CHANGE, "sync-config-change", NULL, NULL)
TD_DEF_MSG_TYPE(TDMT_SYNC_SNAPSHOT_SEND, "sync-snapshot-send", NULL, NULL)
TD_DEF_MSG_TYPE(TDMT_SYNC_SNAPSHOT_RSP, "sync-snapshot-rsp", NULL, NULL)
#if defined(TD_MSG_NUMBER_)
TDMT_MAX

4
include/libs/catalog/catalog.h
View File

@ -67,6 +67,7 @@ typedef struct SCatalogReq {
SArray *pUdf; // element is udf name
SArray *pIndex; // element is index name
SArray *pUser; // element is SUserAuthInfo
SArray *pTableIndex; // element is SNAME
bool qNodeRequired; // valid qnode
bool forceUpdate;
} SCatalogReq;
@ -82,6 +83,7 @@ typedef struct SMetaData {
SArray *pDbInfo; // pRes = SDbInfo*
SArray *pTableMeta; // pRes = STableMeta*
SArray *pTableHash; // pRes = SVgroupInfo*
SArray *pTableIndex; // pRes = SArray<STableIndexInfo>*
SArray *pUdfList; // pRes = SFuncInfo*
SArray *pIndex; // pRes = SIndexInfo*
SArray *pUser; // pRes = bool*
@ -277,7 +279,7 @@ int32_t catalogGetDBCfg(SCatalog* pCtg, void *pRpc, const SEpSet* pMgmtEps, cons
int32_t catalogGetIndexMeta(SCatalog* pCtg, void *pRpc, const SEpSet* pMgmtEps, const char* indexName, SIndexInfo* pInfo);
int32_t catalogGetTableIndex(SCatalog* pCtg, void *pTrans, const SEpSet* pMgmtEps, const char* tbFName, SArray** pRes);
int32_t catalogGetTableIndex(SCatalog* pCtg, void *pTrans, const SEpSet* pMgmtEps, const SName* pTableName, SArray** pRes);
int32_t catalogGetUdfInfo(SCatalog* pCtg, void *pRpc, const SEpSet* pMgmtEps, const char* funcName, SFuncInfo* pInfo);

1
include/libs/function/function.h
View File

@ -146,6 +146,7 @@ typedef struct SqlFunctionCtx {
struct SDiskbasedBuf *pBuf;
struct SSDataBlock *pSrcBlock;
int32_t curBufPage;
bool increase;
char udfName[TSDB_FUNC_NAME_LEN];
} SqlFunctionCtx;

4
include/libs/function/functionMgt.h
View File

@ -129,6 +129,10 @@ typedef enum EFunctionType {
FUNCTION_TYPE_SPREAD_MERGE,
FUNCTION_TYPE_HISTOGRAM_PARTIAL,
FUNCTION_TYPE_HISTOGRAM_MERGE,
FUNCTION_TYPE_HYPERLOGLOG_PARTIAL,
FUNCTION_TYPE_HYPERLOGLOG_MERGE,
FUNCTION_TYPE_ELAPSED_PARTIAL,
FUNCTION_TYPE_ELAPSED_MERGE,
// user defined funcion
FUNCTION_TYPE_UDF = 10000

1
include/libs/nodes/plannodes.h
View File

@ -304,6 +304,7 @@ typedef struct SDownstreamSourceNode {
typedef struct SExchangePhysiNode {
SPhysiNode node;
int32_t srcGroupId; // group id of datasource suplans
bool singleChannel;
SNodeList* pSrcEndPoints; // element is SDownstreamSource, scheduler fill by calling qSetSuplanExecutionNode
} SExchangePhysiNode;

2
include/libs/stream/tstream.h
View File

@ -308,9 +308,11 @@ static FORCE_INLINE int32_t streamTaskOutput(SStreamTask* pTask, SStreamDataBloc
if (pTask->sinkType == TASK_SINK__TABLE) {
ASSERT(pTask->dispatchType == TASK_DISPATCH__NONE);
pTask->tbSink.tbSinkFunc(pTask, pTask->tbSink.vnode, 0, pBlock->blocks);
taosFreeQitem(pBlock);
} else if (pTask->sinkType == TASK_SINK__SMA) {
ASSERT(pTask->dispatchType == TASK_DISPATCH__NONE);
pTask->smaSink.smaSink(pTask->smaSink.vnode, pTask->smaSink.smaId, pBlock->blocks);
taosFreeQitem(pBlock);
} else {
ASSERT(pTask->dispatchType != TASK_DISPATCH__NONE);
taosWriteQitem(pTask->outputQueue->queue, pBlock);

27
include/libs/sync/sync.h
View File

@ -88,11 +88,16 @@ typedef struct SReConfigCbMeta {
} SReConfigCbMeta;
typedef struct SSnapshot {
void *data;
void* data;
SyncIndex lastApplyIndex;
SyncTerm lastApplyTerm;
SyncIndex lastConfigIndex;
} SSnapshot;
typedef struct SSnapshotMeta {
SyncIndex lastConfigIndex;
} SSnapshotMeta;
typedef struct SSyncFSM {
void* data;
@ -141,10 +146,28 @@ typedef struct SSyncLogStore {
// return commit index of log
SyncIndex (*getCommitIndex)(struct SSyncLogStore* pLogStore);
// refactor, log[0 .. n] ==> log[m .. n]
int32_t (*syncLogSetBeginIndex)(struct SSyncLogStore* pLogStore, SyncIndex beginIndex);
int32_t (*syncLogResetBeginIndex)(struct SSyncLogStore* pLogStore);
SyncIndex (*syncLogBeginIndex)(struct SSyncLogStore* pLogStore);
SyncIndex (*syncLogEndIndex)(struct SSyncLogStore* pLogStore);
bool (*syncLogIsEmpty)(struct SSyncLogStore* pLogStore);
int32_t (*syncLogEntryCount)(struct SSyncLogStore* pLogStore);
bool (*syncLogInRange)(struct SSyncLogStore* pLogStore, SyncIndex index);
SyncIndex (*syncLogWriteIndex)(struct SSyncLogStore* pLogStore);
SyncIndex (*syncLogLastIndex)(struct SSyncLogStore* pLogStore);
SyncTerm (*syncLogLastTerm)(struct SSyncLogStore* pLogStore);
int32_t (*syncLogAppendEntry)(struct SSyncLogStore* pLogStore, SSyncRaftEntry* pEntry);
int32_t (*syncLogGetEntry)(struct SSyncLogStore* pLogStore, SyncIndex index, SSyncRaftEntry** ppEntry);
int32_t (*syncLogTruncate)(struct SSyncLogStore* pLogStore, SyncIndex fromIndex);
} SSyncLogStore;
typedef struct SSyncInfo {
bool isStandBy;
bool snapshotEnable;
SyncGroupId vgId;
SSyncCfg syncCfg;
char path[TSDB_FILENAME_LEN];
@ -172,6 +195,8 @@ bool syncEnvIsStart();
const char* syncStr(ESyncState state);
bool syncIsRestoreFinish(int64_t rid);
int32_t syncGetSnapshotMeta(int64_t rid, struct SSnapshotMeta* sMeta);
// to be moved to static
void syncStartNormal(int64_t rid);
void syncStartStandBy(int64_t rid);

94
include/libs/sync/syncTools.h
View File

@ -301,6 +301,7 @@ typedef struct SyncAppendEntries {
SyncIndex prevLogIndex;
SyncTerm prevLogTerm;
SyncIndex commitIndex;
SyncTerm privateTerm;
uint32_t dataLen;
char data[];
} SyncAppendEntries;
@ -332,6 +333,7 @@ typedef struct SyncAppendEntriesReply {
SRaftId destId;
// private data
SyncTerm term;
SyncTerm privateTerm;
bool success;
SyncIndex matchIndex;
} SyncAppendEntriesReply;
@ -385,6 +387,75 @@ void syncApplyMsgPrint2(char* s, const SyncApplyMsg* pMsg);
void syncApplyMsgLog(const SyncApplyMsg* pMsg);
void syncApplyMsgLog2(char* s, const SyncApplyMsg* pMsg);
// ---------------------------------------------
typedef struct SyncSnapshotSend {
uint32_t bytes;
int32_t vgId;
uint32_t msgType;
SRaftId srcId;
SRaftId destId;
SyncTerm term;
SyncIndex lastIndex; // lastIndex of snapshot
SyncTerm lastTerm; // lastTerm of snapshot
SyncTerm privateTerm;
int32_t seq;
uint32_t dataLen;
char data[];
} SyncSnapshotSend;
SyncSnapshotSend* syncSnapshotSendBuild(uint32_t dataLen, int32_t vgId);
void syncSnapshotSendDestroy(SyncSnapshotSend* pMsg);
void syncSnapshotSendSerialize(const SyncSnapshotSend* pMsg, char* buf, uint32_t bufLen);
void syncSnapshotSendDeserialize(const char* buf, uint32_t len, SyncSnapshotSend* pMsg);
char* syncSnapshotSendSerialize2(const SyncSnapshotSend* pMsg, uint32_t* len);
SyncSnapshotSend* syncSnapshotSendDeserialize2(const char* buf, uint32_t len);
void syncSnapshotSend2RpcMsg(const SyncSnapshotSend* pMsg, SRpcMsg* pRpcMsg);
void syncSnapshotSendFromRpcMsg(const SRpcMsg* pRpcMsg, SyncSnapshotSend* pMsg);
SyncSnapshotSend* syncSnapshotSendFromRpcMsg2(const SRpcMsg* pRpcMsg);
cJSON* syncSnapshotSend2Json(const SyncSnapshotSend* pMsg);
char* syncSnapshotSend2Str(const SyncSnapshotSend* pMsg);
// for debug ----------------------
void syncSnapshotSendPrint(const SyncSnapshotSend* pMsg);
void syncSnapshotSendPrint2(char* s, const SyncSnapshotSend* pMsg);
void syncSnapshotSendLog(const SyncSnapshotSend* pMsg);
void syncSnapshotSendLog2(char* s, const SyncSnapshotSend* pMsg);
// ---------------------------------------------
typedef struct SyncSnapshotRsp {
uint32_t bytes;
int32_t vgId;
uint32_t msgType;
SRaftId srcId;
SRaftId destId;
SyncTerm term;
SyncIndex lastIndex;
SyncTerm lastTerm;
SyncTerm privateTerm;
int32_t ack;
int32_t code;
} SyncSnapshotRsp;
SyncSnapshotRsp* syncSnapshotRspBuild(int32_t vgId);
void syncSnapshotRspDestroy(SyncSnapshotRsp* pMsg);
void syncSnapshotRspSerialize(const SyncSnapshotRsp* pMsg, char* buf, uint32_t bufLen);
void syncSnapshotRspDeserialize(const char* buf, uint32_t len, SyncSnapshotRsp* pMsg);
char* syncSnapshotRspSerialize2(const SyncSnapshotRsp* pMsg, uint32_t* len);
SyncSnapshotRsp* syncSnapshotRspDeserialize2(const char* buf, uint32_t len);
void syncSnapshotRsp2RpcMsg(const SyncSnapshotRsp* pMsg, SRpcMsg* pRpcMsg);
void syncSnapshotRspFromRpcMsg(const SRpcMsg* pRpcMsg, SyncSnapshotRsp* pMsg);
SyncSnapshotRsp* syncSnapshotRspFromRpcMsg2(const SRpcMsg* pRpcMsg);
cJSON* syncSnapshotRsp2Json(const SyncSnapshotRsp* pMsg);
char* syncSnapshotRsp2Str(const SyncSnapshotRsp* pMsg);
// for debug ----------------------
void syncSnapshotRspPrint(const SyncSnapshotRsp* pMsg);
void syncSnapshotRspPrint2(char* s, const SyncSnapshotRsp* pMsg);
void syncSnapshotRspLog(const SyncSnapshotRsp* pMsg);
void syncSnapshotRspLog2(char* s, const SyncSnapshotRsp* pMsg);
// on message ----------------------
int32_t syncNodeOnPingCb(SSyncNode* ths, SyncPing* pMsg);
int32_t syncNodeOnPingReplyCb(SSyncNode* ths, SyncPingReply* pMsg);
@ -395,6 +466,29 @@ int32_t syncNodeOnRequestVoteReplyCb(SSyncNode* ths, SyncRequestVoteReply* pMsg)
int32_t syncNodeOnAppendEntriesCb(SSyncNode* ths, SyncAppendEntries* pMsg);
int32_t syncNodeOnAppendEntriesReplyCb(SSyncNode* ths, SyncAppendEntriesReply* pMsg);
int32_t syncNodeOnRequestVoteSnapshotCb(SSyncNode* ths, SyncRequestVote* pMsg);
int32_t syncNodeOnRequestVoteReplySnapshotCb(SSyncNode* ths, SyncRequestVoteReply* pMsg);
int32_t syncNodeOnAppendEntriesSnapshotCb(SSyncNode* ths, SyncAppendEntries* pMsg);
int32_t syncNodeOnAppendEntriesReplySnapshotCb(SSyncNode* ths, SyncAppendEntriesReply* pMsg);
int32_t syncNodeOnSnapshotSendCb(SSyncNode* ths, SyncSnapshotSend* pMsg);
int32_t syncNodeOnSnapshotRspCb(SSyncNode* ths, SyncSnapshotRsp* pMsg);
// -----------------------------------------
typedef int32_t (*FpOnPingCb)(SSyncNode* ths, SyncPing* pMsg);
typedef int32_t (*FpOnPingReplyCb)(SSyncNode* ths, SyncPingReply* pMsg);
typedef int32_t (*FpOnClientRequestCb)(SSyncNode* ths, SyncClientRequest* pMsg);
typedef int32_t (*FpOnRequestVoteCb)(SSyncNode* ths, SyncRequestVote* pMsg);
typedef int32_t (*FpOnRequestVoteReplyCb)(SSyncNode* ths, SyncRequestVoteReply* pMsg);
typedef int32_t (*FpOnAppendEntriesCb)(SSyncNode* ths, SyncAppendEntries* pMsg);
typedef int32_t (*FpOnAppendEntriesReplyCb)(SSyncNode* ths, SyncAppendEntriesReply* pMsg);
typedef int32_t (*FpOnTimeoutCb)(SSyncNode* pSyncNode, SyncTimeout* pMsg);
typedef int32_t (*FpOnSnapshotSendCb)(SSyncNode* ths, SyncSnapshotSend* pMsg);
typedef int32_t (*FpOnSnapshotRspCb)(SSyncNode* ths, SyncSnapshotRsp* pMsg);
// option ----------------------------------
bool syncNodeSnapshotEnable(SSyncNode* pSyncNode);
// ---------------------------------------------
#ifdef __cplusplus

14
include/libs/wal/wal.h
View File

@ -141,6 +141,8 @@ typedef struct SWal {
// ctl
int64_t refId;
TdThreadMutex mutex;
// ref
SHashObj *pRefHash; // ref -> SWalRef
// path
char path[WAL_PATH_LEN];
// reusable write head
@ -184,7 +186,7 @@ int32_t walRollback(SWal *, int64_t ver);
// notify that previous logs can be pruned safely
int32_t walBeginSnapshot(SWal *, int64_t ver);
int32_t walEndSnapshot(SWal *);
void walRestoreFromSnapshot(SWal *, int64_t ver);
int32_t walRestoreFromSnapshot(SWal *, int64_t ver);
// int32_t walDataCorrupted(SWal*);
// read
@ -199,6 +201,16 @@ int32_t walFetchHead(SWalReadHandle *pRead, int64_t ver, SWalHead *pHead);
int32_t walFetchBody(SWalReadHandle *pRead, SWalHead **ppHead);
int32_t walSkipFetchBody(SWalReadHandle *pRead, const SWalHead *pHead);
typedef struct {
int64_t refId;
int64_t ver;
} SWalRef;
SWalRef *walOpenRef(SWal *);
void walCloseRef(SWalRef *);
int32_t walRefVer(SWalRef *, int64_t ver);
int32_t walUnrefVer(SWal *);
// deprecated
#if 0
int32_t walRead(SWal *, SWalHead **, int64_t ver);

53
source/client/src/clientSml.c
View File

@ -303,7 +303,7 @@ static int32_t smlApplySchemaAction(SSmlHandle* info, SSchemaAction* action) {
uError("SML:0x%" PRIx64 " apply schema action. reset query cache. error: %s", info->id, taos_errstr(res2));
}
taos_free_result(res2);
taosMsleep(500);
taosMsleep(10);
}
break;
}
@ -327,7 +327,7 @@ static int32_t smlApplySchemaAction(SSmlHandle* info, SSchemaAction* action) {
uError("SML:0x%" PRIx64 " apply schema action. reset query cache. error: %s", info->id, taos_errstr(res2));
}
taos_free_result(res2);
taosMsleep(500);
taosMsleep(10);
}
break;
}
@ -350,7 +350,7 @@ static int32_t smlApplySchemaAction(SSmlHandle* info, SSchemaAction* action) {
uError("SML:0x%" PRIx64 " apply schema action. reset query cache. error: %s", info->id, taos_errstr(res2));
}
taos_free_result(res2);
taosMsleep(500);
taosMsleep(10);
}
break;
}
@ -373,7 +373,7 @@ static int32_t smlApplySchemaAction(SSmlHandle* info, SSchemaAction* action) {
uError("SML:0x%" PRIx64 " apply schema action. reset query cache. error: %s", info->id, taos_errstr(res2));
}
taos_free_result(res2);
taosMsleep(500);
taosMsleep(10);
}
break;
}
@ -424,7 +424,7 @@ static int32_t smlApplySchemaAction(SSmlHandle* info, SSchemaAction* action) {
uError("SML:0x%" PRIx64 " apply schema action. reset query cache. error: %s", info->id, taos_errstr(res2));
}
taos_free_result(res2);
taosMsleep(500);
taosMsleep(10);
}
break;
}
@ -461,18 +461,18 @@ static int32_t smlProcessSchemaAction(SSmlHandle* info, SSchema* schemaField, SH
static int32_t smlModifyDBSchemas(SSmlHandle* info) {
int32_t code = 0;
SEpSet ep = getEpSet_s(&info->taos->pAppInfo->mgmtEp);
SName pName = {TSDB_TABLE_NAME_T, info->taos->acctId, {0}, {0}};
strcpy(pName.dbname, info->pRequest->pDb);
SSmlSTableMeta** tableMetaSml = (SSmlSTableMeta**)taosHashIterate(info->superTables, NULL);
while (tableMetaSml) {
SSmlSTableMeta* sTableData = *tableMetaSml;
STableMeta *pTableMeta = NULL;
SEpSet ep = getEpSet_s(&info->taos->pAppInfo->mgmtEp);
size_t superTableLen = 0;
void *superTable = taosHashGetKey(tableMetaSml, &superTableLen);
SName pName = {TSDB_TABLE_NAME_T, info->taos->acctId, {0}, {0}};
strcpy(pName.dbname, info->pRequest->pDb);
memset(pName.tname, 0, TSDB_TABLE_NAME_LEN);
memcpy(pName.tname, superTable, superTableLen);
code = catalogGetSTableMeta(info->pCatalog, info->taos->pAppInfo->pTransporter, &ep, &pName, &pTableMeta);
@ -487,7 +487,7 @@ static int32_t smlModifyDBSchemas(SSmlHandle* info) {
code = smlApplySchemaAction(info, &schemaAction);
if (code != TSDB_CODE_SUCCESS) {
uError("SML:0x%"PRIx64" smlApplySchemaAction failed. can not create %s", info->id, schemaAction.createSTable.sTableName);
return code;
goto end;
}
info->cost.numOfCreateSTables++;
}else if (code == TSDB_CODE_SUCCESS) {
@ -502,7 +502,7 @@ static int32_t smlModifyDBSchemas(SSmlHandle* info) {
code = smlProcessSchemaAction(info, pTableMeta->schema, hashTmp, sTableData->tags, &schemaAction, true);
if (code != TSDB_CODE_SUCCESS) {
taosHashCleanup(hashTmp);
return code;
goto end;
}
taosHashClear(hashTmp);
@ -512,29 +512,33 @@ static int32_t smlModifyDBSchemas(SSmlHandle* info) {
code = smlProcessSchemaAction(info, pTableMeta->schema, hashTmp, sTableData->cols, &schemaAction, false);
taosHashCleanup(hashTmp);
if (code != TSDB_CODE_SUCCESS) {
return code;
goto end;
}
code = catalogRefreshTableMeta(info->pCatalog, info->taos->pAppInfo->pTransporter, &ep, &pName, -1);
code = catalogRefreshTableMeta(info->pCatalog, info->taos->pAppInfo->pTransporter, &ep, &pName, 1);
if (code != TSDB_CODE_SUCCESS) {
return code;
goto end;
}
} else {
uError("SML:0x%"PRIx64" load table meta error: %s", info->id, tstrerror(code));
return code;
goto end;
}
if(pTableMeta) taosMemoryFree(pTableMeta);
code = catalogGetSTableMeta(info->pCatalog, info->taos->pAppInfo->pTransporter, &ep, &pName, &pTableMeta);
if (code != TSDB_CODE_SUCCESS) {
uError("SML:0x%"PRIx64" catalogGetSTableMeta failed. super table name %s", info->id, (char*)superTable);
return code;
goto end;
}
sTableData->tableMeta = pTableMeta;
tableMetaSml = (SSmlSTableMeta**)taosHashIterate(info->superTables, tableMetaSml);
}
return 0;
end:
catalogRefreshTableMeta(info->pCatalog, info->taos->pAppInfo->pTransporter, &ep, &pName, 1);
return code;
}
//=========================================================================
@ -1544,7 +1548,7 @@ static int32_t smlParseTSFromJSONObj(SSmlHandle *info, cJSON *root, int64_t *tsV
}
size_t typeLen = strlen(type->valuestring);
if (typeLen == 1 && type->valuestring[0] == 's') {
if (typeLen == 1 && (type->valuestring[0] == 's' || type->valuestring[0] == 'S')) {
//seconds
timeDouble = timeDouble * 1e9;
if(smlDoubleToInt64OverFlow(timeDouble)){
@ -1552,9 +1556,10 @@ static int32_t smlParseTSFromJSONObj(SSmlHandle *info, cJSON *root, int64_t *tsV
return TSDB_CODE_TSC_INVALID_TIME_STAMP;
}
*tsVal = timeDouble;
} else if (typeLen == 2 && type->valuestring[1] == 's') {
} else if (typeLen == 2 && (type->valuestring[1] == 's' || type->valuestring[1] == 'S')) {
switch (type->valuestring[0]) {
case 'm':
case 'M':
//milliseconds
timeDouble = timeDouble * 1e6;
if(smlDoubleToInt64OverFlow(timeDouble)){
@ -1564,6 +1569,7 @@ static int32_t smlParseTSFromJSONObj(SSmlHandle *info, cJSON *root, int64_t *tsV
*tsVal = timeDouble;
break;
case 'u':
case 'U':
//microseconds
timeDouble = timeDouble * 1e3;
if(smlDoubleToInt64OverFlow(timeDouble)){
@ -1573,6 +1579,7 @@ static int32_t smlParseTSFromJSONObj(SSmlHandle *info, cJSON *root, int64_t *tsV
*tsVal = timeDouble;
break;
case 'n':
case 'N':
//nanoseconds
*tsVal = timeDouble;
break;
@ -2285,6 +2292,8 @@ static int32_t smlParseLine(SSmlHandle *info, char* lines[], int numLines){
static int smlProcess(SSmlHandle *info, char* lines[], int numLines) {
int32_t code = TSDB_CODE_SUCCESS;
int32_t retryNum = 0;
info->cost.parseTime = taosGetTimestampUs();
code = smlParseLine(info, lines, numLines);
@ -2298,7 +2307,12 @@ static int smlProcess(SSmlHandle *info, char* lines[], int numLines) {
info->cost.numOfCTables = taosHashGetSize(info->childTables);
info->cost.schemaTime = taosGetTimestampUs();
code = smlModifyDBSchemas(info);
do{
code = smlModifyDBSchemas(info);
if (code == 0) break;
} while (retryNum++ < taosHashGetSize(info->superTables));
if (code != 0) {
uError("SML:0x%"PRIx64" smlModifyDBSchemas error : %s", info->id, tstrerror(code));
goto cleanup;
@ -2409,6 +2423,7 @@ TAOS_RES* taos_schemaless_insert(TAOS* taos, char* lines[], int numLines, int pr
info->pRequest->code = smlProcess(info, lines, numLines);
end:
info->taos->schemalessType = 0;
uDebug("result:%s", info->msgBuf.buf);
smlDestroyInfo(info);
return (TAOS_RES*)request;

56
source/client/test/smlTest.cpp
View File

@ -1272,14 +1272,40 @@ TEST(testCase, sml_params_Test) {
};
TAOS_RES* res = taos_schemaless_insert(taos, (char**)sql, 1, TSDB_SML_LINE_PROTOCOL, TSDB_SML_TIMESTAMP_MILLI_SECONDS);
ASSERT_EQ(taos_errno(res), TSDB_CODE_PAR_DB_NOT_SPECIFIED);
taos_free_result(pRes);
taos_free_result(res);
pRes = taos_query(taos, "use param");
taos_free_result(pRes);
taos_free_result(res);
res = taos_schemaless_insert(taos, (char**)sql, 1, TSDB_SML_LINE_PROTOCOL, TSDB_SML_TIMESTAMP_MILLI_SECONDS);
ASSERT_EQ(taos_errno(res), TSDB_CODE_SML_INVALID_DB_CONF);
taos_free_result(res);
}
TEST(testCase, sml_16384_Test) {
TAOS *taos = taos_connect("localhost", "root", "taosdata", NULL, 0);
ASSERT_NE(taos, nullptr);
TAOS_RES* pRes = taos_query(taos, "create database if not exists d16384 schemaless 1");
taos_free_result(pRes);
const char *sql[] = {
"qelhxo,id=pnnqhsa,t0=t,t1=127i8 c0=t,c1=127i8 1626006833639000000",
};
pRes = taos_query(taos, "use d16384");
taos_free_result(pRes);
TAOS_RES* res = taos_schemaless_insert(taos, (char**)sql, 1, TSDB_SML_LINE_PROTOCOL, 0);
ASSERT_EQ(taos_errno(res), 0);
taos_free_result(res);
const char *sql1[] = {
"qelhxo,id=pnnqhsa,t0=t,t1=127i8 c0=f,c1=127i8,c11=L\"ncharColValue\",c10=t 1626006833639000000",
};
TAOS_RES* res1 = taos_schemaless_insert(taos, (char**)sql1, 1, TSDB_SML_LINE_PROTOCOL, 0);
ASSERT_EQ(taos_errno(res1), 0);
taos_free_result(res1);
}
TEST(testCase, sml_oom_Test) {
@ -1303,3 +1329,29 @@ TEST(testCase, sml_oom_Test) {
ASSERT_EQ(taos_errno(res), 0);
taos_free_result(pRes);
}
TEST(testCase, sml_16368_Test) {
TAOS *taos = taos_connect("localhost", "root", "taosdata", NULL, 0);
ASSERT_NE(taos, nullptr);
TAOS_RES* pRes = taos_query(taos, "create database if not exists d16368 schemaless 1");
taos_free_result(pRes);
pRes = taos_query(taos, "use d16368");
taos_free_result(pRes);
const char *sql[] = {
"[{\"metric\": \"st123456\", \"timestamp\": {\"value\": 1626006833639000, \"type\": \"us\"}, \"value\": 1, \"tags\": {\"t1\": 3, \"t2\": {\"value\": 4, \"type\": \"double\"}, \"t3\": {\"value\": \"t3\", \"type\": \"binary\"}}},\n"
"{\"metric\": \"st123456\", \"timestamp\": {\"value\": 1626006833739000, \"type\": \"us\"}, \"value\": 2, \"tags\": {\"t1\": {\"value\": 4, \"type\": \"double\"}, \"t3\": {\"value\": \"t4\", \"type\": \"binary\"}, \"t2\": {\"value\": 5, \"type\": \"double\"}, \"t4\": {\"value\": 5, \"type\": \"double\"}}},\n"
"{\"metric\": \"stb_name\", \"timestamp\": {\"value\": 1626006833639100, \"type\": \"us\"}, \"value\": 3, \"tags\": {\"t2\": {\"value\": 5, \"type\": \"double\"}, \"t3\": {\"value\": \"ste\", \"type\": \"nchar\"}}},\n"
"{\"metric\": \"stf567890\", \"timestamp\": {\"value\": 1626006833639200, \"type\": \"us\"}, \"value\": 4, \"tags\": {\"t1\": {\"value\": 4, \"type\": \"bigint\"}, \"t3\": {\"value\": \"t4\", \"type\": \"binary\"}, \"t2\": {\"value\": 5, \"type\": \"double\"}, \"t4\": {\"value\": 5, \"type\": \"double\"}}},\n"
"{\"metric\": \"st123456\", \"timestamp\": {\"value\": 1626006833639300, \"type\": \"us\"}, \"value\": {\"value\": 5, \"type\": \"double\"}, \"tags\": {\"t1\": {\"value\": 4, \"type\": \"double\"}, \"t2\": 5.0, \"t3\": {\"value\": \"t4\", \"type\": \"binary\"}}},\n"
"{\"metric\": \"stb_name\", \"timestamp\": {\"value\": 1626006833639400, \"type\": \"us\"}, \"value\": {\"value\": 6, \"type\": \"double\"}, \"tags\": {\"t2\": 5.0, \"t3\": {\"value\": \"ste2\", \"type\": \"nchar\"}}},\n"
"{\"metric\": \"stb_name\", \"timestamp\": {\"value\": 1626006834639400, \"type\": \"us\"}, \"value\": {\"value\": 7, \"type\": \"double\"}, \"tags\": {\"t2\": {\"value\": 5.0, \"type\": \"double\"}, \"t3\": {\"value\": \"ste2\", \"type\": \"nchar\"}}},\n"
"{\"metric\": \"st123456\", \"timestamp\": {\"value\": 1626006833839006, \"type\": \"us\"}, \"value\": {\"value\": 8, \"type\": \"double\"}, \"tags\": {\"t1\": {\"value\": 4, \"type\": \"double\"}, \"t3\": {\"value\": \"t4\", \"type\": \"binary\"}, \"t2\": {\"value\": 5, \"type\": \"double\"}, \"t4\": {\"value\": 5, \"type\": \"double\"}}},\n"
"{\"metric\": \"st123456\", \"timestamp\": {\"value\": 1626006833939007, \"type\": \"us\"}, \"value\": {\"value\": 9, \"type\": \"double\"}, \"tags\": {\"t1\": 4, \"t3\": {\"value\": \"t4\", \"type\": \"binary\"}, \"t2\": {\"value\": 5, \"type\": \"double\"}, \"t4\": {\"value\": 5, \"type\": \"double\"}}}]"
};
pRes = taos_schemaless_insert(taos, (char**)sql, sizeof(sql)/sizeof(sql[0]), TSDB_SML_JSON_PROTOCOL, TSDB_SML_TIMESTAMP_MICRO_SECONDS);
ASSERT_EQ(taos_errno(pRes), 0);
taos_free_result(pRes);
}

12
source/common/src/tdatablock.c
View File

@ -1752,7 +1752,7 @@ SSubmitReq* tdBlockToSubmit(const SArray* pBlocks, const STSchema* pTSchema, boo
taosArrayClear(tagArray);
taosArrayPush(tagArray, &tagVal);
tTagNew(tagArray, 1, false, &pTag);
if (!pTag) {
if (pTag == NULL) {
tdDestroySVCreateTbReq(&createTbReq);
taosArrayDestroy(tagArray);
return NULL;
@ -1763,9 +1763,7 @@ SSubmitReq* tdBlockToSubmit(const SArray* pBlocks, const STSchema* pTSchema, boo
tEncodeSize(tEncodeSVCreateTbReq, &createTbReq, schemaLen, code);
tdDestroySVCreateTbReq(&createTbReq);
if (code < 0) {
tdDestroySVCreateTbReq(&createTbReq);
taosArrayDestroy(tagArray);
return NULL;
}
@ -1775,6 +1773,7 @@ SSubmitReq* tdBlockToSubmit(const SArray* pBlocks, const STSchema* pTSchema, boo
}
// assign data
// TODO
ret = taosMemoryCalloc(1, cap + 46);
ret = POINTER_SHIFT(ret, 46);
ret->header.vgId = vgId;
@ -1804,8 +1803,7 @@ SSubmitReq* tdBlockToSubmit(const SArray* pBlocks, const STSchema* pTSchema, boo
int32_t schemaLen = 0;
if (createTb) {
SVCreateTbReq createTbReq = {0};
char* cname = taosMemoryCalloc(1, TSDB_TABLE_FNAME_LEN);
snprintf(cname, TSDB_TABLE_FNAME_LEN, "%s:%ld", stbFullName, pDataBlock->info.groupId);
char* cname = buildCtbNameByGroupId(stbFullName, pDataBlock->info.groupId);
createTbReq.name = cname;
createTbReq.flags = 0;
createTbReq.type = TSDB_CHILD_TABLE;
@ -1819,7 +1817,7 @@ SSubmitReq* tdBlockToSubmit(const SArray* pBlocks, const STSchema* pTSchema, boo
taosArrayPush(tagArray, &tagVal);
STag* pTag = NULL;
tTagNew(tagArray, 1, false, &pTag);
if (!pTag) {
if (pTag == NULL) {
tdDestroySVCreateTbReq(&createTbReq);
taosArrayDestroy(tagArray);
taosMemoryFreeClear(ret);
@ -1945,7 +1943,6 @@ void blockCompressEncode(const SSDataBlock* pBlock, char* data, int32_t* dataLen
const char* blockCompressDecode(SSDataBlock* pBlock, int32_t numOfCols, int32_t numOfRows, const char* pData) {
blockDataEnsureCapacity(pBlock, numOfRows);
pBlock->info.rows = numOfRows;
const char* pStart = pData;
@ -2019,6 +2016,7 @@ const char* blockCompressDecode(SSDataBlock* pBlock, int32_t numOfCols, int32_t
pStart += colLen[i];
}
pBlock->info.rows = numOfRows;
ASSERT(pStart - pData == dataLen);
return pStart;
}

24
source/common/src/tdataformat.c
View File

@ -244,7 +244,7 @@ int32_t tTSRowNew(STSRowBuilder *pBuilder, SArray *pArray, STSchema *pTSchema, S
}
}
// ASSERT(flags); // only 1 column(ts)
ASSERT(flags);
// decide
uint32_t nData = 0;
@ -268,8 +268,8 @@ int32_t tTSRowNew(STSRowBuilder *pBuilder, SArray *pArray, STSchema *pTSchema, S
nDataT = BIT2_SIZE(pTSchema->numOfCols - 1) + pTSchema->flen + ntv;
break;
default:
break; // only ts column
// ASSERT(0);
break;
ASSERT(0);
}
uint8_t tflags = 0;
@ -374,7 +374,7 @@ int32_t tTSRowNew(STSRowBuilder *pBuilder, SArray *pArray, STSchema *pTSchema, S
ptv = pf + pTSchema->flen;
break;
default:
// ASSERT(0);
ASSERT(0);
break;
}
} else {
@ -421,12 +421,26 @@ int32_t tTSRowNew(STSRowBuilder *pBuilder, SArray *pArray, STSchema *pTSchema, S
_set_none:
if ((flags & 0xf0) == 0) {
setBitMap(pb, 0, iColumn - 1, flags);
if (flags & TSROW_HAS_VAL) { // set 0
if (IS_VAR_DATA_TYPE(pTColumn->type)) {
*(VarDataOffsetT *)(pf + pTColumn->offset) = 0;
} else {
tPutValue(pf + pTColumn->offset, &((SValue){0}), pTColumn->type);
}
}
}
continue;
_set_null:
if ((flags & 0xf0) == 0) {
setBitMap(pb, 1, iColumn - 1, flags);
if (flags & TSROW_HAS_VAL) { // set 0
if (IS_VAR_DATA_TYPE(pTColumn->type)) {
*(VarDataOffsetT *)(pf + pTColumn->offset) = 0;
} else {
tPutValue(pf + pTColumn->offset, &((SValue){0}), pTColumn->type);
}
}
} else {
SET_IDX(pidx, pTSKVRow->nCols, nkv, flags);
pTSKVRow->nCols++;
@ -497,7 +511,7 @@ void tTSRowGet(STSRow2 *pRow, STSchema *pTSchema, int32_t iCol, SColVal *pColVal
SValue value;
ASSERT(iCol < pTSchema->numOfCols);
// ASSERT(flags); // only 1 ts column
ASSERT(flags);
ASSERT(pRow->sver == pTSchema->version);
if (iCol == 0) {

9
source/common/src/tmsg.c
View File

@ -2491,6 +2491,15 @@ int32_t tDeserializeSTableIndexRsp(void *buf, int32_t bufLen, STableIndexRsp *pR
return 0;
}
void tFreeSTableIndexInfo(void* info) {
if (NULL == info) {
return;
}
STableIndexInfo *pInfo = (STableIndexInfo*)info;
taosMemoryFree(pInfo->expr);
}
int32_t tSerializeSShowReq(void *buf, int32_t bufLen, SShowReq *pReq) {
SEncoder encoder = {0};

84
source/common/test/dataformatTest.cpp
View File

@ -52,61 +52,61 @@ STSchema *genSTSchema(int16_t nCols) {
switch (i) {
case 0: {
pSchema[0].type = TSDB_DATA_TYPE_TIMESTAMP;
pSchema[0].bytes = TYPE_BYTES[pSchema[0].type];
pSchema[i].type = TSDB_DATA_TYPE_TIMESTAMP;
pSchema[i].bytes = TYPE_BYTES[pSchema[i].type];
} break;
case 1: {
pSchema[1].type = TSDB_DATA_TYPE_INT;
pSchema[1].bytes = TYPE_BYTES[pSchema[1].type];
pSchema[i].type = TSDB_DATA_TYPE_INT;
pSchema[i].bytes = TYPE_BYTES[pSchema[i].type];
;
} break;
case 2: {
pSchema[2].type = TSDB_DATA_TYPE_BIGINT;
pSchema[2].bytes = TYPE_BYTES[pSchema[2].type];
pSchema[i].type = TSDB_DATA_TYPE_BIGINT;
pSchema[i].bytes = TYPE_BYTES[pSchema[i].type];
} break;
case 3: {
pSchema[3].type = TSDB_DATA_TYPE_FLOAT;
pSchema[3].bytes = TYPE_BYTES[pSchema[3].type];
pSchema[i].type = TSDB_DATA_TYPE_FLOAT;
pSchema[i].bytes = TYPE_BYTES[pSchema[i].type];
} break;
case 4: {
pSchema[4].type = TSDB_DATA_TYPE_DOUBLE;
pSchema[4].bytes = TYPE_BYTES[pSchema[4].type];
pSchema[i].type = TSDB_DATA_TYPE_DOUBLE;
pSchema[i].bytes = TYPE_BYTES[pSchema[i].type];
} break;
case 5: {
pSchema[5].type = TSDB_DATA_TYPE_BINARY;
pSchema[5].bytes = 12;
pSchema[i].type = TSDB_DATA_TYPE_BINARY;
pSchema[i].bytes = 12;
} break;
case 6: {
pSchema[6].type = TSDB_DATA_TYPE_NCHAR;
pSchema[6].bytes = 42;
pSchema[i].type = TSDB_DATA_TYPE_NCHAR;
pSchema[i].bytes = 42;
} break;
case 7: {
pSchema[7].type = TSDB_DATA_TYPE_TINYINT;
pSchema[7].bytes = TYPE_BYTES[pSchema[7].type];
pSchema[i].type = TSDB_DATA_TYPE_TINYINT;
pSchema[i].bytes = TYPE_BYTES[pSchema[i].type];
} break;
case 8: {
pSchema[8].type = TSDB_DATA_TYPE_SMALLINT;
pSchema[8].bytes = TYPE_BYTES[pSchema[8].type];
pSchema[i].type = TSDB_DATA_TYPE_SMALLINT;
pSchema[i].bytes = TYPE_BYTES[pSchema[i].type];
} break;
case 9: {
pSchema[9].type = TSDB_DATA_TYPE_BOOL;
pSchema[9].bytes = TYPE_BYTES[pSchema[9].type];
pSchema[i].type = TSDB_DATA_TYPE_BOOL;
pSchema[i].bytes = TYPE_BYTES[pSchema[i].type];
} break;
case 10: {
pSchema[10].type = TSDB_DATA_TYPE_UTINYINT;
pSchema[10].bytes = TYPE_BYTES[pSchema[10].type];
pSchema[i].type = TSDB_DATA_TYPE_UTINYINT;
pSchema[i].bytes = TYPE_BYTES[pSchema[i].type];
} break;
case 11: {
pSchema[11].type = TSDB_DATA_TYPE_USMALLINT;
pSchema[11].bytes = TYPE_BYTES[pSchema[11].type];
pSchema[i].type = TSDB_DATA_TYPE_USMALLINT;
pSchema[i].bytes = TYPE_BYTES[pSchema[i].type];
} break;
case 12: {
pSchema[12].type = TSDB_DATA_TYPE_UINT;
pSchema[12].bytes = TYPE_BYTES[pSchema[12].type];
pSchema[i].type = TSDB_DATA_TYPE_UINT;
pSchema[i].bytes = TYPE_BYTES[pSchema[i].type];
} break;
case 13: {
pSchema[13].type = TSDB_DATA_TYPE_UBIGINT;
pSchema[13].bytes = TYPE_BYTES[pSchema[13].type];
pSchema[i].type = TSDB_DATA_TYPE_UBIGINT;
pSchema[i].bytes = TYPE_BYTES[pSchema[i].type];
} break;
default:
@ -146,9 +146,9 @@ static int32_t genTestData(const char **data, int16_t nCols, SArray **pArray) {
case 0:
sscanf(data[i], "%" PRIi64, &colVal.value.ts);
break;
case 1: {
case 1:
sscanf(data[i], "%" PRIi32, &colVal.value.i32);
} break;
break;
case 2:
sscanf(data[i], "%" PRIi64, &colVal.value.i64);
break;
@ -274,9 +274,6 @@ int32_t debugPrintSColVal(SColVal *cv, int8_t type) {
case TSDB_DATA_TYPE_MEDIUMBLOB:
printf("MedBLOB ");
break;
// case TSDB_DATA_TYPE_BINARY:
// printf("BINARY ");
// break;
case TSDB_DATA_TYPE_MAX:
printf("UNDEF ");
break;
@ -404,9 +401,10 @@ static void checkTSRow(const char **data, STSRow2 *row, STSchema *pTSchema) {
}
TEST(testCase, AllNormTest) {
int16_t nCols = 1;
STSRow2 *row = nullptr;
SArray *pArray = taosArrayInit(nCols, sizeof(SColVal));
int16_t nCols = 14;
STSRowBuilder rb = {0};
STSRow2 *row = nullptr;
SArray *pArray = taosArrayInit(nCols, sizeof(SColVal));
EXPECT_NE(pArray, nullptr);
STSchema *pTSchema = genSTSchema(nCols);
@ -414,15 +412,16 @@ TEST(testCase, AllNormTest) {
// ts timestamp, c1 int, c2 bigint, c3 float, c4 double, c5 binary(10), c6 nchar(10), c7 tinyint, c8 smallint,
// c9 bool
char *data[10] = {"1653694220000", "10", "20", "10.1", "10.1", "binary10", "nchar10", "10", "10", "1"};
char *data[14] = {"1653694220000", "no", "nu", "nu", "nu", "nu", "nu", "nu", "nu", "no", "no", "no", "no", "no"};
genTestData((const char **)&data, nCols, &pArray);
tTSRowNew(NULL, pArray, pTSchema, &row);
tTSRowNew(&rb, pArray, pTSchema, &row);
debugPrintTSRow(row, pTSchema, __func__, __LINE__);
checkTSRow((const char **)&data, row, pTSchema);
tsRowBuilderClear(&rb);
taosArrayDestroy(pArray);
taosMemoryFree(pTSchema);
}
@ -443,12 +442,12 @@ TEST(testCase, NoneTest) {
const char *data[nRows][nCols] = {
{"1653694220000", "no", "20", "10.1", "10.1", "binary10", "no", "10", "10", "nu", "10", "20", "30", "40"},
{"1653694220001", "no", "no", "no", "no", "no", "no", "no", "no", "no", "no", "no", "no", "no"},
{"1653694220002", "no", "no", "no", "no", "no", "nu", "no", "no", "no", "no", "no", "no", "nu"},
{"1653694220003", "nu", "no", "no", "no", "no", "nu", "no", "no", "no", "no", "no", "no", "no"},
{"1653694220002", "10", "no", "no", "no", "no", "no", "no", "no", "no", "no", "no", "no", "no"},
{"1653694220003", "10", "10", "no", "no", "no", "no", "no", "no", "no", "no", "no", "no", "no"},
{"1653694220004", "no", "20", "no", "no", "no", "nchar10", "no", "no", "no", "no", "no", "no", "no"},
{"1653694220005", "nu", "nu", "nu", "nu", "nu", "nu", "nu", "nu", "nu", "nu", "nu", "nu", "nu"},
{"1653694220006", "no", "nu", "nu", "nu", "nu", "nu", "nu", "nu", "nu", "nu", "nu", "nu", "nu"},
{"1653694220007", "no", "nu", "nu", "nu", "nu", "nu", "nu", "nu", "nu", "nu", "nu", "nu", "no"},
{"1653694220007", "no", "nu", "nu", "nu", "nu", "nu", "nu", "nu", "no", "no", "no", "no", "no"},
{"1653694220008", "no", "nu", "nu", "nu", "binary10", "nu", "nu", "nu", "nu", "nu", "nu", "nu", "no"},
{"1653694220009", "no", "nu", "nu", "nu", "binary10", "nu", "nu", "10", "no", "nu", "nu", "nu", "100"},
{"1653694220010", "-1", "-1", "-1", "-1", "binary10", "nu", "-1", "0", "0", "0", "0", "0", "0"},
@ -465,13 +464,12 @@ TEST(testCase, NoneTest) {
{"1653694220019", "no", "9223372036854775807", "nu", "nu", "bin10", "nu", "nu", "10", "no", "254", "nu", "nu",
"no"}};
for (int r = 0; r < nRows; ++r) {
genTestData((const char **)&data[r], nCols, &pArray);
tTSRowNew(NULL, pArray, pTSchema, &row);
debugPrintTSRow(row, pTSchema, __func__, __LINE__); // debug print
checkTSRow((const char **)&data[r], row, pTSchema); // check
taosMemoryFreeClear(row);
tTSRowFree(row);
taosArrayClear(pArray);
}

9
source/dnode/mgmt/mgmt_mnode/src/mmHandle.c
View File

@ -215,8 +215,12 @@ SArray *mmGetMsgHandles() {
if (dmSetMgmtHandle(pArray, TDMT_VND_ALTER_CONFIG_RSP, mmPutNodeMsgToWriteQueue, 0) == NULL) goto _OVER;
if (dmSetMgmtHandle(pArray, TDMT_VND_ALTER_REPLICA_RSP, mmPutNodeMsgToWriteQueue, 0) == NULL) goto _OVER;
if (dmSetMgmtHandle(pArray, TDMT_VND_ALTER_CONFIRM_RSP, mmPutNodeMsgToWriteQueue, 0) == NULL) goto _OVER;
if (dmSetMgmtHandle(pArray, TDMT_VND_ALTER_HASHRANGE_RSP, mmPutNodeMsgToWriteQueue, 0) == NULL) goto _OVER;
if (dmSetMgmtHandle(pArray, TDMT_VND_COMPACT_RSP, mmPutNodeMsgToWriteQueue, 0) == NULL) goto _OVER;
if (dmSetMgmtHandle(pArray, TDMT_MON_MM_INFO, mmPutNodeMsgToMonitorQueue, 0) == NULL) goto _OVER;
if (dmSetMgmtHandle(pArray, TDMT_MON_MM_LOAD, mmPutNodeMsgToMonitorQueue, 0) == NULL) goto _OVER;
if (dmSetMgmtHandle(pArray, TDMT_SYNC_TIMEOUT, mmPutNodeMsgToSyncQueue, 1) == NULL) goto _OVER;
if (dmSetMgmtHandle(pArray, TDMT_SYNC_PING, mmPutNodeMsgToSyncQueue, 1) == NULL) goto _OVER;
if (dmSetMgmtHandle(pArray, TDMT_SYNC_PING_REPLY, mmPutNodeMsgToSyncQueue, 1) == NULL) goto _OVER;
@ -226,9 +230,8 @@ SArray *mmGetMsgHandles() {
if (dmSetMgmtHandle(pArray, TDMT_SYNC_REQUEST_VOTE_REPLY, mmPutNodeMsgToSyncQueue, 1) == NULL) goto _OVER;
if (dmSetMgmtHandle(pArray, TDMT_SYNC_APPEND_ENTRIES, mmPutNodeMsgToSyncQueue, 1) == NULL) goto _OVER;
if (dmSetMgmtHandle(pArray, TDMT_SYNC_APPEND_ENTRIES_REPLY, mmPutNodeMsgToSyncQueue, 1) == NULL) goto _OVER;
if (dmSetMgmtHandle(pArray, TDMT_MON_MM_INFO, mmPutNodeMsgToMonitorQueue, 0) == NULL) goto _OVER;
if (dmSetMgmtHandle(pArray, TDMT_MON_MM_LOAD, mmPutNodeMsgToMonitorQueue, 0) == NULL) goto _OVER;
if (dmSetMgmtHandle(pArray, TDMT_SYNC_SNAPSHOT_SEND, mmPutNodeMsgToSyncQueue, 1) == NULL) goto _OVER;
if (dmSetMgmtHandle(pArray, TDMT_SYNC_SNAPSHOT_RSP, mmPutNodeMsgToSyncQueue, 1) == NULL) goto _OVER;
code = 0;

2
source/dnode/mgmt/mgmt_mnode/src/mmWorker.c
View File

@ -71,7 +71,7 @@ static void mmProcessSyncQueue(SQueueInfo *pInfo, SRpcMsg *pMsg) {
}
static int32_t mmPutNodeMsgToWorker(SSingleWorker *pWorker, SRpcMsg *pMsg) {
dTrace("msg:%p, put into worker %s, type:%s", pMsg, pWorker->name, TMSG_INFO(pMsg->msgType));
dTrace("msg:%p, put into %s queue, type:%s", pMsg, pWorker->name, TMSG_INFO(pMsg->msgType));
taosWriteQitem(pWorker->queue, pMsg);
return 0;
}

1
source/dnode/mgmt/mgmt_vnode/src/vmHandle.c
View File

@ -359,6 +359,7 @@ SArray *vmGetMsgHandles() {
if (dmSetMgmtHandle(pArray, TDMT_VND_ALTER_REPLICA, vmPutMsgToWriteQueue, 0) == NULL) goto _OVER;
if (dmSetMgmtHandle(pArray, TDMT_VND_ALTER_CONFIG, vmPutMsgToWriteQueue, 0) == NULL) goto _OVER;
if (dmSetMgmtHandle(pArray, TDMT_VND_ALTER_CONFIRM, vmPutMsgToWriteQueue, 0) == NULL) goto _OVER;
if (dmSetMgmtHandle(pArray, TDMT_VND_ALTER_HASHRANGE, vmPutMsgToWriteQueue, 0) == NULL) goto _OVER;
if (dmSetMgmtHandle(pArray, TDMT_VND_COMPACT, vmPutMsgToWriteQueue, 0) == NULL) goto _OVER;
if (dmSetMgmtHandle(pArray, TDMT_DND_CREATE_VNODE, vmPutMsgToMgmtQueue, 0) == NULL) goto _OVER;
if (dmSetMgmtHandle(pArray, TDMT_DND_DROP_VNODE, vmPutMsgToMgmtQueue, 0) == NULL) goto _OVER;

3
source/dnode/mnode/impl/inc/mndInt.h
View File

@ -19,6 +19,7 @@
#include "mndDef.h"
#include "sdb.h"
#include "sync.h"
#include "syncTools.h"
#include "tcache.h"
#include "tdatablock.h"
@ -75,7 +76,6 @@ typedef struct {
} STelemMgmt;
typedef struct {
SWal *pWal;
sem_t syncSem;
int64_t sync;
bool standby;
@ -108,6 +108,7 @@ typedef struct SMnode {
SQHandle *pQuery;
SHashObj *infosMeta;
SHashObj *perfsMeta;
SWal *pWal;
SShowMgmt showMgmt;
SProfileMgmt profileMgmt;
STelemMgmt telemMgmt;

1
source/dnode/mnode/impl/inc/mndVgroup.h
View File

@ -41,6 +41,7 @@ int32_t mndAddAlterVnodeAction(SMnode *, STrans *pTrans, SDbObj *pDb, SVgObj *pV
int32_t mndAddDropVnodeAction(SMnode *, STrans *pTrans, SDbObj *pDb, SVgObj *pVgroup, SVnodeGid *pVgid, bool isRedo);
int32_t mndSetMoveVgroupInfoToTrans(SMnode *, STrans *pTrans, SDbObj *pDb, SVgObj *pVgroup, int32_t vn, SArray *pArray);
int32_t mndSetMoveVgroupsInfoToTrans(SMnode *, STrans *pTrans, int32_t dropDnodeId);
int32_t mndBuildAlterVgroupAction(SMnode *pMnode, STrans *pTrans, SDbObj *pDb, SVgObj *pVgroup, SArray *pArray);
void *mndBuildCreateVnodeReq(SMnode *, SDnodeObj *pDnode, SDbObj *pDb, SVgObj *pVgroup, int32_t *cntlen, bool standby);
void *mndBuildDropVnodeReq(SMnode *, SDnodeObj *pDnode, SDbObj *pDb, SVgObj *pVgroup, int32_t *pContLen);

51
source/dnode/mnode/impl/src/mndDb.c
View File

@ -636,57 +636,6 @@ static int32_t mndSetAlterDbCommitLogs(SMnode *pMnode, STrans *pTrans, SDbObj *p
return 0;
}
static int32_t mndBuildAlterVgroupAction(SMnode *pMnode, STrans *pTrans, SDbObj *pDb, SVgObj *pVgroup, SArray *pArray) {
if (pVgroup->replica <= 0 || pVgroup->replica == pDb->cfg.replications) {
if (mndAddAlterVnodeAction(pMnode, pTrans, pDb, pVgroup, TDMT_VND_ALTER_CONFIG) != 0) {
return -1;
}
} else {
SVgObj newVgroup = {0};
memcpy(&newVgroup, pVgroup, sizeof(SVgObj));
mndTransSetSerial(pTrans);
if (newVgroup.replica < pDb->cfg.replications) {
mInfo("db:%s, vgId:%d, vn:0 dnode:%d, will add 2 vnodes", pVgroup->dbName, pVgroup->vgId,
pVgroup->vnodeGid[0].dnodeId);
if (mndAddVnodeToVgroup(pMnode, &newVgroup, pArray) != 0) return -1;
if (mndAddCreateVnodeAction(pMnode, pTrans, pDb, &newVgroup, &newVgroup.vnodeGid[1], true) != 0) return -1;
if (mndAddAlterVnodeAction(pMnode, pTrans, pDb, &newVgroup, TDMT_VND_ALTER_REPLICA) != 0) return -1;
if (mndAddAlterVnodeConfirmAction(pMnode, pTrans, pDb, &newVgroup) != 0) return -1;
if (mndAddVnodeToVgroup(pMnode, &newVgroup, pArray) != 0) return -1;
if (mndAddCreateVnodeAction(pMnode, pTrans, pDb, &newVgroup, &newVgroup.vnodeGid[2], true) != 0) return -1;
if (mndAddAlterVnodeAction(pMnode, pTrans, pDb, &newVgroup, TDMT_VND_ALTER_REPLICA) != 0) return -1;
if (mndAddAlterVnodeConfirmAction(pMnode, pTrans, pDb, &newVgroup) != 0) return -1;
} else {
mInfo("db:%s, vgId:%d, will remove 2 vnodes", pVgroup->dbName, pVgroup->vgId);
SVnodeGid del1 = {0};
if (mndRemoveVnodeFromVgroup(pMnode, &newVgroup, pArray, &del1) != 0) return -1;
if (mndAddAlterVnodeAction(pMnode, pTrans, pDb, &newVgroup, TDMT_VND_ALTER_REPLICA) != 0) return -1;
if (mndAddDropVnodeAction(pMnode, pTrans, pDb, &newVgroup, &del1, true) != 0) return -1;
if (mndAddAlterVnodeConfirmAction(pMnode, pTrans, pDb, &newVgroup) != 0) return -1;
SVnodeGid del2 = {0};
if (mndRemoveVnodeFromVgroup(pMnode, &newVgroup, pArray, &del2) != 0) return -1;
if (mndAddAlterVnodeAction(pMnode, pTrans, pDb, &newVgroup, TDMT_VND_ALTER_REPLICA) != 0) return -1;
if (mndAddDropVnodeAction(pMnode, pTrans, pDb, &newVgroup, &del2, true) != 0) return -1;
if (mndAddAlterVnodeConfirmAction(pMnode, pTrans, pDb, &newVgroup) != 0) return -1;
}
SSdbRaw *pVgRaw = mndVgroupActionEncode(&newVgroup);
if (pVgRaw == NULL) return -1;
if (mndTransAppendCommitlog(pTrans, pVgRaw) != 0) {
sdbFreeRaw(pVgRaw);
return -1;
}
sdbSetRawStatus(pVgRaw, SDB_STATUS_READY);
}
return 0;
}
static int32_t mndSetAlterDbRedoActions(SMnode *pMnode, STrans *pTrans, SDbObj *pOld, SDbObj *pNew) {
SSdb *pSdb = pMnode->pSdb;
void *pIter = NULL;

2
source/dnode/mnode/impl/src/mndDnode.c
View File

@ -566,9 +566,11 @@ static int32_t mndDropDnode(SMnode *pMnode, SRpcMsg *pReq, SDnodeObj *pDnode, SM
pRaw = NULL;
if (pMObj != NULL) {
mDebug("trans:%d, mnode on dnode:%d will be dropped", pTrans->id, pDnode->id);
if (mndSetDropMnodeInfoToTrans(pMnode, pTrans, pMObj) != 0) goto _OVER;
}
if (numOfVnodes > 0) {
mDebug("trans:%d, %d vnodes on dnode:%d will be dropped", pTrans->id, numOfVnodes, pDnode->id);
if (mndSetMoveVgroupsInfoToTrans(pMnode, pTrans, pDnode->id) != 0) goto _OVER;
}
if (mndTransPrepare(pMnode, pTrans) != 0) goto _OVER;

154
source/dnode/mnode/impl/src/mndMain.c
View File

@ -139,10 +139,40 @@ static int32_t mndCreateDir(SMnode *pMnode, const char *path) {
return 0;
}
static int32_t mndInitWal(SMnode *pMnode) {
char path[PATH_MAX + 20] = {0};
snprintf(path, sizeof(path), "%s%swal", pMnode->path, TD_DIRSEP);
SWalCfg cfg = {
.vgId = 1,
.fsyncPeriod = 0,
.rollPeriod = -1,
.segSize = -1,
.retentionPeriod = -1,
.retentionSize = -1,
.level = TAOS_WAL_FSYNC,
};
pMnode->pWal = walOpen(path, &cfg);
if (pMnode->pWal == NULL) {
mError("failed to open wal since %s", terrstr());
return -1;
}
return 0;
}
static void mndCloseWal(SMnode *pMnode) {
if (pMnode->pWal != NULL) {
walClose(pMnode->pWal);
pMnode->pWal = NULL;
}
}
static int32_t mndInitSdb(SMnode *pMnode) {
SSdbOpt opt = {0};
opt.path = pMnode->path;
opt.pMnode = pMnode;
opt.pWal = pMnode->pWal;
pMnode->pSdb = sdbInit(&opt);
if (pMnode->pSdb == NULL) {
@ -156,7 +186,6 @@ static int32_t mndOpenSdb(SMnode *pMnode) {
if (!pMnode->deploy) {
return sdbReadFile(pMnode->pSdb);
} else {
// return sdbDeploy(pMnode->pSdb);;
return 0;
}
}
@ -182,6 +211,7 @@ static int32_t mndAllocStep(SMnode *pMnode, char *name, MndInitFp initFp, MndCle
}
static int32_t mndInitSteps(SMnode *pMnode) {
if (mndAllocStep(pMnode, "mnode-wal", mndInitWal, mndCloseWal) != 0) return -1;
if (mndAllocStep(pMnode, "mnode-sdb", mndInitSdb, mndCleanupSdb) != 0) return -1;
if (mndAllocStep(pMnode, "mnode-trans", mndInitTrans, mndCleanupTrans) != 0) return -1;
if (mndAllocStep(pMnode, "mnode-cluster", mndInitCluster, mndCleanupCluster) != 0) return -1;
@ -201,7 +231,7 @@ static int32_t mndInitSteps(SMnode *pMnode) {
if (mndAllocStep(pMnode, "mnode-offset", mndInitOffset, mndCleanupOffset) != 0) return -1;
if (mndAllocStep(pMnode, "mnode-vgroup", mndInitVgroup, mndCleanupVgroup) != 0) return -1;
if (mndAllocStep(pMnode, "mnode-stb", mndInitStb, mndCleanupStb) != 0) return -1;
if (mndAllocStep(pMnode, "mnode-stb", mndInitSma, mndCleanupSma) != 0) return -1;
if (mndAllocStep(pMnode, "mnode-sma", mndInitSma, mndCleanupSma) != 0) return -1;
if (mndAllocStep(pMnode, "mnode-infos", mndInitInfos, mndCleanupInfos) != 0) return -1;
if (mndAllocStep(pMnode, "mnode-perfs", mndInitPerfs, mndCleanupPerfs) != 0) return -1;
if (mndAllocStep(pMnode, "mnode-db", mndInitDb, mndCleanupDb) != 0) return -1;
@ -376,41 +406,93 @@ int32_t mndProcessSyncMsg(SRpcMsg *pMsg) {
syncRpcMsgLog2(logBuf, pMsg);
taosMemoryFree(syncNodeStr);
if (pMsg->msgType == TDMT_SYNC_TIMEOUT) {
SyncTimeout *pSyncMsg = syncTimeoutFromRpcMsg2(pMsg);
code = syncNodeOnTimeoutCb(pSyncNode, pSyncMsg);
syncTimeoutDestroy(pSyncMsg);
} else if (pMsg->msgType == TDMT_SYNC_PING) {
SyncPing *pSyncMsg = syncPingFromRpcMsg2(pMsg);
code = syncNodeOnPingCb(pSyncNode, pSyncMsg);
syncPingDestroy(pSyncMsg);
} else if (pMsg->msgType == TDMT_SYNC_PING_REPLY) {
SyncPingReply *pSyncMsg = syncPingReplyFromRpcMsg2(pMsg);
code = syncNodeOnPingReplyCb(pSyncNode, pSyncMsg);
syncPingReplyDestroy(pSyncMsg);
} else if (pMsg->msgType == TDMT_SYNC_CLIENT_REQUEST) {
SyncClientRequest *pSyncMsg = syncClientRequestFromRpcMsg2(pMsg);
code = syncNodeOnClientRequestCb(pSyncNode, pSyncMsg);
syncClientRequestDestroy(pSyncMsg);
} else if (pMsg->msgType == TDMT_SYNC_REQUEST_VOTE) {
SyncRequestVote *pSyncMsg = syncRequestVoteFromRpcMsg2(pMsg);
code = syncNodeOnRequestVoteCb(pSyncNode, pSyncMsg);
syncRequestVoteDestroy(pSyncMsg);
} else if (pMsg->msgType == TDMT_SYNC_REQUEST_VOTE_REPLY) {
SyncRequestVoteReply *pSyncMsg = syncRequestVoteReplyFromRpcMsg2(pMsg);
code = syncNodeOnRequestVoteReplyCb(pSyncNode, pSyncMsg);
syncRequestVoteReplyDestroy(pSyncMsg);
} else if (pMsg->msgType == TDMT_SYNC_APPEND_ENTRIES) {
SyncAppendEntries *pSyncMsg = syncAppendEntriesFromRpcMsg2(pMsg);
code = syncNodeOnAppendEntriesCb(pSyncNode, pSyncMsg);
syncAppendEntriesDestroy(pSyncMsg);
} else if (pMsg->msgType == TDMT_SYNC_APPEND_ENTRIES_REPLY) {
SyncAppendEntriesReply *pSyncMsg = syncAppendEntriesReplyFromRpcMsg2(pMsg);
code = syncNodeOnAppendEntriesReplyCb(pSyncNode, pSyncMsg);
syncAppendEntriesReplyDestroy(pSyncMsg);
// ToDo: ugly! use function pointer
if (syncNodeSnapshotEnable(pSyncNode)) {
if (pMsg->msgType == TDMT_SYNC_TIMEOUT) {
SyncTimeout *pSyncMsg = syncTimeoutFromRpcMsg2(pMsg);
code = syncNodeOnTimeoutCb(pSyncNode, pSyncMsg);
syncTimeoutDestroy(pSyncMsg);
} else if (pMsg->msgType == TDMT_SYNC_PING) {
SyncPing *pSyncMsg = syncPingFromRpcMsg2(pMsg);
code = syncNodeOnPingCb(pSyncNode, pSyncMsg);
syncPingDestroy(pSyncMsg);
} else if (pMsg->msgType == TDMT_SYNC_PING_REPLY) {
SyncPingReply *pSyncMsg = syncPingReplyFromRpcMsg2(pMsg);
code = syncNodeOnPingReplyCb(pSyncNode, pSyncMsg);
syncPingReplyDestroy(pSyncMsg);
} else if (pMsg->msgType == TDMT_SYNC_CLIENT_REQUEST) {
SyncClientRequest *pSyncMsg = syncClientRequestFromRpcMsg2(pMsg);
code = syncNodeOnClientRequestCb(pSyncNode, pSyncMsg);
syncClientRequestDestroy(pSyncMsg);
} else if (pMsg->msgType == TDMT_SYNC_REQUEST_VOTE) {
SyncRequestVote *pSyncMsg = syncRequestVoteFromRpcMsg2(pMsg);
code = syncNodeOnRequestVoteSnapshotCb(pSyncNode, pSyncMsg);
syncRequestVoteDestroy(pSyncMsg);
} else if (pMsg->msgType == TDMT_SYNC_REQUEST_VOTE_REPLY) {
SyncRequestVoteReply *pSyncMsg = syncRequestVoteReplyFromRpcMsg2(pMsg);
code = syncNodeOnRequestVoteReplySnapshotCb(pSyncNode, pSyncMsg);
syncRequestVoteReplyDestroy(pSyncMsg);
} else if (pMsg->msgType == TDMT_SYNC_APPEND_ENTRIES) {
SyncAppendEntries *pSyncMsg = syncAppendEntriesFromRpcMsg2(pMsg);
code = syncNodeOnAppendEntriesSnapshotCb(pSyncNode, pSyncMsg);
syncAppendEntriesDestroy(pSyncMsg);
} else if (pMsg->msgType == TDMT_SYNC_APPEND_ENTRIES_REPLY) {
SyncAppendEntriesReply *pSyncMsg = syncAppendEntriesReplyFromRpcMsg2(pMsg);
code = syncNodeOnAppendEntriesReplySnapshotCb(pSyncNode, pSyncMsg);
syncAppendEntriesReplyDestroy(pSyncMsg);
} else if (pMsg->msgType == TDMT_SYNC_SNAPSHOT_SEND) {
SyncSnapshotSend *pSyncMsg = syncSnapshotSendFromRpcMsg2(pMsg);
code = syncNodeOnSnapshotSendCb(pSyncNode, pSyncMsg);
syncSnapshotSendDestroy(pSyncMsg);
} else if (pMsg->msgType == TDMT_SYNC_SNAPSHOT_RSP) {
SyncSnapshotRsp *pSyncMsg = syncSnapshotRspFromRpcMsg2(pMsg);
code = syncNodeOnSnapshotRspCb(pSyncNode, pSyncMsg);
syncSnapshotRspDestroy(pSyncMsg);
} else {
mError("failed to process msg:%p since invalid type:%s", pMsg, TMSG_INFO(pMsg->msgType));
code = TAOS_SYNC_PROPOSE_OTHER_ERROR;
}
} else {
mError("failed to process msg:%p since invalid type:%s", pMsg, TMSG_INFO(pMsg->msgType));
code = TAOS_SYNC_PROPOSE_OTHER_ERROR;
if (pMsg->msgType == TDMT_SYNC_TIMEOUT) {
SyncTimeout *pSyncMsg = syncTimeoutFromRpcMsg2(pMsg);
code = syncNodeOnTimeoutCb(pSyncNode, pSyncMsg);
syncTimeoutDestroy(pSyncMsg);
} else if (pMsg->msgType == TDMT_SYNC_PING) {
SyncPing *pSyncMsg = syncPingFromRpcMsg2(pMsg);
code = syncNodeOnPingCb(pSyncNode, pSyncMsg);
syncPingDestroy(pSyncMsg);
} else if (pMsg->msgType == TDMT_SYNC_PING_REPLY) {
SyncPingReply *pSyncMsg = syncPingReplyFromRpcMsg2(pMsg);
code = syncNodeOnPingReplyCb(pSyncNode, pSyncMsg);
syncPingReplyDestroy(pSyncMsg);
} else if (pMsg->msgType == TDMT_SYNC_CLIENT_REQUEST) {
SyncClientRequest *pSyncMsg = syncClientRequestFromRpcMsg2(pMsg);
code = syncNodeOnClientRequestCb(pSyncNode, pSyncMsg);
syncClientRequestDestroy(pSyncMsg);
} else if (pMsg->msgType == TDMT_SYNC_REQUEST_VOTE) {
SyncRequestVote *pSyncMsg = syncRequestVoteFromRpcMsg2(pMsg);
code = syncNodeOnRequestVoteCb(pSyncNode, pSyncMsg);
syncRequestVoteDestroy(pSyncMsg);
} else if (pMsg->msgType == TDMT_SYNC_REQUEST_VOTE_REPLY) {
SyncRequestVoteReply *pSyncMsg = syncRequestVoteReplyFromRpcMsg2(pMsg);
code = syncNodeOnRequestVoteReplyCb(pSyncNode, pSyncMsg);
syncRequestVoteReplyDestroy(pSyncMsg);
} else if (pMsg->msgType == TDMT_SYNC_APPEND_ENTRIES) {
SyncAppendEntries *pSyncMsg = syncAppendEntriesFromRpcMsg2(pMsg);
code = syncNodeOnAppendEntriesCb(pSyncNode, pSyncMsg);
syncAppendEntriesDestroy(pSyncMsg);
} else if (pMsg->msgType == TDMT_SYNC_APPEND_ENTRIES_REPLY) {
SyncAppendEntriesReply *pSyncMsg = syncAppendEntriesReplyFromRpcMsg2(pMsg);
code = syncNodeOnAppendEntriesReplyCb(pSyncNode, pSyncMsg);
syncAppendEntriesReplyDestroy(pSyncMsg);
} else {
mError("failed to process msg:%p since invalid type:%s", pMsg, TMSG_INFO(pMsg->msgType));
code = TAOS_SYNC_PROPOSE_OTHER_ERROR;
}
}
mndReleaseSyncRef(pMnode);

21
source/dnode/mnode/impl/src/mndScheduler.c
View File

@ -35,6 +35,13 @@
extern bool tsStreamSchedV;
static int32_t mndAddTaskToTaskSet(SArray* pArray, SStreamTask* pTask) {
int32_t childId = taosArrayGetSize(pArray);
pTask->childId = childId;
taosArrayPush(pArray, &pTask);
return 0;
}
int32_t mndConvertRSmaTask(const char* ast, int64_t uid, int8_t triggerType, int64_t watermark, char** pStr,
int32_t* pLen, double filesFactor) {
SNode* pAst = NULL;
@ -186,7 +193,7 @@ int32_t mndAddShuffledSinkToStream(SMnode* pMnode, STrans* pTrans, SStreamObj* p
SVgObj* pVgroup;
pIter = sdbFetch(pSdb, SDB_VGROUP, pIter, (void**)&pVgroup);
if (pIter == NULL) break;
if (pVgroup->dbUid != pStream->dbUid) {
if (strcmp(pVgroup->dbName, pStream->targetDb) != 0) {
sdbRelease(pSdb, pVgroup);
continue;
}
@ -195,7 +202,7 @@ int32_t mndAddShuffledSinkToStream(SMnode* pMnode, STrans* pTrans, SStreamObj* p
terrno = TSDB_CODE_OUT_OF_MEMORY;
return -1;
}
taosArrayPush(tasks, &pTask);
mndAddTaskToTaskSet(tasks, pTask);
pTask->nodeId = pVgroup->vgId;
pTask->epSet = mndGetVgroupEpset(pMnode, pVgroup);
@ -235,7 +242,7 @@ int32_t mndAddFixedSinkToStream(SMnode* pMnode, STrans* pTrans, SStreamObj* pStr
terrno = TSDB_CODE_OUT_OF_MEMORY;
return -1;
}
taosArrayPush(tasks, &pTask);
mndAddTaskToTaskSet(tasks, pTask);
pTask->nodeId = pStream->fixedSinkVgId;
#if 0
@ -286,7 +293,8 @@ int32_t mndScheduleStream(SMnode* pMnode, STrans* pTrans, SStreamObj* pStream) {
pStream->tasks = taosArrayInit(totLevel, sizeof(void*));
bool hasExtraSink = false;
if (totLevel == 2 || strcmp(pStream->sourceDb, pStream->targetDb) != 0) {
bool externalTargetDB = strcmp(pStream->sourceDb, pStream->targetDb) != 0;
if (totLevel == 2 || externalTargetDB) {
SArray* taskOneLevel = taosArrayInit(0, sizeof(void*));
taosArrayPush(pStream->tasks, &taskOneLevel);
// add extra sink
@ -377,7 +385,7 @@ int32_t mndScheduleStream(SMnode* pMnode, STrans* pTrans, SStreamObj* pStream) {
return -1;
}
sdbRelease(pSdb, pVgroup);
taosArrayPush(taskOneLevel, &pTask);
mndAddTaskToTaskSet(taskOneLevel, pTask);
}
} else {
// merge plan
@ -405,7 +413,6 @@ int32_t mndScheduleStream(SMnode* pMnode, STrans* pTrans, SStreamObj* pStream) {
if (pStream->fixedSinkVgId == 0) {
pTask->dispatchType = TASK_DISPATCH__SHUFFLE;
/*pTask->dispatchMsgType = TDMT_VND_TASK_WRITE_EXEC;*/
pTask->dispatchMsgType = TDMT_STREAM_TASK_DISPATCH;
SDbObj* pDb = mndAcquireDb(pMnode, pStream->targetDb);
ASSERT(pDb);
@ -426,10 +433,8 @@ int32_t mndScheduleStream(SMnode* pMnode, STrans* pTrans, SStreamObj* pStream) {
SVgroupInfo* pVgInfo = taosArrayGet(pVgs, i);
for (int32_t j = 0; j < sinkLvSize; j++) {
SStreamTask* pLastLevelTask = taosArrayGetP(sinkLv, j);
/*printf("vgid %d node id %d\n", pVgInfo->vgId, pTask->nodeId);*/
if (pLastLevelTask->nodeId == pVgInfo->vgId) {
pVgInfo->taskId = pLastLevelTask->taskId;
/*printf("taskid %d set to %d\n", pVgInfo->taskId, pTask->taskId);*/
break;
}
}

64
source/dnode/mnode/impl/src/mndSync.c
View File

@ -17,15 +17,27 @@
#include "mndSync.h"
#include "mndTrans.h"
int32_t mndSyncEqMsg(const SMsgCb *msgcb, SRpcMsg *pMsg) {
static int32_t mndSyncEqMsg(const SMsgCb *msgcb, SRpcMsg *pMsg) {
SMsgHead *pHead = pMsg->pCont;
pHead->contLen = htonl(pHead->contLen);
pHead->vgId = htonl(pHead->vgId);
return tmsgPutToQueue(msgcb, SYNC_QUEUE, pMsg);
int32_t code = tmsgPutToQueue(msgcb, SYNC_QUEUE, pMsg);
if (code != 0) {
rpcFreeCont(pMsg->pCont);
pMsg->pCont = NULL;
}
return code;
}
int32_t mndSyncSendMsg(const SEpSet *pEpSet, SRpcMsg *pMsg) { return tmsgSendReq(pEpSet, pMsg); }
static int32_t mndSyncSendMsg(const SEpSet *pEpSet, SRpcMsg *pMsg) {
int32_t code = tmsgSendReq(pEpSet, pMsg);
if (code != 0) {
rpcFreeCont(pMsg->pCont);
pMsg->pCont = NULL;
}
return code;
}
void mndSyncCommitMsg(struct SSyncFSM *pFsm, const SRpcMsg *pMsg, SFsmCbMeta cbMeta) {
SMnode *pMnode = pFsm->data;
@ -34,7 +46,7 @@ void mndSyncCommitMsg(struct SSyncFSM *pFsm, const SRpcMsg *pMsg, SFsmCbMeta cbM
int32_t transId = sdbGetIdFromRaw(pMnode->pSdb, pRaw);
pMgmt->errCode = cbMeta.code;
mTrace("trans:%d, is proposed, savedTransId:%d code:0x%x, ver:%" PRId64 " term:%" PRId64 " role:%s raw:%p", transId,
mDebug("trans:%d, is proposed, saved:%d code:0x%x, index:%" PRId64 " term:%" PRId64 " role:%s raw:%p", transId,
pMgmt->transId, cbMeta.code, cbMeta.index, cbMeta.term, syncStr(cbMeta.state), pRaw);
if (pMgmt->errCode == 0) {
@ -50,6 +62,10 @@ void mndSyncCommitMsg(struct SSyncFSM *pFsm, const SRpcMsg *pMsg, SFsmCbMeta cbM
tsem_post(&pMgmt->syncSem);
} else {
if (cbMeta.index - sdbGetApplyIndex(pMnode->pSdb) > 100) {
SSnapshotMeta sMeta = {0};
if (syncGetSnapshotMeta(pMnode->syncMgmt.sync, &sMeta) == 0) {
sdbSetCurConfig(pMnode->pSdb, sMeta.lastConfigIndex);
}
sdbWriteFile(pMnode->pSdb);
}
}
@ -57,13 +73,20 @@ void mndSyncCommitMsg(struct SSyncFSM *pFsm, const SRpcMsg *pMsg, SFsmCbMeta cbM
int32_t mndSyncGetSnapshot(struct SSyncFSM *pFsm, SSnapshot *pSnapshot) {
SMnode *pMnode = pFsm->data;
pSnapshot->lastApplyIndex = sdbGetApplyIndex(pMnode->pSdb);
pSnapshot->lastApplyTerm = sdbGetApplyTerm(pMnode->pSdb);
pSnapshot->lastApplyIndex = sdbGetCommitIndex(pMnode->pSdb);
pSnapshot->lastApplyTerm = sdbGetCommitTerm(pMnode->pSdb);
pSnapshot->lastConfigIndex = sdbGetCurConfig(pMnode->pSdb);
return 0;
}
void mndRestoreFinish(struct SSyncFSM *pFsm) {
SMnode *pMnode = pFsm->data;
SSnapshotMeta sMeta = {0};
if (syncGetSnapshotMeta(pMnode->syncMgmt.sync, &sMeta) == 0) {
sdbSetCurConfig(pMnode->pSdb, sMeta.lastConfigIndex);
}
if (!pMnode->deploy) {
mInfo("mnode sync restore finished, and will handle outstanding transactions");
mndTransPullup(pMnode);
@ -78,8 +101,8 @@ void mndReConfig(struct SSyncFSM *pFsm, SSyncCfg newCfg, SReConfigCbMeta cbMeta)
SSyncMgmt *pMgmt = &pMnode->syncMgmt;
pMgmt->errCode = cbMeta.code;
mInfo("trans:-1, sync reconfig is proposed, savedTransId:%d code:0x%x, curTerm:%" PRId64 " term:%" PRId64,
pMgmt->transId, cbMeta.code, cbMeta.index, cbMeta.term);
mInfo("trans:-1, sync reconfig is proposed, saved:%d code:0x%x, index:%" PRId64 " term:%" PRId64, pMgmt->transId,
cbMeta.code, cbMeta.index, cbMeta.term);
if (pMgmt->transId == -1) {
if (pMgmt->errCode != 0) {
@ -144,29 +167,12 @@ SSyncFSM *mndSyncMakeFsm(SMnode *pMnode) {
int32_t mndInitSync(SMnode *pMnode) {
SSyncMgmt *pMgmt = &pMnode->syncMgmt;
char path[PATH_MAX + 20] = {0};
snprintf(path, sizeof(path), "%s%swal", pMnode->path, TD_DIRSEP);
SWalCfg cfg = {
.vgId = 1,
.fsyncPeriod = 0,
.rollPeriod = -1,
.segSize = -1,
.retentionPeriod = -1,
.retentionSize = -1,
.level = TAOS_WAL_FSYNC,
};
pMgmt->pWal = walOpen(path, &cfg);
if (pMgmt->pWal == NULL) {
mError("failed to open wal since %s", terrstr());
return -1;
}
SSyncInfo syncInfo = {.vgId = 1, .FpSendMsg = mndSyncSendMsg, .FpEqMsg = mndSyncEqMsg};
snprintf(syncInfo.path, sizeof(syncInfo.path), "%s%ssync", pMnode->path, TD_DIRSEP);
syncInfo.pWal = pMgmt->pWal;
syncInfo.pWal = pMnode->pWal;
syncInfo.pFsm = mndSyncMakeFsm(pMnode);
syncInfo.isStandBy = pMgmt->standby;
syncInfo.snapshotEnable = true;
SSyncCfg *pCfg = &syncInfo.syncCfg;
pCfg->replicaNum = pMnode->replica;
@ -196,10 +202,6 @@ void mndCleanupSync(SMnode *pMnode) {
mDebug("mnode sync is stopped, id:%" PRId64, pMgmt->sync);
tsem_destroy(&pMgmt->syncSem);
if (pMgmt->pWal != NULL) {
walClose(pMgmt->pWal);
}
memset(pMgmt, 0, sizeof(SSyncMgmt));
}

8
source/dnode/mnode/impl/src/mndTrans.c
View File

@ -922,7 +922,7 @@ static int32_t mndTransSendSingleMsg(SMnode *pMnode, STrans *pTrans, STransActio
char detail[1024] = {0};
int32_t len = snprintf(detail, sizeof(detail), "msgType:%s numOfEps:%d inUse:%d", TMSG_INFO(pAction->msgType),
pAction->epSet.numOfEps, pAction->epSet.inUse);
for (int32_t i = 0; i < pTrans->lastErrorEpset.numOfEps; ++i) {
for (int32_t i = 0; i < pAction->epSet.numOfEps; ++i) {
len += snprintf(detail + len, sizeof(detail) - len, " ep:%d-%s:%u", i, pAction->epSet.eps[i].fqdn,
pAction->epSet.eps[i].port);
}
@ -1085,6 +1085,8 @@ static int32_t mndTransExecuteRedoActionsSerial(SMnode *pMnode, STrans *pTrans)
}
if (code == 0) {
if (!pMnode->deploy && !mndIsMaster(pMnode)) break;
pTrans->code = 0;
pTrans->redoActionPos++;
mDebug("trans:%d, %s:%d is executed and need sync to other mnodes", pTrans->id, mndTransStr(pAction->stage),
@ -1386,6 +1388,10 @@ void mndTransPullup(SMnode *pMnode) {
mndReleaseTrans(pMnode, pTrans);
}
SSnapshotMeta sMeta = {0};
if (syncGetSnapshotMeta(pMnode->syncMgmt.sync, &sMeta) == 0) {
sdbSetCurConfig(pMnode->pSdb, sMeta.lastConfigIndex);
}
sdbWriteFile(pMnode->pSdb);
taosArrayDestroy(pArray);
}

151
source/dnode/mnode/impl/src/mndVgroup.c
View File

@ -55,6 +55,7 @@ int32_t mndInitVgroup(SMnode *pMnode) {
mndSetMsgHandle(pMnode, TDMT_VND_ALTER_REPLICA_RSP, mndTransProcessRsp);
mndSetMsgHandle(pMnode, TDMT_VND_ALTER_CONFIG_RSP, mndTransProcessRsp);
mndSetMsgHandle(pMnode, TDMT_VND_ALTER_CONFIRM_RSP, mndTransProcessRsp);
mndSetMsgHandle(pMnode, TDMT_VND_ALTER_HASHRANGE_RSP, mndTransProcessRsp);
mndSetMsgHandle(pMnode, TDMT_DND_DROP_VNODE_RSP, mndTransProcessRsp);
mndSetMsgHandle(pMnode, TDMT_VND_COMPACT_RSP, mndTransProcessRsp);
@ -1166,7 +1167,155 @@ _OVER:
return code;
}
static int32_t mndProcessSplitVgroupMsg(SRpcMsg *pReq) { return 0; }
int32_t mndBuildAlterVgroupAction(SMnode *pMnode, STrans *pTrans, SDbObj *pDb, SVgObj *pVgroup, SArray *pArray) {
if (pVgroup->replica <= 0 || pVgroup->replica == pDb->cfg.replications) {
if (mndAddAlterVnodeAction(pMnode, pTrans, pDb, pVgroup, TDMT_VND_ALTER_CONFIG) != 0) {
return -1;
}
} else {
SVgObj newVgroup = {0};
memcpy(&newVgroup, pVgroup, sizeof(SVgObj));
mndTransSetSerial(pTrans);
if (newVgroup.replica < pDb->cfg.replications) {
mInfo("db:%s, vgId:%d, vn:0 dnode:%d, will add 2 vnodes", pVgroup->dbName, pVgroup->vgId,
pVgroup->vnodeGid[0].dnodeId);
if (mndAddVnodeToVgroup(pMnode, &newVgroup, pArray) != 0) return -1;
if (mndAddCreateVnodeAction(pMnode, pTrans, pDb, &newVgroup, &newVgroup.vnodeGid[1], true) != 0) return -1;
if (mndAddAlterVnodeAction(pMnode, pTrans, pDb, &newVgroup, TDMT_VND_ALTER_REPLICA) != 0) return -1;
if (mndAddAlterVnodeConfirmAction(pMnode, pTrans, pDb, &newVgroup) != 0) return -1;
if (mndAddVnodeToVgroup(pMnode, &newVgroup, pArray) != 0) return -1;
if (mndAddCreateVnodeAction(pMnode, pTrans, pDb, &newVgroup, &newVgroup.vnodeGid[2], true) != 0) return -1;
if (mndAddAlterVnodeAction(pMnode, pTrans, pDb, &newVgroup, TDMT_VND_ALTER_REPLICA) != 0) return -1;
if (mndAddAlterVnodeConfirmAction(pMnode, pTrans, pDb, &newVgroup) != 0) return -1;
} else if (newVgroup.replica > pDb->cfg.replications) {
mInfo("db:%s, vgId:%d, will remove 2 vnodes", pVgroup->dbName, pVgroup->vgId);
SVnodeGid del1 = {0};
if (mndRemoveVnodeFromVgroup(pMnode, &newVgroup, pArray, &del1) != 0) return -1;
if (mndAddAlterVnodeAction(pMnode, pTrans, pDb, &newVgroup, TDMT_VND_ALTER_REPLICA) != 0) return -1;
if (mndAddDropVnodeAction(pMnode, pTrans, pDb, &newVgroup, &del1, true) != 0) return -1;
if (mndAddAlterVnodeConfirmAction(pMnode, pTrans, pDb, &newVgroup) != 0) return -1;
SVnodeGid del2 = {0};
if (mndRemoveVnodeFromVgroup(pMnode, &newVgroup, pArray, &del2) != 0) return -1;
if (mndAddAlterVnodeAction(pMnode, pTrans, pDb, &newVgroup, TDMT_VND_ALTER_REPLICA) != 0) return -1;
if (mndAddDropVnodeAction(pMnode, pTrans, pDb, &newVgroup, &del2, true) != 0) return -1;
if (mndAddAlterVnodeConfirmAction(pMnode, pTrans, pDb, &newVgroup) != 0) return -1;
} else {
}
SSdbRaw *pVgRaw = mndVgroupActionEncode(&newVgroup);
if (pVgRaw == NULL) return -1;
if (mndTransAppendCommitlog(pTrans, pVgRaw) != 0) {
sdbFreeRaw(pVgRaw);
return -1;
}
sdbSetRawStatus(pVgRaw, SDB_STATUS_READY);
}
return 0;
}
static int32_t mndAddAdjustVnodeHashRangeAction(SMnode *pMnode, STrans *pTrans, SDbObj *pDb, SVgObj *pVgroup) {
return 0;
}
static int32_t mndSplitVgroup(SMnode *pMnode, SRpcMsg *pReq, SDbObj *pDb, SVgObj *pVgroup) {
int32_t code = -1;
SSdbRaw *pRaw = NULL;
STrans *pTrans = NULL;
SArray *pArray = mndBuildDnodesArray(pMnode, 0);
pTrans = mndTransCreate(pMnode, TRN_POLICY_RETRY, TRN_CONFLICT_GLOBAL, pReq);
if (pTrans == NULL) goto _OVER;
mndTransSetSerial(pTrans);
mDebug("trans:%d, used to split vgroup, vgId:%d", pTrans->id, pVgroup->vgId);
SVgObj newVg1 = {0};
memcpy(&newVg1, pVgroup, sizeof(SVgObj));
mInfo("vgId:%d, vgroup info before split, replica:%d hashBegin:%u hashEnd:%u", newVg1.vgId, newVg1.replica,
newVg1.hashBegin, newVg1.hashEnd);
for (int32_t i = 0; i < newVg1.replica; ++i) {
mInfo("vgId:%d, vnode:%d dnode:%d", newVg1.vgId, i, newVg1.vnodeGid[i].dnodeId);
}
if (newVg1.replica == 1) {
if (mndAddVnodeToVgroup(pMnode, &newVg1, pArray) != 0) goto _OVER;
if (mndAddCreateVnodeAction(pMnode, pTrans, pDb, &newVg1, &newVg1.vnodeGid[1], true) != 0) goto _OVER;
if (mndAddAlterVnodeAction(pMnode, pTrans, pDb, &newVg1, TDMT_VND_ALTER_REPLICA) != 0) goto _OVER;
if (mndAddAlterVnodeConfirmAction(pMnode, pTrans, pDb, &newVg1) != 0) goto _OVER;
} else if (newVg1.replica == 3) {
SVnodeGid del1 = {0};
if (mndRemoveVnodeFromVgroup(pMnode, &newVg1, pArray, &del1) != 0) goto _OVER;
if (mndAddAlterVnodeAction(pMnode, pTrans, pDb, &newVg1, TDMT_VND_ALTER_REPLICA) != 0) goto _OVER;
if (mndAddDropVnodeAction(pMnode, pTrans, pDb, &newVg1, &del1, true) != 0) goto _OVER;
if (mndAddAlterVnodeConfirmAction(pMnode, pTrans, pDb, &newVg1) != 0) goto _OVER;
} else {
goto _OVER;
}
SVgObj newVg2 = {0};
memcpy(&newVg1, &newVg2, sizeof(SVgObj));
newVg1.replica = 1;
newVg1.hashEnd = (newVg1.hashBegin + newVg1.hashEnd) / 2;
memset(&newVg1.vnodeGid[1], 0, sizeof(SVnodeGid));
newVg2.replica = 1;
newVg2.hashBegin = newVg1.hashEnd + 1;
memcpy(&newVg2.vnodeGid[0], &newVg2.vnodeGid[1], sizeof(SVnodeGid));
memset(&newVg1.vnodeGid[1], 0, sizeof(SVnodeGid));
if (mndAddAlterVnodeAction(pMnode, pTrans, pDb, &newVg1, TDMT_VND_ALTER_HASHRANGE) != 0) goto _OVER;
if (mndAddAlterVnodeAction(pMnode, pTrans, pDb, &newVg2, TDMT_VND_ALTER_HASHRANGE) != 0) goto _OVER;
// adjust vgroup
if (mndBuildAlterVgroupAction(pMnode, pTrans, pDb, &newVg1, pArray) != 0) goto _OVER;
if (mndBuildAlterVgroupAction(pMnode, pTrans, pDb, &newVg2, pArray) != 0) goto _OVER;
_OVER:
mndTransDrop(pTrans);
sdbFreeRaw(pRaw);
return code;
}
static int32_t mndProcessSplitVgroupMsg(SRpcMsg *pReq) {
SMnode *pMnode = pReq->info.node;
int32_t code = -1;
int32_t vgId = 2;
SUserObj *pUser = NULL;
SVgObj *pVgroup = NULL;
SDbObj *pDb = NULL;
mDebug("vgId:%d, start to split", vgId);
pVgroup = mndAcquireVgroup(pMnode, vgId);
if (pVgroup == NULL) goto _OVER;
pDb = mndAcquireDb(pMnode, pVgroup->dbName);
if (pDb == NULL) goto _OVER;
pUser = mndAcquireUser(pMnode, pReq->conn.user);
if (pUser == NULL) {
terrno = TSDB_CODE_MND_NO_USER_FROM_CONN;
goto _OVER;
}
if (mndCheckNodeAuth(pUser) != 0) {
goto _OVER;
}
code = mndSplitVgroup(pMnode, pReq, pDb, pVgroup);
if (code == 0) code = TSDB_CODE_ACTION_IN_PROGRESS;
_OVER:
mndReleaseUser(pMnode, pUser);
mndReleaseVgroup(pMnode, pVgroup);
mndReleaseDb(pMnode, pDb);
return code;
}
static int32_t mndSetBalanceVgroupInfoToTrans(SMnode *pMnode, STrans *pTrans, SDbObj *pDb, SVgObj *pVgroup,
SDnodeObj *pSrc, SDnodeObj *pDst) {

10
source/dnode/mnode/sdb/inc/sdb.h
View File

@ -22,6 +22,7 @@
#include "tlockfree.h"
#include "tlog.h"
#include "tmsg.h"
#include "wal.h"
#ifdef __cplusplus
extern "C" {
@ -165,12 +166,14 @@ typedef struct SSdbRow {
typedef struct SSdb {
SMnode *pMnode;
SWal *pWal;
char *currDir;
char *tmpDir;
int64_t lastCommitVer;
int64_t lastCommitTerm;
int64_t curVer;
int64_t curTerm;
int64_t curConfig;
int64_t tableVer[SDB_MAX];
int64_t maxId[SDB_MAX];
EKeyType keyTypes[SDB_MAX];
@ -205,6 +208,7 @@ typedef struct {
typedef struct SSdbOpt {
const char *path;
SMnode *pMnode;
SWal *pWal;
} SSdbOpt;
/**
@ -358,9 +362,13 @@ int64_t sdbGetTableVer(SSdb *pSdb, ESdbType type);
* @return int32_t The current index of sdb
*/
void sdbSetApplyIndex(SSdb *pSdb, int64_t index);
int64_t sdbGetApplyIndex(SSdb *pSdb);
void sdbSetApplyTerm(SSdb *pSdb, int64_t term);
void sdbSetCurConfig(SSdb *pSdb, int64_t config);
int64_t sdbGetApplyIndex(SSdb *pSdb);
int64_t sdbGetApplyTerm(SSdb *pSdb);
int64_t sdbGetCommitIndex(SSdb *pSdb);
int64_t sdbGetCommitTerm(SSdb *pSdb);
int64_t sdbGetCurConfig(SSdb *pSdb);
SSdbRaw *sdbAllocRaw(ESdbType type, int8_t sver, int32_t dataLen);
void sdbFreeRaw(SSdbRaw *pRaw);

14
source/dnode/mnode/sdb/src/sdb.c
View File

@ -52,10 +52,12 @@ SSdb *sdbInit(SSdbOpt *pOption) {
pSdb->keyTypes[i] = SDB_KEY_INT32;
}
pSdb->pWal = pOption->pWal;
pSdb->curVer = -1;
pSdb->curTerm = -1;
pSdb->lastCommitVer = -1;
pSdb->lastCommitTerm = -1;
pSdb->curConfig = -1;
pSdb->pMnode = pOption->pMnode;
taosThreadMutexInit(&pSdb->filelock, NULL);
mDebug("sdb init successfully");
@ -159,8 +161,16 @@ static int32_t sdbCreateDir(SSdb *pSdb) {
void sdbSetApplyIndex(SSdb *pSdb, int64_t index) { pSdb->curVer = index; }
int64_t sdbGetApplyIndex(SSdb *pSdb) { return pSdb->curVer; }
void sdbSetApplyTerm(SSdb *pSdb, int64_t term) { pSdb->curTerm = term; }
void sdbSetCurConfig(SSdb *pSdb, int64_t config) { pSdb->curConfig = config; }
int64_t sdbGetApplyIndex(SSdb *pSdb) { return pSdb->curVer; }
int64_t sdbGetApplyTerm(SSdb *pSdb) { return pSdb->curTerm; }
int64_t sdbGetCommitIndex(SSdb *pSdb) { return pSdb->lastCommitVer; }
int64_t sdbGetCommitTerm(SSdb *pSdb) { return pSdb->lastCommitTerm; }
int64_t sdbGetCurConfig(SSdb *pSdb) { return pSdb->curConfig; }

38
source/dnode/mnode/sdb/src/sdbFile.c
View File

@ -110,6 +110,16 @@ static int32_t sdbReadFileHead(SSdb *pSdb, TdFilePtr pFile) {
return -1;
}
ret = taosReadFile(pFile, &pSdb->curConfig, sizeof(int64_t));
if (ret < 0) {
terrno = TAOS_SYSTEM_ERROR(errno);
return -1;
}
if (ret != sizeof(int64_t)) {
terrno = TSDB_CODE_FILE_CORRUPTED;
return -1;
}
for (int32_t i = 0; i < SDB_TABLE_SIZE; ++i) {
int64_t maxId = 0;
ret = taosReadFile(pFile, &maxId, sizeof(int64_t));
@ -173,6 +183,11 @@ static int32_t sdbWriteFileHead(SSdb *pSdb, TdFilePtr pFile) {
return -1;
}
if (taosWriteFile(pFile, &pSdb->curConfig, sizeof(int64_t)) != sizeof(int64_t)) {
terrno = TAOS_SYSTEM_ERROR(errno);
return -1;
}
for (int32_t i = 0; i < SDB_TABLE_SIZE; ++i) {
int64_t maxId = 0;
if (i < SDB_MAX) {
@ -288,8 +303,8 @@ static int32_t sdbReadFileImp(SSdb *pSdb) {
pSdb->lastCommitVer = pSdb->curVer;
pSdb->lastCommitTerm = pSdb->curTerm;
memcpy(pSdb->tableVer, tableVer, sizeof(tableVer));
mDebug("read sdb file:%s successfully, ver:%" PRId64 " term:%" PRId64, file, pSdb->lastCommitVer,
pSdb->lastCommitTerm);
mDebug("read sdb file:%s successfully, index:%" PRId64 " term:%" PRId64 " config:%" PRId64, file, pSdb->lastCommitVer,
pSdb->lastCommitTerm, pSdb->curConfig);
_OVER:
taosCloseFile(&pFile);
@ -426,12 +441,23 @@ static int32_t sdbWriteFileImp(SSdb *pSdb) {
}
int32_t sdbWriteFile(SSdb *pSdb) {
int32_t code = 0;
if (pSdb->curVer == pSdb->lastCommitVer) {
return 0;
}
taosThreadMutexLock(&pSdb->filelock);
int32_t code = sdbWriteFileImp(pSdb);
if (pSdb->pWal != NULL) {
code = walBeginSnapshot(pSdb->pWal, pSdb->curVer);
}
if (code == 0) {
code = sdbWriteFileImp(pSdb);
}
if (code == 0) {
if (pSdb->pWal != NULL) {
code = walEndSnapshot(pSdb->pWal);
}
}
if (code != 0) {
mError("failed to write sdb file since %s", terrstr());
}
@ -496,6 +522,9 @@ int32_t sdbStartRead(SSdb *pSdb, SSdbIter **ppIter) {
snprintf(datafile, sizeof(datafile), "%s%ssdb.data", pSdb->currDir, TD_DIRSEP);
taosThreadMutexLock(&pSdb->filelock);
int64_t commitIndex = pSdb->lastCommitVer;
int64_t commitTerm = pSdb->lastCommitTerm;
int64_t curConfig = pSdb->curConfig;
if (taosCopyFile(datafile, pIter->name) < 0) {
taosThreadMutexUnlock(&pSdb->filelock);
terrno = TAOS_SYSTEM_ERROR(errno);
@ -514,7 +543,8 @@ int32_t sdbStartRead(SSdb *pSdb, SSdbIter **ppIter) {
}
*ppIter = pIter;
mInfo("sdbiter:%p, is created to read snapshot, file:%s", pIter, pIter->name);
mInfo("sdbiter:%p, is created to read snapshot, index:%" PRId64 " term:%" PRId64 " config:%" PRId64 " file:%s", pIter,
commitIndex, commitTerm, curConfig, pIter->name);
return 0;
}

14
source/dnode/vnode/src/vnd/vnodeSvr.c
View File

@ -24,6 +24,7 @@ static int32_t vnodeProcessDropTbReq(SVnode *pVnode, int64_t version, void *pReq
static int32_t vnodeProcessSubmitReq(SVnode *pVnode, int64_t version, void *pReq, int32_t len, SRpcMsg *pRsp);
static int32_t vnodeProcessCreateTSmaReq(SVnode *pVnode, int64_t version, void *pReq, int32_t len, SRpcMsg *pRsp);
static int32_t vnodeProcessAlterConfirmReq(SVnode *pVnode, int64_t version, void *pReq, int32_t len, SRpcMsg *pRsp);
static int32_t vnodeProcessAlterHasnRangeReq(SVnode *pVnode, int64_t version, void *pReq, int32_t len, SRpcMsg *pRsp);
static int32_t vnodeProcessWriteMsg(SVnode *pVnode, int64_t version, SRpcMsg *pMsg, SRpcMsg *pRsp);
int32_t vnodePreprocessReq(SVnode *pVnode, SRpcMsg *pMsg) {
@ -163,6 +164,9 @@ int32_t vnodeProcessWriteReq(SVnode *pVnode, SRpcMsg *pMsg, int64_t version, SRp
case TDMT_VND_ALTER_CONFIRM:
vnodeProcessAlterConfirmReq(pVnode, version, pReq, len, pRsp);
break;
case TDMT_VND_ALTER_HASHRANGE:
vnodeProcessAlterHasnRangeReq(pVnode, version, pReq, len, pRsp);
break;
case TDMT_VND_ALTER_CONFIG:
break;
default:
@ -889,3 +893,13 @@ static int32_t vnodeProcessAlterConfirmReq(SVnode *pVnode, int64_t version, void
return 0;
}
static int32_t vnodeProcessAlterHasnRangeReq(SVnode *pVnode, int64_t version, void *pReq, int32_t len, SRpcMsg *pRsp) {
vInfo("vgId:%d, alter hashrange msg will be processed", TD_VID(pVnode));
// todo
// 1. stop work
// 2. adjust hash range / compact / remove wals / rename vgroups
// 3. reload sync
return 0;
}

2
source/dnode/vnode/src/vnd/vnodeSync.c
View File

@ -252,6 +252,8 @@ static SSyncFSM *vnodeSyncMakeFsm(SVnode *pVnode) {
int32_t vnodeSyncOpen(SVnode *pVnode, char *path) {
SSyncInfo syncInfo = {
.isStandBy = false,
.snapshotEnable = false,
.vgId = pVnode->config.vgId,
.isStandBy = pVnode->config.standby,
.syncCfg = pVnode->config.syncCfg,

8
source/libs/catalog/inc/catalogInt.h
View File

@ -67,6 +67,7 @@ typedef enum {
CTG_TASK_GET_DB_INFO,
CTG_TASK_GET_TB_META,
CTG_TASK_GET_TB_HASH,
CTG_TASK_GET_TB_INDEX,
CTG_TASK_GET_INDEX,
CTG_TASK_GET_UDF,
CTG_TASK_GET_USER,
@ -93,6 +94,10 @@ typedef struct SCtgTbMetaCtx {
int32_t flag;
} SCtgTbMetaCtx;
typedef struct SCtgTbIndexCtx {
SName* pName;
} SCtgTbIndexCtx;
typedef struct SCtgDbVgCtx {
char dbFName[TSDB_DB_FNAME_LEN];
} SCtgDbVgCtx;
@ -189,6 +194,7 @@ typedef struct SCtgJob {
int32_t indexNum;
int32_t userNum;
int32_t dbInfoNum;
int32_t tbIndexNum;
} SCtgJob;
typedef struct SCtgMsgCtx {
@ -490,7 +496,7 @@ int32_t ctgGetDBVgInfoFromMnode(CTG_PARAMS, SBuildUseDBInput *input, SUseDbOutpu
int32_t ctgGetQnodeListFromMnode(CTG_PARAMS, SArray *out, SCtgTask* pTask);
int32_t ctgGetDBCfgFromMnode(CTG_PARAMS, const char *dbFName, SDbCfgInfo *out, SCtgTask* pTask);
int32_t ctgGetIndexInfoFromMnode(CTG_PARAMS, const char *indexName, SIndexInfo *out, SCtgTask* pTask);
int32_t ctgGetTbIndexFromMnode(CTG_PARAMS, const char *tbFName, SArray** out, SCtgTask* pTask);
int32_t ctgGetTbIndexFromMnode(CTG_PARAMS, SName* name, SArray** out, SCtgTask* pTask);
int32_t ctgGetUdfInfoFromMnode(CTG_PARAMS, const char *funcName, SFuncInfo *out, SCtgTask* pTask);
int32_t ctgGetUserDbAuthFromMnode(CTG_PARAMS, const char *user, SGetUserAuthRsp *out, SCtgTask* pTask);
int32_t ctgGetTbMetaFromMnodeImpl(CTG_PARAMS, char *dbFName, char* tbName, STableMetaOutput* out, SCtgTask* pTask);

6
source/libs/catalog/src/catalog.c
View File

@ -1136,14 +1136,14 @@ int32_t catalogGetIndexMeta(SCatalog* pCtg, void *pTrans, const SEpSet* pMgmtEps
CTG_API_LEAVE(ctgGetIndexInfoFromMnode(CTG_PARAMS_LIST(), indexName, pInfo, NULL));
}
int32_t catalogGetTableIndex(SCatalog* pCtg, void *pTrans, const SEpSet* pMgmtEps, const char* tbFName, SArray** pRes) {
int32_t catalogGetTableIndex(SCatalog* pCtg, void *pTrans, const SEpSet* pMgmtEps, const SName* pTableName, SArray** pRes) {
CTG_API_ENTER();
if (NULL == pCtg || NULL == pTrans || NULL == pMgmtEps || NULL == tbFName || NULL == pRes) {
if (NULL == pCtg || NULL == pTrans || NULL == pMgmtEps || NULL == pTableName || NULL == pRes) {
CTG_API_LEAVE(TSDB_CODE_CTG_INVALID_INPUT);
}
CTG_API_LEAVE(ctgGetTbIndexFromMnode(CTG_PARAMS_LIST(), tbFName, pRes, NULL));
CTG_API_LEAVE(ctgGetTbIndexFromMnode(CTG_PARAMS_LIST(), (SName*)pTableName, pRes, NULL));
}

101
source/libs/catalog/src/ctgAsync.c
View File

@ -44,7 +44,7 @@ int32_t ctgInitGetTbMetaTask(SCtgJob *pJob, int32_t taskIdx, SName *name) {
taosArrayPush(pJob->pTasks, &task);
qDebug("QID:0x%" PRIx64 " the %d task type %s initialized, tableName:%s", pJob->queryId, taskIdx, ctgTaskTypeStr(task.type), name->tname);
qDebug("QID:0x%" PRIx64 " the %d task type %s initialized, tbName:%s", pJob->queryId, taskIdx, ctgTaskTypeStr(task.type), name->tname);
return TSDB_CODE_SUCCESS;
}
@ -232,6 +232,35 @@ int32_t ctgInitGetUserTask(SCtgJob *pJob, int32_t taskIdx, SUserAuthInfo *user)
return TSDB_CODE_SUCCESS;
}
int32_t ctgInitGetTbIndexTask(SCtgJob *pJob, int32_t taskIdx, SName *name) {
SCtgTask task = {0};
task.type = CTG_TASK_GET_TB_INDEX;
task.taskId = taskIdx;
task.pJob = pJob;
task.taskCtx = taosMemoryCalloc(1, sizeof(SCtgTbIndexCtx));
if (NULL == task.taskCtx) {
CTG_ERR_RET(TSDB_CODE_OUT_OF_MEMORY);
}
SCtgTbIndexCtx* ctx = task.taskCtx;
ctx->pName = taosMemoryMalloc(sizeof(*name));
if (NULL == ctx->pName) {
taosMemoryFree(task.taskCtx);
CTG_ERR_RET(TSDB_CODE_OUT_OF_MEMORY);
}
memcpy(ctx->pName, name, sizeof(*name));
taosArrayPush(pJob->pTasks, &task);
qDebug("QID:%" PRIx64 " the %d task type %s initialized, tbName:%s", pJob->queryId, taskIdx, ctgTaskTypeStr(task.type), name->tname);
return TSDB_CODE_SUCCESS;
}
int32_t ctgHandleForceUpdate(SCatalog* pCtg, SCtgJob *pJob, const SCatalogReq* pReq) {
int32_t dbNum = pJob->dbCfgNum + pJob->dbVgNum + pJob->dbInfoNum;
if (dbNum > 0) {
@ -329,8 +358,9 @@ int32_t ctgInitJob(CTG_PARAMS, SCtgJob** job, uint64_t reqId, const SCatalogReq*
int32_t indexNum = (int32_t)taosArrayGetSize(pReq->pIndex);
int32_t userNum = (int32_t)taosArrayGetSize(pReq->pUser);
int32_t dbInfoNum = (int32_t)taosArrayGetSize(pReq->pDbInfo);
int32_t tbIndexNum = (int32_t)taosArrayGetSize(pReq->pTableIndex);
*taskNum = tbMetaNum + dbVgNum + udfNum + tbHashNum + qnodeNum + dbCfgNum + indexNum + userNum + dbInfoNum;
*taskNum = tbMetaNum + dbVgNum + udfNum + tbHashNum + qnodeNum + dbCfgNum + indexNum + userNum + dbInfoNum + tbIndexNum;
if (*taskNum <= 0) {
ctgDebug("Empty input for job, no need to retrieve meta, reqId:0x%" PRIx64, reqId);
return TSDB_CODE_SUCCESS;
@ -360,6 +390,7 @@ int32_t ctgInitJob(CTG_PARAMS, SCtgJob** job, uint64_t reqId, const SCatalogReq*
pJob->indexNum = indexNum;
pJob->userNum = userNum;
pJob->dbInfoNum = dbInfoNum;
pJob->tbIndexNum = tbIndexNum;
pJob->pTasks = taosArrayInit(*taskNum, sizeof(SCtgTask));
@ -398,6 +429,11 @@ int32_t ctgInitJob(CTG_PARAMS, SCtgJob** job, uint64_t reqId, const SCatalogReq*
CTG_ERR_JRET(ctgInitGetTbHashTask(pJob, taskIdx++, name));
}
for (int32_t i = 0; i < tbIndexNum; ++i) {
SName* name = taosArrayGet(pReq->pTableIndex, i);
CTG_ERR_JRET(ctgInitGetTbIndexTask(pJob, taskIdx++, name));
}
for (int32_t i = 0; i < indexNum; ++i) {
char* indexName = taosArrayGet(pReq->pIndex, i);
CTG_ERR_JRET(ctgInitGetIndexTask(pJob, taskIdx++, indexName));
@ -479,6 +515,21 @@ int32_t ctgDumpTbHashRes(SCtgTask* pTask) {
return TSDB_CODE_SUCCESS;
}
int32_t ctgDumpTbIndexRes(SCtgTask* pTask) {
SCtgJob* pJob = pTask->pJob;
if (NULL == pJob->jobRes.pTableIndex) {
pJob->jobRes.pTableIndex = taosArrayInit(pJob->tbIndexNum, sizeof(SMetaRes));
if (NULL == pJob->jobRes.pTableIndex) {
CTG_ERR_RET(TSDB_CODE_OUT_OF_MEMORY);
}
}
SMetaRes res = {.code = pTask->code, .pRes = pTask->res};
taosArrayPush(pJob->jobRes.pTableHash, &res);
return TSDB_CODE_SUCCESS;
}
int32_t ctgDumpIndexRes(SCtgTask* pTask) {
SCtgJob* pJob = pTask->pJob;
if (NULL == pJob->jobRes.pIndex) {
@ -817,6 +868,20 @@ _return:
CTG_RET(code);
}
int32_t ctgHandleGetTbIndexRsp(SCtgTask* pTask, int32_t reqType, const SDataBuf *pMsg, int32_t rspCode) {
int32_t code = 0;
CTG_ERR_JRET(ctgProcessRspMsg(pTask->msgCtx.out, reqType, pMsg->pData, pMsg->len, rspCode, pTask->msgCtx.target));
TSWAP(pTask->res, pTask->msgCtx.out);
_return:
ctgHandleTaskEnd(pTask, code);
CTG_RET(code);
}
int32_t ctgHandleGetDbCfgRsp(SCtgTask* pTask, int32_t reqType, const SDataBuf *pMsg, int32_t rspCode) {
int32_t code = 0;
CTG_ERR_JRET(ctgProcessRspMsg(pTask->msgCtx.out, reqType, pMsg->pData, pMsg->len, rspCode, pTask->msgCtx.target));
@ -1056,13 +1121,24 @@ _return:
CTG_RET(code);
}
int32_t ctgLaunchGetTbIndexTask(SCtgTask *pTask) {
int32_t code = 0;
SCatalog* pCtg = pTask->pJob->pCtg;
void *pTrans = pTask->pJob->pTrans;
const SEpSet* pMgmtEps = &pTask->pJob->pMgmtEps;
SCtgTbIndexCtx* pCtx = (SCtgTbIndexCtx*)pTask->taskCtx;
CTG_ERR_RET(ctgGetTbIndexFromMnode(CTG_PARAMS_LIST(), pCtx->pName, NULL, pTask));
return TSDB_CODE_SUCCESS;
}
int32_t ctgLaunchGetQnodeTask(SCtgTask *pTask) {
SCatalog* pCtg = pTask->pJob->pCtg;
void *pTrans = pTask->pJob->pTrans;
const SEpSet* pMgmtEps = &pTask->pJob->pMgmtEps;
CTG_ERR_RET(ctgGetQnodeListFromMnode(CTG_PARAMS_LIST(), NULL, pTask));
return TSDB_CODE_SUCCESS;
}
@ -1168,15 +1244,16 @@ int32_t ctgRelaunchGetTbMetaTask(SCtgTask *pTask) {
}
SCtgAsyncFps gCtgAsyncFps[] = {
{ctgLaunchGetQnodeTask, ctgHandleGetQnodeRsp, ctgDumpQnodeRes},
{ctgLaunchGetDbVgTask, ctgHandleGetDbVgRsp, ctgDumpDbVgRes},
{ctgLaunchGetDbCfgTask, ctgHandleGetDbCfgRsp, ctgDumpDbCfgRes},
{ctgLaunchGetDbInfoTask, ctgHandleGetDbInfoRsp, ctgDumpDbInfoRes},
{ctgLaunchGetTbMetaTask, ctgHandleGetTbMetaRsp, ctgDumpTbMetaRes},
{ctgLaunchGetTbHashTask, ctgHandleGetTbHashRsp, ctgDumpTbHashRes},
{ctgLaunchGetIndexTask, ctgHandleGetIndexRsp, ctgDumpIndexRes},
{ctgLaunchGetUdfTask, ctgHandleGetUdfRsp, ctgDumpUdfRes},
{ctgLaunchGetUserTask, ctgHandleGetUserRsp, ctgDumpUserRes},
{ctgLaunchGetQnodeTask, ctgHandleGetQnodeRsp, ctgDumpQnodeRes},
{ctgLaunchGetDbVgTask, ctgHandleGetDbVgRsp, ctgDumpDbVgRes},
{ctgLaunchGetDbCfgTask, ctgHandleGetDbCfgRsp, ctgDumpDbCfgRes},
{ctgLaunchGetDbInfoTask, ctgHandleGetDbInfoRsp, ctgDumpDbInfoRes},
{ctgLaunchGetTbMetaTask, ctgHandleGetTbMetaRsp, ctgDumpTbMetaRes},
{ctgLaunchGetTbHashTask, ctgHandleGetTbHashRsp, ctgDumpTbHashRes},
{ctgLaunchGetTbIndexTask, ctgHandleGetTbIndexRsp, ctgDumpTbIndexRes},
{ctgLaunchGetIndexTask, ctgHandleGetIndexRsp, ctgDumpIndexRes},
{ctgLaunchGetUdfTask, ctgHandleGetUdfRsp, ctgDumpUdfRes},
{ctgLaunchGetUserTask, ctgHandleGetUserRsp, ctgDumpUserRes},
};
int32_t ctgMakeAsyncRes(SCtgJob *pJob) {

4
source/libs/catalog/src/ctgRemote.c
View File

@ -427,11 +427,13 @@ int32_t ctgGetIndexInfoFromMnode(CTG_PARAMS, const char *indexName, SIndexInfo *
return TSDB_CODE_SUCCESS;
}
int32_t ctgGetTbIndexFromMnode(CTG_PARAMS, const char *tbFName, SArray** out, SCtgTask* pTask) {
int32_t ctgGetTbIndexFromMnode(CTG_PARAMS, SName *name, SArray** out, SCtgTask* pTask) {
char *msg = NULL;
int32_t msgLen = 0;
int32_t reqType = TDMT_MND_GET_TABLE_INDEX;
void*(*mallocFp)(int32_t) = pTask ? taosMemoryMalloc : rpcMallocCont;
char tbFName[TSDB_TABLE_FNAME_LEN];
tNameExtractFullName(name, tbFName);
ctgDebug("try to get tb index from mnode, tbFName:%s", tbFName);

18
source/libs/catalog/src/ctgUtil.c
View File

@ -59,6 +59,9 @@ void ctgFreeSMetaData(SMetaData* pData) {
taosArrayDestroy(pData->pTableHash);
pData->pTableHash = NULL;
taosArrayDestroy(pData->pTableIndex);
pData->pTableIndex = NULL;
taosArrayDestroy(pData->pUdfList);
pData->pUdfList = NULL;
@ -248,6 +251,14 @@ void ctgFreeMsgCtx(SCtgMsgCtx* pCtx) {
taosMemoryFreeClear(pCtx->out);
break;
}
case TDMT_MND_GET_TABLE_INDEX: {
SArray** pOut = (SArray**)pCtx->out;
if (pOut) {
taosArrayDestroyEx(*pOut, tFreeSTableIndexInfo);
taosMemoryFreeClear(pCtx->out);
}
break;
}
case TDMT_MND_RETRIEVE_FUNC: {
SFuncInfo* pOut = (SFuncInfo*)pCtx->out;
taosMemoryFree(pOut->pCode);
@ -344,6 +355,13 @@ void ctgFreeTask(SCtgTask* pTask) {
taosMemoryFreeClear(pTask->res);
break;
}
case CTG_TASK_GET_TB_INDEX: {
SCtgTbIndexCtx* taskCtx = (SCtgTbIndexCtx*)pTask->taskCtx;
taosMemoryFreeClear(taskCtx->pName);
taosMemoryFreeClear(pTask->taskCtx);
taosArrayDestroyEx(pTask->res, tFreeSTableIndexInfo);
break;
}
case CTG_TASK_GET_INDEX: {
taosMemoryFreeClear(pTask->taskCtx);
taosMemoryFreeClear(pTask->res);

6
source/libs/command/inc/commandInt.h
View File

@ -32,15 +32,17 @@ extern "C" {
#define EXPLAIN_PROJECTION_FORMAT "Projection"
#define EXPLAIN_JOIN_FORMAT "%s"
#define EXPLAIN_AGG_FORMAT "Aggragate"
#define EXPLAIN_INDEF_ROWS_FORMAT "Indefinite Rows Function"
#define EXPLAIN_EXCHANGE_FORMAT "Data Exchange %d:1"
#define EXPLAIN_SORT_FORMAT "Sort"
#define EXPLAIN_INTERVAL_FORMAT "Interval on Column %s"
#define EXPLAIN_FILL_FORMAT "Fill"
#define EXPLAIN_SESSION_FORMAT "Session"
#define EXPLAIN_STATE_WINDOW_FORMAT "StateWindow on Column %s"
#define EXPLAIN_PARITION_FORMAT "Partition on Column %s"
#define EXPLAIN_ORDER_FORMAT "Order: %s"
#define EXPLAIN_FILTER_FORMAT "Filter: "
#define EXPLAIN_FILL_FORMAT "Fill: %s"
#define EXPLAIN_FILL_VALUE_FORMAT "Fill Values: "
#define EXPLAIN_ON_CONDITIONS_FORMAT "Join Cond: "
#define EXPLAIN_TIMERANGE_FORMAT "Time Range: [%" PRId64 ", %" PRId64 "]"
#define EXPLAIN_OUTPUT_FORMAT "Output: "
@ -66,6 +68,8 @@ extern "C" {
#define EXPLAIN_WIDTH_FORMAT "width=%d"
#define EXPLAIN_FUNCTIONS_FORMAT "functions=%d"
#define EXPLAIN_EXECINFO_FORMAT "cost=%.3f..%.3f rows=%" PRIu64
#define EXPLAIN_MODE_FORMAT "mode=%s"
#define EXPLAIN_STRING_TYPE_FORMAT "%s"
typedef struct SExplainGroup {
int32_t nodeNum;

156
source/libs/command/src/explain.c
View File

@ -179,6 +179,21 @@ int32_t qExplainGenerateResChildren(SPhysiNode *pNode, SExplainGroup *group, SNo
pPhysiChildren = mergePhysiNode->node.pChildren;
break;
}
case QUERY_NODE_PHYSICAL_PLAN_INDEF_ROWS_FUNC: {
SIndefRowsFuncPhysiNode *indefPhysiNode = (SIndefRowsFuncPhysiNode *)pNode;
pPhysiChildren = indefPhysiNode->node.pChildren;
break;
}
case QUERY_NODE_PHYSICAL_PLAN_MERGE_INTERVAL: {
SMergeIntervalPhysiNode *intPhysiNode = (SMergeIntervalPhysiNode *)pNode;
pPhysiChildren = intPhysiNode->window.node.pChildren;
break;
}
case QUERY_NODE_PHYSICAL_PLAN_FILL: {
SFillPhysiNode *fillPhysiNode = (SFillPhysiNode *)pNode;
pPhysiChildren = fillPhysiNode->node.pChildren;
break;
}
default:
qError("not supported physical node type %d", pNode->type);
QRY_ERR_RET(TSDB_CODE_QRY_APP_ERROR);
@ -212,12 +227,15 @@ int32_t qExplainGenerateResNodeExecInfo(SArray **pExecInfo, SExplainGroup *group
SExplainRsp *rsp = NULL;
for (int32_t i = 0; i < group->nodeNum; ++i) {
rsp = taosArrayGet(group->nodeExecInfo, i);
/*
if (group->physiPlanExecIdx >= rsp->numOfPlans) {
qError("physiPlanIdx %d exceed plan num %d", group->physiPlanExecIdx, rsp->numOfPlans);
return TSDB_CODE_QRY_APP_ERROR;
}
taosArrayPush(*pExecInfo, rsp->subplanInfo + group->physiPlanExecIdx);
*/
taosArrayPush(*pExecInfo, rsp->subplanInfo);
}
++group->physiPlanExecIdx;
@ -599,6 +617,42 @@ int32_t qExplainResNodeToRowsImpl(SExplainResNode *pResNode, SExplainCtx *ctx, i
}
break;
}
case QUERY_NODE_PHYSICAL_PLAN_INDEF_ROWS_FUNC: {
SIndefRowsFuncPhysiNode *pIndefNode = (SIndefRowsFuncPhysiNode *)pNode;
EXPLAIN_ROW_NEW(level, EXPLAIN_INDEF_ROWS_FORMAT);
EXPLAIN_ROW_APPEND(EXPLAIN_LEFT_PARENTHESIS_FORMAT);
if (pResNode->pExecInfo) {
QRY_ERR_RET(qExplainBufAppendExecInfo(pResNode->pExecInfo, tbuf, &tlen));
EXPLAIN_ROW_APPEND(EXPLAIN_BLANK_FORMAT);
}
if (pIndefNode->pVectorFuncs) {
EXPLAIN_ROW_APPEND(EXPLAIN_FUNCTIONS_FORMAT, pIndefNode->pVectorFuncs->length);
EXPLAIN_ROW_APPEND(EXPLAIN_BLANK_FORMAT);
}
EXPLAIN_ROW_APPEND(EXPLAIN_WIDTH_FORMAT, pIndefNode->node.pOutputDataBlockDesc->totalRowSize);
EXPLAIN_ROW_APPEND(EXPLAIN_RIGHT_PARENTHESIS_FORMAT);
EXPLAIN_ROW_END();
QRY_ERR_RET(qExplainResAppendRow(ctx, tbuf, tlen, level));
if (verbose) {
EXPLAIN_ROW_NEW(level + 1, EXPLAIN_OUTPUT_FORMAT);
EXPLAIN_ROW_APPEND(EXPLAIN_COLUMNS_FORMAT,
nodesGetOutputNumFromSlotList(pIndefNode->node.pOutputDataBlockDesc->pSlots));
EXPLAIN_ROW_APPEND(EXPLAIN_BLANK_FORMAT);
EXPLAIN_ROW_APPEND(EXPLAIN_WIDTH_FORMAT, pIndefNode->node.pOutputDataBlockDesc->outputRowSize);
EXPLAIN_ROW_END();
QRY_ERR_RET(qExplainResAppendRow(ctx, tbuf, tlen, level + 1));
if (pIndefNode->node.pConditions) {
EXPLAIN_ROW_NEW(level + 1, EXPLAIN_FILTER_FORMAT);
QRY_ERR_RET(nodesNodeToSQL(pIndefNode->node.pConditions, tbuf + VARSTR_HEADER_SIZE,
TSDB_EXPLAIN_RESULT_ROW_SIZE, &tlen));
EXPLAIN_ROW_END();
QRY_ERR_RET(qExplainResAppendRow(ctx, tbuf, tlen, level + 1));
}
}
break;
}
case QUERY_NODE_PHYSICAL_PLAN_EXCHANGE: {
SExchangePhysiNode *pExchNode = (SExchangePhysiNode *)pNode;
SExplainGroup *group = taosHashGet(ctx->groupHash, &pExchNode->srcGroupId, sizeof(pExchNode->srcGroupId));
@ -607,7 +661,7 @@ int32_t qExplainResNodeToRowsImpl(SExplainResNode *pResNode, SExplainCtx *ctx, i
QRY_ERR_RET(TSDB_CODE_QRY_APP_ERROR);
}
EXPLAIN_ROW_NEW(level, EXPLAIN_EXCHANGE_FORMAT, group->nodeNum);
EXPLAIN_ROW_NEW(level, EXPLAIN_EXCHANGE_FORMAT, pExchNode->singleChannel ? 1 : group->nodeNum);
EXPLAIN_ROW_APPEND(EXPLAIN_LEFT_PARENTHESIS_FORMAT);
if (pResNode->pExecInfo) {
QRY_ERR_RET(qExplainBufAppendExecInfo(pResNode->pExecInfo, tbuf, &tlen));
@ -750,6 +804,106 @@ int32_t qExplainResNodeToRowsImpl(SExplainResNode *pResNode, SExplainCtx *ctx, i
}
break;
}
case QUERY_NODE_PHYSICAL_PLAN_MERGE_INTERVAL: {
SMergeIntervalPhysiNode *pIntNode = (SMergeIntervalPhysiNode *)pNode;
EXPLAIN_ROW_NEW(level, EXPLAIN_INTERVAL_FORMAT, nodesGetNameFromColumnNode(pIntNode->window.pTspk));
EXPLAIN_ROW_APPEND(EXPLAIN_LEFT_PARENTHESIS_FORMAT);
if (pResNode->pExecInfo) {
QRY_ERR_RET(qExplainBufAppendExecInfo(pResNode->pExecInfo, tbuf, &tlen));
EXPLAIN_ROW_APPEND(EXPLAIN_BLANK_FORMAT);
}
EXPLAIN_ROW_APPEND(EXPLAIN_FUNCTIONS_FORMAT, pIntNode->window.pFuncs->length);
EXPLAIN_ROW_APPEND(EXPLAIN_BLANK_FORMAT);
EXPLAIN_ROW_APPEND(EXPLAIN_WIDTH_FORMAT, pIntNode->window.node.pOutputDataBlockDesc->totalRowSize);
EXPLAIN_ROW_APPEND(EXPLAIN_RIGHT_PARENTHESIS_FORMAT);
EXPLAIN_ROW_END();
QRY_ERR_RET(qExplainResAppendRow(ctx, tbuf, tlen, level));
if (verbose) {
EXPLAIN_ROW_NEW(level + 1, EXPLAIN_OUTPUT_FORMAT);
EXPLAIN_ROW_APPEND(EXPLAIN_COLUMNS_FORMAT,
nodesGetOutputNumFromSlotList(pIntNode->window.node.pOutputDataBlockDesc->pSlots));
EXPLAIN_ROW_APPEND(EXPLAIN_BLANK_FORMAT);
EXPLAIN_ROW_APPEND(EXPLAIN_WIDTH_FORMAT, pIntNode->window.node.pOutputDataBlockDesc->outputRowSize);
EXPLAIN_ROW_END();
QRY_ERR_RET(qExplainResAppendRow(ctx, tbuf, tlen, level + 1));
uint8_t precision = getIntervalPrecision(pIntNode);
EXPLAIN_ROW_NEW(level + 1, EXPLAIN_TIME_WINDOWS_FORMAT,
INVERAL_TIME_FROM_PRECISION_TO_UNIT(pIntNode->interval, pIntNode->intervalUnit, precision),
pIntNode->intervalUnit, pIntNode->offset, getPrecisionUnit(precision),
INVERAL_TIME_FROM_PRECISION_TO_UNIT(pIntNode->sliding, pIntNode->slidingUnit, precision),
pIntNode->slidingUnit);
EXPLAIN_ROW_END();
QRY_ERR_RET(qExplainResAppendRow(ctx, tbuf, tlen, level + 1));
if (pIntNode->window.node.pConditions) {
EXPLAIN_ROW_NEW(level + 1, EXPLAIN_FILTER_FORMAT);
QRY_ERR_RET(nodesNodeToSQL(pIntNode->window.node.pConditions, tbuf + VARSTR_HEADER_SIZE,
TSDB_EXPLAIN_RESULT_ROW_SIZE, &tlen));
EXPLAIN_ROW_END();
QRY_ERR_RET(qExplainResAppendRow(ctx, tbuf, tlen, level + 1));
}
}
break;
}
case QUERY_NODE_PHYSICAL_PLAN_FILL: {
SFillPhysiNode *pFillNode = (SFillPhysiNode *)pNode;
EXPLAIN_ROW_NEW(level, EXPLAIN_FILL_FORMAT);
EXPLAIN_ROW_APPEND(EXPLAIN_LEFT_PARENTHESIS_FORMAT);
if (pResNode->pExecInfo) {
QRY_ERR_RET(qExplainBufAppendExecInfo(pResNode->pExecInfo, tbuf, &tlen));
EXPLAIN_ROW_APPEND(EXPLAIN_BLANK_FORMAT);
}
EXPLAIN_ROW_APPEND(EXPLAIN_MODE_FORMAT, nodesGetFillModeString(pFillNode->mode));
EXPLAIN_ROW_APPEND(EXPLAIN_BLANK_FORMAT);
EXPLAIN_ROW_APPEND(EXPLAIN_WIDTH_FORMAT, pFillNode->node.pOutputDataBlockDesc->totalRowSize);
EXPLAIN_ROW_APPEND(EXPLAIN_RIGHT_PARENTHESIS_FORMAT);
EXPLAIN_ROW_END();
QRY_ERR_RET(qExplainResAppendRow(ctx, tbuf, tlen, level));
if (verbose) {
EXPLAIN_ROW_NEW(level + 1, EXPLAIN_OUTPUT_FORMAT);
EXPLAIN_ROW_APPEND(EXPLAIN_COLUMNS_FORMAT,
nodesGetOutputNumFromSlotList(pFillNode->node.pOutputDataBlockDesc->pSlots));
EXPLAIN_ROW_APPEND(EXPLAIN_BLANK_FORMAT);
EXPLAIN_ROW_APPEND(EXPLAIN_WIDTH_FORMAT, pFillNode->node.pOutputDataBlockDesc->outputRowSize);
EXPLAIN_ROW_END();
QRY_ERR_RET(qExplainResAppendRow(ctx, tbuf, tlen, level + 1));
if (pFillNode->pValues) {
SNodeListNode *pValues = (SNodeListNode*)pFillNode->pValues;
EXPLAIN_ROW_NEW(level + 1, EXPLAIN_FILL_VALUE_FORMAT);
SNode* tNode = NULL;
int32_t i = 0;
FOREACH(tNode, pValues->pNodeList) {
if (i) {
EXPLAIN_ROW_APPEND(EXPLAIN_BLANK_FORMAT);
}
SValueNode* tValue = (SValueNode*)tNode;
char *value = nodesGetStrValueFromNode(tValue);
EXPLAIN_ROW_APPEND(EXPLAIN_STRING_TYPE_FORMAT, value);
taosMemoryFree(value);
++i;
}
EXPLAIN_ROW_END();
QRY_ERR_RET(qExplainResAppendRow(ctx, tbuf, tlen, level + 1));
}
EXPLAIN_ROW_NEW(level + 1, EXPLAIN_TIMERANGE_FORMAT, pFillNode->timeRange.skey,
pFillNode->timeRange.ekey);
EXPLAIN_ROW_END();
QRY_ERR_RET(qExplainResAppendRow(ctx, tbuf, tlen, level + 1));
if (pFillNode->node.pConditions) {
EXPLAIN_ROW_NEW(level + 1, EXPLAIN_FILTER_FORMAT);
QRY_ERR_RET(nodesNodeToSQL(pFillNode->node.pConditions, tbuf + VARSTR_HEADER_SIZE,
TSDB_EXPLAIN_RESULT_ROW_SIZE, &tlen));
EXPLAIN_ROW_END();
QRY_ERR_RET(qExplainResAppendRow(ctx, tbuf, tlen, level + 1));
}
}
break;
}
case QUERY_NODE_PHYSICAL_PLAN_MERGE_SESSION: {
SSessionWinodwPhysiNode *pSessNode = (SSessionWinodwPhysiNode *)pNode;
EXPLAIN_ROW_NEW(level, EXPLAIN_SESSION_FORMAT);

6
source/libs/executor/inc/executorimpl.h
View File

@ -804,7 +804,7 @@ SOperatorInfo* createSortedMergeOperatorInfo(SOperatorInfo** downstream, int32_t
SOperatorInfo* createIntervalOperatorInfo(SOperatorInfo* downstream, SExprInfo* pExprInfo, int32_t numOfCols,
SSDataBlock* pResBlock, SInterval* pInterval, int32_t primaryTsSlotId,
STimeWindowAggSupp *pTwAggSupp, SExecTaskInfo* pTaskInfo);
STimeWindowAggSupp *pTwAggSupp, SExecTaskInfo* pTaskInfo, bool isStream);
SOperatorInfo* createStreamFinalIntervalOperatorInfo(SOperatorInfo* downstream,
SPhysiNode* pPhyNode, SExecTaskInfo* pTaskInfo, int32_t numOfChild);
SOperatorInfo* createStreamIntervalOperatorInfo(SOperatorInfo* downstream, SExprInfo* pExprInfo, int32_t numOfCols,
@ -892,8 +892,8 @@ STimeWindow getActiveTimeWindow(SDiskbasedBuf* pBuf, SResultRowInfo* pResultRowI
int32_t getNumOfRowsInTimeWindow(SDataBlockInfo* pDataBlockInfo, TSKEY* pPrimaryColumn, int32_t startPos, TSKEY ekey,
__block_search_fn_t searchFn, STableQueryInfo* item, int32_t order);
int32_t binarySearchForKey(char* pValue, int num, TSKEY key, int order);
int32_t initSessionAggSupporter(SStreamAggSupporter* pSup, const char* pKey);
int32_t initStateAggSupporter(SStreamAggSupporter* pSup, const char* pKey);
int32_t initStreamAggSupporter(SStreamAggSupporter* pSup, const char* pKey,
SqlFunctionCtx* pCtx, int32_t numOfOutput, size_t size);
SResultRow* getNewResultRow(SDiskbasedBuf* pResultBuf, int64_t tableGroupId, int32_t interBufSize);
SResultWindowInfo* getSessionTimeWindow(SArray* pWinInfos, TSKEY ts, int64_t gap, int32_t* pIndex);
int32_t updateSessionWindowInfo(SResultWindowInfo* pWinInfo, TSKEY* pTs, int32_t rows,

33
source/libs/executor/src/executorimpl.c
View File

@ -1130,6 +1130,7 @@ SqlFunctionCtx* createSqlFunctionCtx(SExprInfo* pExprInfo, int32_t numOfOutput,
pCtx->start.key = INT64_MIN;
pCtx->end.key = INT64_MIN;
pCtx->numOfParams = pExpr->base.numOfParams;
pCtx->increase = false;
pCtx->param = pFunct->pParam;
// for (int32_t j = 0; j < pCtx->numOfParams; ++j) {
@ -2008,8 +2009,16 @@ int32_t doCopyToSDataBlock(SExecTaskInfo* pTaskInfo, SSDataBlock* pBlock, SExprI
// the _wstartts needs to copy to 20 following rows, since the results of top-k expands to 20 different rows.
SColumnInfoData* pColInfoData = taosArrayGet(pBlock->pDataBlock, slotId);
char* in = GET_ROWCELL_INTERBUF(pCtx[j].resultInfo);
for (int32_t k = 0; k < pRow->numOfRows; ++k) {
colDataAppend(pColInfoData, pBlock->info.rows + k, in, pCtx[j].resultInfo->isNullRes);
if (pCtx[j].increase) {
int64_t ts = *(int64_t*) in;
for (int32_t k = 0; k < pRow->numOfRows; ++k) {
colDataAppend(pColInfoData, pBlock->info.rows + k, (const char *)&ts, pCtx[j].resultInfo->isNullRes);
ts++;
}
} else {
for (int32_t k = 0; k < pRow->numOfRows; ++k) {
colDataAppend(pColInfoData, pBlock->info.rows + k, in, pCtx[j].resultInfo->isNullRes);
}
}
}
}
@ -4676,7 +4685,8 @@ SOperatorInfo* createOperatorTree(SPhysiNode* pPhyNode, SExecTaskInfo* pTaskInfo
}
int32_t tsSlotId = ((SColumnNode*)pIntervalPhyNode->window.pTspk)->slotId;
pOptr = createIntervalOperatorInfo(ops[0], pExprInfo, num, pResBlock, &interval, tsSlotId, &as, pTaskInfo);
bool isStream = (QUERY_NODE_PHYSICAL_PLAN_STREAM_INTERVAL == type);
pOptr = createIntervalOperatorInfo(ops[0], pExprInfo, num, pResBlock, &interval, tsSlotId, &as, pTaskInfo, isStream);
} else if (QUERY_NODE_PHYSICAL_PLAN_STREAM_SEMI_INTERVAL == type) {
int32_t children = 8;
@ -5339,7 +5349,8 @@ int32_t getOperatorExplainExecInfo(SOperatorInfo* operatorInfo, SExplainExecInfo
return TSDB_CODE_SUCCESS;
}
int32_t initStreamAggSupporter(SStreamAggSupporter* pSup, const char* pKey, size_t size) {
int32_t initStreamAggSupporter(SStreamAggSupporter* pSup, const char* pKey, SqlFunctionCtx* pCtx, int32_t numOfOutput, size_t size) {
pSup->resultRowSize = getResultRowSize(pCtx, numOfOutput);
pSup->keySize = sizeof(int64_t) + sizeof(TSKEY);
pSup->pKeyBuf = taosMemoryCalloc(1, pSup->keySize);
pSup->pResultRows = taosArrayInit(1024, size);
@ -5358,15 +5369,11 @@ int32_t initStreamAggSupporter(SStreamAggSupporter* pSup, const char* pKey, size
if (bufSize <= pageSize) {
bufSize = pageSize * 4;
}
return createDiskbasedBuf(&pSup->pResultBuf, pageSize, bufSize, pKey, TD_TMP_DIR_PATH);
}
int32_t initSessionAggSupporter(SStreamAggSupporter* pSup, const char* pKey) {
return initStreamAggSupporter(pSup, pKey, sizeof(SResultWindowInfo));
}
int32_t initStateAggSupporter(SStreamAggSupporter* pSup, const char* pKey) {
return initStreamAggSupporter(pSup, pKey, sizeof(SStateWindowInfo));
int32_t code = createDiskbasedBuf(&pSup->pResultBuf, pageSize, bufSize, pKey, TD_TMP_DIR_PATH);
for(int32_t i = 0; i < numOfOutput; ++i) {
pCtx[i].pBuf = pSup->pResultBuf;
}
return code;
}
int64_t getSmaWaterMark(int64_t interval, double filesFactor) {

1
source/libs/executor/src/scanoperator.c
View File

@ -1756,6 +1756,7 @@ SOperatorInfo* createTagScanOperatorInfo(SReadHandle* pReadHandle, STagScanPhysi
;
pInfo->readHandle = *pReadHandle;
pInfo->curPos = 0;
pInfo->pFilterNode = pPhyNode->node.pConditions;
pOperator->name = "TagScanOperator";
pOperator->operatorType = QUERY_NODE_PHYSICAL_PLAN_TAG_SCAN;
pOperator->blocking = false;

32
source/libs/executor/src/timewindowoperator.c
View File

@ -1438,9 +1438,15 @@ static bool timeWindowinterpNeeded(SqlFunctionCtx* pCtx, int32_t numOfCols, SInt
return needed;
}
void increaseTs(SqlFunctionCtx* pCtx) {
if (pCtx[0].pExpr->pExpr->_function.pFunctNode->funcType == FUNCTION_TYPE_WSTARTTS) {
pCtx[0].increase = true;
}
}
SOperatorInfo* createIntervalOperatorInfo(SOperatorInfo* downstream, SExprInfo* pExprInfo, int32_t numOfCols,
SSDataBlock* pResBlock, SInterval* pInterval, int32_t primaryTsSlotId,
STimeWindowAggSupp* pTwAggSupp, SExecTaskInfo* pTaskInfo) {
STimeWindowAggSupp* pTwAggSupp, SExecTaskInfo* pTaskInfo, bool isStream) {
SIntervalAggOperatorInfo* pInfo = taosMemoryCalloc(1, sizeof(SIntervalAggOperatorInfo));
SOperatorInfo* pOperator = taosMemoryCalloc(1, sizeof(SOperatorInfo));
if (pInfo == NULL || pOperator == NULL) {
@ -1460,6 +1466,11 @@ SOperatorInfo* createIntervalOperatorInfo(SOperatorInfo* downstream, SExprInfo*
int32_t code =
initAggInfo(&pInfo->binfo, &pInfo->aggSup, pExprInfo, numOfCols, pResBlock, keyBufSize, pTaskInfo->id.str);
if (isStream) {
ASSERT(numOfCols > 0);
increaseTs(pInfo->binfo.pCtx);
}
initExecTimeWindowInfo(&pInfo->twAggSup.timeWindowData, &pInfo->win);
@ -2128,6 +2139,8 @@ SOperatorInfo* createStreamFinalIntervalOperatorInfo(SOperatorInfo* downstream,
SSDataBlock* pResBlock = createResDataBlock(pPhyNode->pOutputDataBlockDesc);
int32_t code = initAggInfo(&pInfo->binfo, &pInfo->aggSup, pExprInfo, numOfCols,
pResBlock, keyBufSize, pTaskInfo->id.str);
ASSERT(numOfCols > 0);
increaseTs(pInfo->binfo.pCtx);
initExecTimeWindowInfo(&pInfo->twAggSup.timeWindowData, &pTaskInfo->window);
if (code != TSDB_CODE_SUCCESS) {
goto _error;
@ -2212,6 +2225,8 @@ int32_t initBiasicInfo(SOptrBasicInfo* pBasicInfo, SExprInfo* pExprInfo,
for (int32_t i = 0; i < numOfCols; ++i) {
pBasicInfo->pCtx[i].pBuf = NULL;
}
ASSERT(numOfCols > 0);
increaseTs(pBasicInfo->pCtx);
return TSDB_CODE_SUCCESS;
}
@ -2228,6 +2243,10 @@ void initDownStream(SOperatorInfo* downstream, SStreamAggSupporter* pAggSup, int
pScanInfo->pUpdateInfo = updateInfoInit(60000, TSDB_TIME_PRECISION_MILLI, waterMark);
}
int32_t initSessionAggSupporter(SStreamAggSupporter* pSup, const char* pKey, SqlFunctionCtx* pCtx, int32_t numOfOutput) {
return initStreamAggSupporter(pSup, pKey, pCtx, numOfOutput, sizeof(SResultWindowInfo));
}
SOperatorInfo* createStreamSessionAggOperatorInfo(SOperatorInfo* downstream, SExprInfo* pExprInfo, int32_t numOfCols,
SSDataBlock* pResBlock, int64_t gap, int32_t tsSlotId,
STimeWindowAggSupp* pTwAggSupp, SExecTaskInfo* pTaskInfo) {
@ -2244,8 +2263,8 @@ SOperatorInfo* createStreamSessionAggOperatorInfo(SOperatorInfo* downstream, SEx
if (code != TSDB_CODE_SUCCESS) {
goto _error;
}
pInfo->streamAggSup.resultRowSize = getResultRowSize(pInfo->binfo.pCtx, numOfCols);
code = initSessionAggSupporter(&pInfo->streamAggSup, "StreamSessionAggOperatorInfo");
code = initSessionAggSupporter(&pInfo->streamAggSup, "StreamSessionAggOperatorInfo", pInfo->binfo.pCtx, numOfCols);
if (code != TSDB_CODE_SUCCESS) {
goto _error;
}
@ -3097,6 +3116,10 @@ static SSDataBlock* doStreamStateAgg(SOperatorInfo* pOperator) {
return pBInfo->pRes->info.rows == 0 ? NULL : pBInfo->pRes;
}
int32_t initStateAggSupporter(SStreamAggSupporter* pSup, const char* pKey, SqlFunctionCtx* pCtx, int32_t numOfOutput) {
return initStreamAggSupporter(pSup, pKey, pCtx, numOfOutput, sizeof(SStateWindowInfo));
}
SOperatorInfo* createStreamStateAggOperatorInfo(SOperatorInfo* downstream, SPhysiNode* pPhyNode,
SExecTaskInfo* pTaskInfo) {
SStreamStateWinodwPhysiNode* pStateNode = (SStreamStateWinodwPhysiNode*)pPhyNode;
@ -3130,8 +3153,7 @@ SOperatorInfo* createStreamStateAggOperatorInfo(SOperatorInfo* downstream, SPhys
goto _error;
}
pInfo->streamAggSup.resultRowSize = getResultRowSize(pInfo->binfo.pCtx, numOfCols);
code = initStateAggSupporter(&pInfo->streamAggSup, "StreamStateAggOperatorInfo");
code = initStateAggSupporter(&pInfo->streamAggSup, "StreamStateAggOperatorInfo", pInfo->binfo.pCtx, numOfCols);
if (code != TSDB_CODE_SUCCESS) {
goto _error;
}

8
source/libs/function/inc/builtinsimpl.h
View File

@ -102,6 +102,8 @@ bool getTopBotFuncEnv(SFunctionNode* UNUSED_PARAM(pFunc), SFuncExecEnv* pEnv);
int32_t topFunction(SqlFunctionCtx *pCtx);
int32_t bottomFunction(SqlFunctionCtx *pCtx);
int32_t topBotFinalize(SqlFunctionCtx* pCtx, SSDataBlock* pBlock);
int32_t topCombine(SqlFunctionCtx* pDestCtx, SqlFunctionCtx* pSourceCtx);
int32_t bottomCombine(SqlFunctionCtx* pDestCtx, SqlFunctionCtx* pSourceCtx);
bool getSpreadFuncEnv(struct SFunctionNode* pFunc, SFuncExecEnv* pEnv);
bool spreadFunctionSetup(SqlFunctionCtx *pCtx, SResultRowEntryInfo* pResultInfo);
@ -114,7 +116,10 @@ int32_t getSpreadInfoSize();
bool getElapsedFuncEnv(struct SFunctionNode* pFunc, SFuncExecEnv* pEnv);
bool elapsedFunctionSetup(SqlFunctionCtx *pCtx, SResultRowEntryInfo* pResultInfo);
int32_t elapsedFunction(SqlFunctionCtx* pCtx);
int32_t elapsedFunctionMerge(SqlFunctionCtx* pCtx);
int32_t elapsedFinalize(SqlFunctionCtx* pCtx, SSDataBlock* pBlock);
int32_t elapsedPartialFinalize(SqlFunctionCtx* pCtx, SSDataBlock* pBlock);
int32_t getElapsedInfoSize();
bool getHistogramFuncEnv(struct SFunctionNode* pFunc, SFuncExecEnv* pEnv);
bool histogramFunctionSetup(SqlFunctionCtx *pCtx, SResultRowEntryInfo* pResultInfo);
@ -126,7 +131,10 @@ int32_t getHistogramInfoSize();
bool getHLLFuncEnv(struct SFunctionNode* pFunc, SFuncExecEnv* pEnv);
int32_t hllFunction(SqlFunctionCtx* pCtx);
int32_t hllFunctionMerge(SqlFunctionCtx* pCtx);
int32_t hllFinalize(SqlFunctionCtx* pCtx, SSDataBlock* pBlock);
int32_t hllPartialFinalize(SqlFunctionCtx* pCtx, SSDataBlock* pBlock);
int32_t getHLLInfoSize();
bool getStateFuncEnv(struct SFunctionNode* pFunc, SFuncExecEnv* pEnv);
bool stateFunctionSetup(SqlFunctionCtx *pCtx, SResultRowEntryInfo* pResultInfo);

180
source/libs/function/src/builtins.c
View File

@ -313,6 +313,7 @@ static int32_t translateApercentileImpl(SFunctionNode* pFunc, char* pErrBuf, int
static int32_t translateApercentilePartial(SFunctionNode* pFunc, char* pErrBuf, int32_t len) {
return translateApercentileImpl(pFunc, pErrBuf, len, true);
}
static int32_t translateApercentileMerge(SFunctionNode* pFunc, char* pErrBuf, int32_t len) {
return translateApercentileImpl(pFunc, pErrBuf, len, false);
}
@ -401,6 +402,7 @@ static int32_t translateSpreadImpl(SFunctionNode* pFunc, char* pErrBuf, int32_t
static int32_t translateSpreadPartial(SFunctionNode* pFunc, char* pErrBuf, int32_t len) {
return translateSpreadImpl(pFunc, pErrBuf, len, true);
}
static int32_t translateSpreadMerge(SFunctionNode* pFunc, char* pErrBuf, int32_t len) {
return translateSpreadImpl(pFunc, pErrBuf, len, false);
}
@ -442,6 +444,64 @@ static int32_t translateElapsed(SFunctionNode* pFunc, char* pErrBuf, int32_t len
return TSDB_CODE_SUCCESS;
}
static int32_t translateElapsedImpl(SFunctionNode* pFunc, char* pErrBuf, int32_t len, bool isPartial) {
int32_t numOfParams = LIST_LENGTH(pFunc->pParameterList);
if (isPartial) {
if (1 != numOfParams && 2 != numOfParams) {
return invaildFuncParaNumErrMsg(pErrBuf, len, pFunc->functionName);
}
uint8_t paraType = ((SExprNode*)nodesListGetNode(pFunc->pParameterList, 0))->resType.type;
if (TSDB_DATA_TYPE_TIMESTAMP != paraType) {
return invaildFuncParaTypeErrMsg(pErrBuf, len, pFunc->functionName);
}
// param1
if (2 == numOfParams) {
SNode* pParamNode1 = nodesListGetNode(pFunc->pParameterList, 1);
if (QUERY_NODE_VALUE != nodeType(pParamNode1)) {
return invaildFuncParaTypeErrMsg(pErrBuf, len, pFunc->functionName);
}
SValueNode* pValue = (SValueNode*)pParamNode1;
pValue->notReserved = true;
paraType = ((SExprNode*)nodesListGetNode(pFunc->pParameterList, 1))->resType.type;
if (!IS_INTEGER_TYPE(paraType)) {
return invaildFuncParaTypeErrMsg(pErrBuf, len, pFunc->functionName);
}
if (pValue->datum.i == 0) {
return buildFuncErrMsg(pErrBuf, len, TSDB_CODE_FUNC_FUNTION_ERROR,
"ELAPSED function time unit parameter should be greater than db precision");
}
}
pFunc->node.resType = (SDataType){.bytes = getElapsedInfoSize() + VARSTR_HEADER_SIZE, .type = TSDB_DATA_TYPE_BINARY};
} else {
if (1 != numOfParams) {
return invaildFuncParaNumErrMsg(pErrBuf, len, pFunc->functionName);
}
uint8_t paraType = ((SExprNode*)nodesListGetNode(pFunc->pParameterList, 0))->resType.type;
if (TSDB_DATA_TYPE_BINARY != paraType) {
return invaildFuncParaTypeErrMsg(pErrBuf, len, pFunc->functionName);
}
pFunc->node.resType = (SDataType){.bytes = tDataTypes[TSDB_DATA_TYPE_DOUBLE].bytes, .type = TSDB_DATA_TYPE_DOUBLE};
}
return TSDB_CODE_SUCCESS;
}
static int32_t translateElapsedPartial(SFunctionNode* pFunc, char* pErrBuf, int32_t len) {
return translateElapsedImpl(pFunc, pErrBuf, len, true);
}
static int32_t translateElapsedMerge(SFunctionNode* pFunc, char* pErrBuf, int32_t len) {
return translateElapsedImpl(pFunc, pErrBuf, len, false);
}
static int32_t translateLeastSQR(SFunctionNode* pFunc, char* pErrBuf, int32_t len) {
int32_t numOfParams = LIST_LENGTH(pFunc->pParameterList);
if (3 != numOfParams) {
@ -551,6 +611,7 @@ static int32_t translateHistogramImpl(SFunctionNode* pFunc, char* pErrBuf, int32
static int32_t translateHistogramPartial(SFunctionNode* pFunc, char* pErrBuf, int32_t len) {
return translateHistogramImpl(pFunc, pErrBuf, len, true);
}
static int32_t translateHistogramMerge(SFunctionNode* pFunc, char* pErrBuf, int32_t len) {
return translateHistogramImpl(pFunc, pErrBuf, len, false);
}
@ -564,6 +625,28 @@ static int32_t translateHLL(SFunctionNode* pFunc, char* pErrBuf, int32_t len) {
return TSDB_CODE_SUCCESS;
}
static int32_t translateHLLImpl(SFunctionNode* pFunc, char* pErrBuf, int32_t len, bool isPartial) {
if (1 != LIST_LENGTH(pFunc->pParameterList)) {
return invaildFuncParaNumErrMsg(pErrBuf, len, pFunc->functionName);
}
if (isPartial) {
pFunc->node.resType = (SDataType){.bytes = getHistogramInfoSize() + VARSTR_HEADER_SIZE, .type = TSDB_DATA_TYPE_BINARY};
} else {
pFunc->node.resType = (SDataType){.bytes = tDataTypes[TSDB_DATA_TYPE_BIGINT].bytes, .type = TSDB_DATA_TYPE_BIGINT};
}
return TSDB_CODE_SUCCESS;
}
static int32_t translateHLLPartial(SFunctionNode* pFunc, char* pErrBuf, int32_t len) {
return translateHLLImpl(pFunc, pErrBuf, len, true);
}
static int32_t translateHLLMerge(SFunctionNode* pFunc, char* pErrBuf, int32_t len) {
return translateHLLImpl(pFunc, pErrBuf, len, false);
}
static bool validateStateOper(const SValueNode* pVal) {
if (TSDB_DATA_TYPE_BINARY != pVal->node.resType.type) {
return false;
@ -1020,6 +1103,22 @@ static int32_t translateCast(SFunctionNode* pFunc, char* pErrBuf, int32_t len) {
*
*/
static bool validateHourRange(int8_t hour) {
if (hour < 0 || hour > 12) {
return false;
}
return true;
}
static bool validateMinuteRange(int8_t hour, int8_t minute, char sign) {
if (minute == 0 || (minute == 30 && (hour == 3 || hour == 5) && sign == '-')) {
return true;
}
return false;
}
static bool validateTimezoneFormat(const SValueNode* pVal) {
if (TSDB_DATA_TYPE_BINARY != pVal->node.resType.type) {
return false;
@ -1028,6 +1127,8 @@ static bool validateTimezoneFormat(const SValueNode* pVal) {
char* tz = varDataVal(pVal->datum.p);
int32_t len = varDataLen(pVal->datum.p);
char buf[3] = {0};
int8_t hour = -1, minute = -1;
if (len == 0) {
return false;
} else if (len == 1 && (tz[0] == 'z' || tz[0] == 'Z')) {
@ -1040,6 +1141,20 @@ static bool validateTimezoneFormat(const SValueNode* pVal) {
if (!isdigit(tz[i])) {
return false;
}
if (i == 2) {
memcpy(buf, &tz[i - 1], 2);
hour = taosStr2Int8(buf, NULL, 10);
if (!validateHourRange(hour)) {
return false;
}
} else if (i == 4) {
memcpy(buf, &tz[i - 1], 2);
minute = taosStr2Int8(buf, NULL, 10);
if (!validateMinuteRange(hour, minute, tz[0])) {
return false;
}
}
}
break;
}
@ -1051,9 +1166,24 @@ static bool validateTimezoneFormat(const SValueNode* pVal) {
}
continue;
}
if (!isdigit(tz[i])) {
return false;
}
if (i == 2) {
memcpy(buf, &tz[i - 1], 2);
hour = taosStr2Int8(buf, NULL, 10);
if (!validateHourRange(hour)) {
return false;
}
} else if (i == 5) {
memcpy(buf, &tz[i - 1], 2);
minute = taosStr2Int8(buf, NULL, 10);
if (!validateMinuteRange(hour, minute, tz[0])) {
return false;
}
}
}
break;
}
@ -1339,6 +1469,7 @@ const SBuiltinFuncDefinition funcMgtBuiltins[] = {
.initFunc = functionSetup,
.processFunc = topFunction,
.finalizeFunc = topBotFinalize,
.combineFunc = topCombine,
},
{
.name = "bottom",
@ -1348,7 +1479,8 @@ const SBuiltinFuncDefinition funcMgtBuiltins[] = {
.getEnvFunc = getTopBotFuncEnv,
.initFunc = functionSetup,
.processFunc = bottomFunction,
.finalizeFunc = topBotFinalize
.finalizeFunc = topBotFinalize,
.combineFunc = bottomCombine,
},
{
.name = "spread",
@ -1394,6 +1526,30 @@ const SBuiltinFuncDefinition funcMgtBuiltins[] = {
.getEnvFunc = getElapsedFuncEnv,
.initFunc = elapsedFunctionSetup,
.processFunc = elapsedFunction,
.finalizeFunc = elapsedFinalize,
.pPartialFunc = "_elapsed_partial",
.pMergeFunc = "_elapsed_merge"
},
{
.name = "_elapsed_partial",
.type = FUNCTION_TYPE_ELAPSED,
.classification = FUNC_MGT_AGG_FUNC,
.dataRequiredFunc = statisDataRequired,
.translateFunc = translateElapsedPartial,
.getEnvFunc = getElapsedFuncEnv,
.initFunc = elapsedFunctionSetup,
.processFunc = elapsedFunction,
.finalizeFunc = elapsedPartialFinalize
},
{
.name = "_elapsed_merge",
.type = FUNCTION_TYPE_ELAPSED,
.classification = FUNC_MGT_AGG_FUNC,
.dataRequiredFunc = statisDataRequired,
.translateFunc = translateElapsedMerge,
.getEnvFunc = getElapsedFuncEnv,
.initFunc = elapsedFunctionSetup,
.processFunc = elapsedFunctionMerge,
.finalizeFunc = elapsedFinalize
},
{
@ -1478,6 +1634,28 @@ const SBuiltinFuncDefinition funcMgtBuiltins[] = {
.getEnvFunc = getHLLFuncEnv,
.initFunc = functionSetup,
.processFunc = hllFunction,
.finalizeFunc = hllFinalize,
.pPartialFunc = "_hyperloglog_partial",
.pMergeFunc = "_hyperloglog_merge"
},
{
.name = "_hyperloglog_partial",
.type = FUNCTION_TYPE_HYPERLOGLOG_PARTIAL,
.classification = FUNC_MGT_AGG_FUNC,
.translateFunc = translateHLLPartial,
.getEnvFunc = getHLLFuncEnv,
.initFunc = functionSetup,
.processFunc = hllFunction,
.finalizeFunc = hllPartialFinalize
},
{
.name = "_hyperloglog_merge",
.type = FUNCTION_TYPE_HYPERLOGLOG_MERGE,
.classification = FUNC_MGT_AGG_FUNC,
.translateFunc = translateHLLMerge,
.getEnvFunc = getHLLFuncEnv,
.initFunc = functionSetup,
.processFunc = hllFunctionMerge,
.finalizeFunc = hllFinalize
},
{

166
source/libs/function/src/builtinsimpl.c
View File

@ -1389,6 +1389,18 @@ void setSelectivityValue(SqlFunctionCtx* pCtx, SSDataBlock* pBlock, const STuple
}
}
void releaseSource(STuplePos* pPos) {
if (pPos->pageId == -1) {
return ;
}
// Todo(liuyao) relase row
}
void replaceTupleData(STuplePos* pDestPos, STuplePos* pSourcePos) {
releaseSource(pDestPos);
*pDestPos = *pSourcePos;
}
int32_t minMaxCombine(SqlFunctionCtx* pDestCtx, SqlFunctionCtx* pSourceCtx, int32_t isMinFunc) {
SResultRowEntryInfo* pDResInfo = GET_RES_INFO(pDestCtx);
SMinmaxResInfo* pDBuf = GET_ROWCELL_INTERBUF(pDResInfo);
@ -1400,10 +1412,12 @@ int32_t minMaxCombine(SqlFunctionCtx* pDestCtx, SqlFunctionCtx* pSourceCtx, int3
if (pSBuf->assign &&
( (((*(double*)&pDBuf->v) < (*(double*)&pSBuf->v)) ^ isMinFunc) || !pDBuf->assign ) ) {
*(double*) &pDBuf->v = *(double*) &pSBuf->v;
replaceTupleData(&pDBuf->tuplePos, &pSBuf->tuplePos);
}
} else {
if ( pSBuf->assign && ( ((pDBuf->v < pSBuf->v) ^ isMinFunc) || !pDBuf->assign ) ) {
pDBuf->v = pSBuf->v;
replaceTupleData(&pDBuf->tuplePos, &pSBuf->tuplePos);
}
}
pDResInfo->numOfRes = TMAX(pDResInfo->numOfRes, pSResInfo->numOfRes);
@ -2856,7 +2870,6 @@ void copyTupleData(SqlFunctionCtx* pCtx, int32_t rowIndex, const SSDataBlock* pS
int32_t topBotFinalize(SqlFunctionCtx* pCtx, SSDataBlock* pBlock) {
SResultRowEntryInfo* pEntryInfo = GET_RES_INFO(pCtx);
STopBotRes* pRes = GET_ROWCELL_INTERBUF(pEntryInfo);
pEntryInfo->complete = true;
int32_t type = pCtx->input.pData[0]->info.type;
int32_t slotId = pCtx->pExpr->base.resSchema.slotId;
@ -2881,6 +2894,67 @@ int32_t topBotFinalize(SqlFunctionCtx* pCtx, SSDataBlock* pBlock) {
return pEntryInfo->numOfRes;
}
void addResult(SqlFunctionCtx* pCtx, STopBotResItem* pSourceItem, int16_t type,
bool isTopQuery) {
SResultRowEntryInfo* pEntryInfo = GET_RES_INFO(pCtx);
STopBotRes* pRes = getTopBotOutputInfo(pCtx);
int32_t maxSize = pCtx->param[1].param.i;
STopBotResItem* pItems = pRes->pItems;
assert(pItems != NULL);
// not full yet
if (pEntryInfo->numOfRes < maxSize) {
STopBotResItem* pItem = &pItems[pEntryInfo->numOfRes];
pItem->v = pSourceItem->v;
pItem->uid = pSourceItem->uid;
pItem->tuplePos.pageId = -1;
replaceTupleData(&pItem->tuplePos, &pSourceItem->tuplePos);
pEntryInfo->numOfRes++;
taosheapsort((void*)pItems, sizeof(STopBotResItem), pEntryInfo->numOfRes, (const void*)&type, topBotResComparFn,
!isTopQuery);
} else { // replace the minimum value in the result
if ((isTopQuery && (
(IS_SIGNED_NUMERIC_TYPE(type) && pSourceItem->v.i > pItems[0].v.i) ||
(IS_UNSIGNED_NUMERIC_TYPE(type) && pSourceItem->v.u > pItems[0].v.u) ||
(IS_FLOAT_TYPE(type) && pSourceItem->v.d > pItems[0].v.d)))
|| (!isTopQuery && (
(IS_SIGNED_NUMERIC_TYPE(type) && pSourceItem->v.i < pItems[0].v.i) ||
(IS_UNSIGNED_NUMERIC_TYPE(type) && pSourceItem->v.u < pItems[0].v.u) ||
(IS_FLOAT_TYPE(type) && pSourceItem->v.d < pItems[0].v.d))
)) {
// replace the old data and the coresponding tuple data
STopBotResItem* pItem = &pItems[0];
pItem->v = pSourceItem->v;
pItem->uid = pSourceItem->uid;
// save the data of this tuple by over writing the old data
replaceTupleData(&pItem->tuplePos, &pSourceItem->tuplePos);
taosheapadjust((void*)pItems, sizeof(STopBotResItem), 0, pEntryInfo->numOfRes - 1, (const void*)&type,
topBotResComparFn, NULL, !isTopQuery);
}
}
}
int32_t topCombine(SqlFunctionCtx* pDestCtx, SqlFunctionCtx* pSourceCtx) {
int32_t type = pDestCtx->input.pData[0]->info.type;
SResultRowEntryInfo* pSResInfo = GET_RES_INFO(pSourceCtx);
STopBotRes* pSBuf = getTopBotOutputInfo(pSourceCtx);
for (int32_t i = 0; i < pSResInfo->numOfRes; i++) {
addResult(pDestCtx, pSBuf->pItems + i, type, true);
}
return TSDB_CODE_SUCCESS;
}
int32_t bottomCombine(SqlFunctionCtx* pDestCtx, SqlFunctionCtx* pSourceCtx) {
int32_t type = pDestCtx->input.pData[0]->info.type;
SResultRowEntryInfo* pSResInfo = GET_RES_INFO(pSourceCtx);
STopBotRes* pSBuf = getTopBotOutputInfo(pSourceCtx);
for (int32_t i = 0; i < pSResInfo->numOfRes; i++) {
addResult(pDestCtx, pSBuf->pItems + i, type, false);
}
return TSDB_CODE_SUCCESS;
}
bool getSpreadFuncEnv(SFunctionNode* UNUSED_PARAM(pFunc), SFuncExecEnv* pEnv) {
pEnv->calcMemSize = sizeof(SSpreadInfo);
return true;
@ -3027,6 +3101,10 @@ int32_t spreadPartialFinalize(SqlFunctionCtx* pCtx, SSDataBlock* pBlock) {
return pResInfo->numOfRes;
}
int32_t getElapsedInfoSize() {
return (int32_t)sizeof(SElapsedInfo);
}
bool getElapsedFuncEnv(SFunctionNode* UNUSED_PARAM(pFunc), SFuncExecEnv* pEnv) {
pEnv->calcMemSize = sizeof(SElapsedInfo);
return true;
@ -3128,6 +3206,30 @@ _elapsed_over:
return TSDB_CODE_SUCCESS;
}
int32_t elapsedFunctionMerge(SqlFunctionCtx *pCtx) {
SInputColumnInfoData* pInput = &pCtx->input;
SColumnInfoData* pCol = pInput->pData[0];
ASSERT(pCol->info.type == TSDB_DATA_TYPE_BINARY);
SElapsedInfo* pInfo = GET_ROWCELL_INTERBUF(GET_RES_INFO(pCtx));
int32_t start = pInput->startRowIndex;
char* data = colDataGetData(pCol, start);
SElapsedInfo* pInputInfo = (SElapsedInfo *)varDataVal(data);
pInfo->timeUnit = pInputInfo->timeUnit;
if (pInfo->min > pInputInfo->min) {
pInfo->min = pInputInfo->min;
}
if (pInfo->max < pInputInfo->max) {
pInfo->max = pInputInfo->max;
}
SET_VAL(GET_RES_INFO(pCtx), 1, 1);
return TSDB_CODE_SUCCESS;
}
int32_t elapsedFinalize(SqlFunctionCtx* pCtx, SSDataBlock* pBlock) {
SElapsedInfo* pInfo = GET_ROWCELL_INTERBUF(GET_RES_INFO(pCtx));
double result = (double)pInfo->max - (double)pInfo->min;
@ -3136,6 +3238,24 @@ int32_t elapsedFinalize(SqlFunctionCtx* pCtx, SSDataBlock* pBlock) {
return functionFinalize(pCtx, pBlock);
}
int32_t elapsedPartialFinalize(SqlFunctionCtx* pCtx, SSDataBlock* pBlock) {
SResultRowEntryInfo* pResInfo = GET_RES_INFO(pCtx);
SElapsedInfo* pInfo = GET_ROWCELL_INTERBUF(GET_RES_INFO(pCtx));
int32_t resultBytes = getElapsedInfoSize();
char *res = taosMemoryCalloc(resultBytes + VARSTR_HEADER_SIZE, sizeof(char));
memcpy(varDataVal(res), pInfo, resultBytes);
varDataSetLen(res, resultBytes);
int32_t slotId = pCtx->pExpr->base.resSchema.slotId;
SColumnInfoData* pCol = taosArrayGet(pBlock->pDataBlock, slotId);
colDataAppend(pCol, pBlock->info.rows, res, false);
taosMemoryFree(res);
return pResInfo->numOfRes;
}
int32_t getHistogramInfoSize() {
return (int32_t)sizeof(SHistoFuncInfo) + HISTOGRAM_MAX_BINS_NUM * sizeof(SHistoFuncBin);
}
@ -3411,7 +3531,6 @@ int32_t histogramFinalize(SqlFunctionCtx* pCtx, SSDataBlock* pBlock) {
}
int32_t histogramPartialFinalize(SqlFunctionCtx* pCtx, SSDataBlock* pBlock) {
SResultRowEntryInfo* pResInfo = GET_RES_INFO(pCtx);
SHistoFuncInfo* pInfo = GET_ROWCELL_INTERBUF(GET_RES_INFO(pCtx));
int32_t resultBytes = getHistogramInfoSize();
char *res = taosMemoryCalloc(resultBytes + VARSTR_HEADER_SIZE, sizeof(char));
@ -3428,6 +3547,10 @@ int32_t histogramPartialFinalize(SqlFunctionCtx* pCtx, SSDataBlock* pBlock) {
return 1;
}
int32_t getHLLInfoSize() {
return (int32_t)sizeof(SHLLInfo);
}
bool getHLLFuncEnv(SFunctionNode* UNUSED_PARAM(pFunc), SFuncExecEnv* pEnv) {
pEnv->calcMemSize = sizeof(SHLLInfo);
return true;
@ -3553,6 +3676,27 @@ int32_t hllFunction(SqlFunctionCtx *pCtx) {
return TSDB_CODE_SUCCESS;
}
int32_t hllFunctionMerge(SqlFunctionCtx *pCtx) {
SInputColumnInfoData* pInput = &pCtx->input;
SColumnInfoData* pCol = pInput->pData[0];
ASSERT(pCol->info.type == TSDB_DATA_TYPE_BINARY);
SHLLInfo* pInfo = GET_ROWCELL_INTERBUF(GET_RES_INFO(pCtx));
int32_t start = pInput->startRowIndex;
char* data = colDataGetData(pCol, start);
SHLLInfo* pInputInfo = (SHLLInfo *)varDataVal(data);
for (int32_t k = 0; k < HLL_BUCKETS; ++k) {
if (pInfo->buckets[k] < pInputInfo->buckets[k]) {
pInfo->buckets[k] = pInputInfo->buckets[k];
}
}
SET_VAL(GET_RES_INFO(pCtx), 1, 1);
return TSDB_CODE_SUCCESS;
}
int32_t hllFinalize(SqlFunctionCtx* pCtx, SSDataBlock* pBlock) {
SResultRowEntryInfo *pInfo = GET_RES_INFO(pCtx);
@ -3565,6 +3709,24 @@ int32_t hllFinalize(SqlFunctionCtx* pCtx, SSDataBlock* pBlock) {
return functionFinalize(pCtx, pBlock);
}
int32_t hllPartialFinalize(SqlFunctionCtx* pCtx, SSDataBlock* pBlock) {
SResultRowEntryInfo* pResInfo = GET_RES_INFO(pCtx);
SHLLInfo* pInfo = GET_ROWCELL_INTERBUF(GET_RES_INFO(pCtx));
int32_t resultBytes = getHLLInfoSize();
char *res = taosMemoryCalloc(resultBytes + VARSTR_HEADER_SIZE, sizeof(char));
memcpy(varDataVal(res), pInfo, resultBytes);
varDataSetLen(res, resultBytes);
int32_t slotId = pCtx->pExpr->base.resSchema.slotId;
SColumnInfoData* pCol = taosArrayGet(pBlock->pDataBlock, slotId);
colDataAppend(pCol, pBlock->info.rows, res, false);
taosMemoryFree(res);
return pResInfo->numOfRes;
}
bool getStateFuncEnv(SFunctionNode* UNUSED_PARAM(pFunc), SFuncExecEnv* pEnv) {
pEnv->calcMemSize = sizeof(SStateInfo);
return true;

22
source/libs/index/inc/indexCache.h
View File

@ -62,25 +62,25 @@ typedef struct CacheTerm {
} CacheTerm;
//
IndexCache* indexCacheCreate(SIndex* idx, uint64_t suid, const char* colName, int8_t type);
IndexCache* idxCacheCreate(SIndex* idx, uint64_t suid, const char* colName, int8_t type);
void indexCacheForceToMerge(void* cache);
void indexCacheDestroy(void* cache);
void indexCacheBroadcast(void* cache);
void indexCacheWait(void* cache);
void idxCacheForceToMerge(void* cache);
void idxCacheDestroy(void* cache);
void idxCacheBroadcast(void* cache);
void idxCacheWait(void* cache);
Iterate* indexCacheIteratorCreate(IndexCache* cache);
Iterate* idxCacheIteratorCreate(IndexCache* cache);
void idxCacheIteratorDestroy(Iterate* iiter);
int indexCachePut(void* cache, SIndexTerm* term, uint64_t uid);
int idxCachePut(void* cache, SIndexTerm* term, uint64_t uid);
// int indexCacheGet(void *cache, uint64_t *rst);
int indexCacheSearch(void* cache, SIndexTermQuery* query, SIdxTRslt* tr, STermValueType* s);
int idxCacheSearch(void* cache, SIndexTermQuery* query, SIdxTRslt* tr, STermValueType* s);
void indexCacheRef(IndexCache* cache);
void indexCacheUnRef(IndexCache* cache);
void idxCacheRef(IndexCache* cache);
void idxCacheUnRef(IndexCache* cache);
void indexCacheDebug(IndexCache* cache);
void idxCacheDebug(IndexCache* cache);
void idxCacheDestroyImm(IndexCache* cache);
#ifdef __cplusplus

4
source/libs/index/inc/indexComm.h
View File

@ -34,11 +34,11 @@ typedef enum { MATCH, CONTINUE, BREAK } TExeCond;
typedef TExeCond (*_cache_range_compare)(void* a, void* b, int8_t type);
__compar_fn_t indexGetCompar(int8_t type);
__compar_fn_t idxGetCompar(int8_t type);
TExeCond tCompare(__compar_fn_t func, int8_t cmpType, void* a, void* b, int8_t dType);
TExeCond tDoCompare(__compar_fn_t func, int8_t cmpType, void* a, void* b);
_cache_range_compare indexGetCompare(RangeType ty);
_cache_range_compare idxGetCompare(RangeType ty);
int32_t idxConvertData(void* src, int8_t type, void** dst);
int32_t idxConvertDataToStr(void* src, int8_t type, void** dst);

24
source/libs/index/inc/indexInt.h
View File

@ -133,24 +133,24 @@ typedef struct TFileCacheKey {
} ICacheKey;
int idxFlushCacheToTFile(SIndex* sIdx, void*, bool quit);
int64_t indexAddRef(void* p);
int32_t indexRemoveRef(int64_t ref);
void indexAcquireRef(int64_t ref);
void indexReleaseRef(int64_t ref);
int64_t idxAddRef(void* p);
int32_t idxRemoveRef(int64_t ref);
void idxAcquireRef(int64_t ref);
void idxReleaseRef(int64_t ref);
int32_t indexSerialCacheKey(ICacheKey* key, char* buf);
int32_t idxSerialCacheKey(ICacheKey* key, char* buf);
// int32_t indexSerialKey(ICacheKey* key, char* buf);
// int32_t indexSerialTermKey(SIndexTerm* itm, char* buf);
#define INDEX_TYPE_CONTAIN_EXTERN_TYPE(ty, exTy) (((ty >> 4) & (exTy)) != 0)
#define IDX_TYPE_CONTAIN_EXTERN_TYPE(ty, exTy) (((ty >> 4) & (exTy)) != 0)
#define INDEX_TYPE_GET_TYPE(ty) (ty & 0x0F)
#define IDX_TYPE_GET_TYPE(ty) (ty & 0x0F)
#define INDEX_TYPE_ADD_EXTERN_TYPE(ty, exTy) \
do { \
uint8_t oldTy = ty; \
ty = (ty >> 4) | exTy; \
ty = (ty << 4) | oldTy; \
#define IDX_TYPE_ADD_EXTERN_TYPE(ty, exTy) \
do { \
uint8_t oldTy = ty; \
ty = (ty >> 4) | exTy; \
ty = (ty << 4) | oldTy; \
} while (0)
#ifdef __cplusplus

8
source/libs/index/inc/indexTfile.h
View File

@ -117,10 +117,10 @@ int tfileWriterPut(TFileWriter* tw, void* data, bool order);
int tfileWriterFinish(TFileWriter* tw);
//
IndexTFile* indexTFileCreate(const char* path);
void indexTFileDestroy(IndexTFile* tfile);
int indexTFilePut(void* tfile, SIndexTerm* term, uint64_t uid);
int indexTFileSearch(void* tfile, SIndexTermQuery* query, SIdxTRslt* tr);
IndexTFile* idxTFileCreate(const char* path);
void idxTFileDestroy(IndexTFile* tfile);
int idxTFilePut(void* tfile, SIndexTerm* term, uint64_t uid);
int idxTFileSearch(void* tfile, SIndexTermQuery* query, SIdxTRslt* tr);
Iterate* tfileIteratorCreate(TFileReader* reader);
void tfileIteratorDestroy(Iterate* iterator);

68
source/libs/index/src/index.c
View File

@ -90,7 +90,7 @@ static void idxMergeCacheAndTFile(SArray* result, IterateValue* icache, IterateV
// static int32_t indexSerialTermKey(SIndexTerm* itm, char* buf);
// int32_t indexSerialKey(ICacheKey* key, char* buf);
static void indexPost(void* idx) {
static void idxPost(void* idx) {
SIndex* pIdx = idx;
tsem_post(&pIdx->sem);
}
@ -106,8 +106,8 @@ int indexOpen(SIndexOpts* opts, const char* path, SIndex** index) {
return -1;
}
// sIdx->cache = (void*)indexCacheCreate(sIdx);
sIdx->tindex = indexTFileCreate(path);
// sIdx->cache = (void*)idxCacheCreate(sIdx);
sIdx->tindex = idxTFileCreate(path);
if (sIdx->tindex == NULL) {
goto END;
}
@ -118,8 +118,8 @@ int indexOpen(SIndexOpts* opts, const char* path, SIndex** index) {
taosThreadMutexInit(&sIdx->mtx, NULL);
tsem_init(&sIdx->sem, 0, 0);
sIdx->refId = indexAddRef(sIdx);
indexAcquireRef(sIdx->refId);
sIdx->refId = idxAddRef(sIdx);
idxAcquireRef(sIdx->refId);
*index = sIdx;
return 0;
@ -136,7 +136,7 @@ void indexDestroy(void* handle) {
SIndex* sIdx = handle;
taosThreadMutexDestroy(&sIdx->mtx);
tsem_destroy(&sIdx->sem);
indexTFileDestroy(sIdx->tindex);
idxTFileDestroy(sIdx->tindex);
taosMemoryFree(sIdx->path);
taosMemoryFree(sIdx);
return;
@ -147,33 +147,33 @@ void indexClose(SIndex* sIdx) {
void* iter = taosHashIterate(sIdx->colObj, NULL);
while (iter) {
IndexCache** pCache = iter;
indexCacheForceToMerge((void*)(*pCache));
idxCacheForceToMerge((void*)(*pCache));
indexInfo("%s wait to merge", (*pCache)->colName);
indexWait((void*)(sIdx));
indexInfo("%s finish to wait", (*pCache)->colName);
iter = taosHashIterate(sIdx->colObj, iter);
indexCacheUnRef(*pCache);
idxCacheUnRef(*pCache);
}
taosHashCleanup(sIdx->colObj);
sIdx->colObj = NULL;
}
indexReleaseRef(sIdx->refId);
indexRemoveRef(sIdx->refId);
idxReleaseRef(sIdx->refId);
idxRemoveRef(sIdx->refId);
}
int64_t indexAddRef(void* p) {
int64_t idxAddRef(void* p) {
// impl
return taosAddRef(indexRefMgt, p);
}
int32_t indexRemoveRef(int64_t ref) {
int32_t idxRemoveRef(int64_t ref) {
// impl later
return taosRemoveRef(indexRefMgt, ref);
}
void indexAcquireRef(int64_t ref) {
void idxAcquireRef(int64_t ref) {
// impl
taosAcquireRef(indexRefMgt, ref);
}
void indexReleaseRef(int64_t ref) {
void idxReleaseRef(int64_t ref) {
// impl
taosReleaseRef(indexRefMgt, ref);
}
@ -186,11 +186,11 @@ int indexPut(SIndex* index, SIndexMultiTerm* fVals, uint64_t uid) {
char buf[128] = {0};
ICacheKey key = {.suid = p->suid, .colName = p->colName, .nColName = strlen(p->colName), .colType = p->colType};
int32_t sz = indexSerialCacheKey(&key, buf);
int32_t sz = idxSerialCacheKey(&key, buf);
IndexCache** cache = taosHashGet(index->colObj, buf, sz);
if (cache == NULL) {
IndexCache* pCache = indexCacheCreate(index, p->suid, p->colName, p->colType);
IndexCache* pCache = idxCacheCreate(index, p->suid, p->colName, p->colType);
taosHashPut(index->colObj, buf, sz, &pCache, sizeof(void*));
}
}
@ -201,12 +201,12 @@ int indexPut(SIndex* index, SIndexMultiTerm* fVals, uint64_t uid) {
char buf[128] = {0};
ICacheKey key = {.suid = p->suid, .colName = p->colName, .nColName = strlen(p->colName), .colType = p->colType};
int32_t sz = indexSerialCacheKey(&key, buf);
int32_t sz = idxSerialCacheKey(&key, buf);
indexDebug("w suid: %" PRIu64 ", colName: %s, colType: %d", key.suid, key.colName, key.colType);
IndexCache** cache = taosHashGet(index->colObj, buf, sz);
assert(*cache != NULL);
int ret = indexCachePut(*cache, p, uid);
int ret = idxCachePut(*cache, p, uid);
if (ret != 0) {
return ret;
}
@ -289,7 +289,7 @@ SIndexTerm* indexTermCreate(int64_t suid, SIndexOperOnColumn oper, uint8_t colTy
tm->nColName = nColName;
char* buf = NULL;
int32_t len = idxConvertDataToStr((void*)colVal, INDEX_TYPE_GET_TYPE(colType), (void**)&buf);
int32_t len = idxConvertDataToStr((void*)colVal, IDX_TYPE_GET_TYPE(colType), (void**)&buf);
assert(len != -1);
tm->colVal = buf;
@ -331,7 +331,7 @@ static int idxTermSearch(SIndex* sIdx, SIndexTermQuery* query, SArray** result)
ICacheKey key = {
.suid = term->suid, .colName = term->colName, .nColName = strlen(term->colName), .colType = term->colType};
indexDebug("r suid: %" PRIu64 ", colName: %s, colType: %d", key.suid, key.colName, key.colType);
int32_t sz = indexSerialCacheKey(&key, buf);
int32_t sz = idxSerialCacheKey(&key, buf);
taosThreadMutexLock(&sIdx->mtx);
IndexCache** pCache = taosHashGet(sIdx->colObj, buf, sz);
@ -345,14 +345,14 @@ static int idxTermSearch(SIndex* sIdx, SIndexTermQuery* query, SArray** result)
int64_t st = taosGetTimestampUs();
SIdxTRslt* tr = idxTRsltCreate();
if (0 == indexCacheSearch(cache, query, tr, &s)) {
if (0 == idxCacheSearch(cache, query, tr, &s)) {
if (s == kTypeDeletion) {
indexInfo("col: %s already drop by", term->colName);
// coloum already drop by other oper, no need to query tindex
return 0;
} else {
st = taosGetTimestampUs();
if (0 != indexTFileSearch(sIdx->tindex, query, tr)) {
if (0 != idxTFileSearch(sIdx->tindex, query, tr)) {
indexError("corrupt at index(TFile) col:%s val: %s", term->colName, term->colVal);
goto END;
}
@ -465,23 +465,23 @@ int idxFlushCacheToTFile(SIndex* sIdx, void* cache, bool quit) {
IndexCache* pCache = (IndexCache*)cache;
while (quit && atomic_load_32(&pCache->merging) == 1) {
}
while (quit && atomic_load_32(&pCache->merging) == 1)
;
TFileReader* pReader = tfileGetReaderByCol(sIdx->tindex, pCache->suid, pCache->colName);
if (pReader == NULL) {
indexWarn("empty tfile reader found");
}
// handle flush
Iterate* cacheIter = indexCacheIteratorCreate(pCache);
Iterate* cacheIter = idxCacheIteratorCreate(pCache);
if (cacheIter == NULL) {
indexError("%p immtable is empty, ignore merge opera", pCache);
idxCacheDestroyImm(pCache);
tfileReaderUnRef(pReader);
atomic_store_32(&pCache->merging, 0);
if (quit) {
indexPost(sIdx);
idxPost(sIdx);
}
indexReleaseRef(sIdx->refId);
idxReleaseRef(sIdx->refId);
return 0;
}
@ -532,7 +532,7 @@ int idxFlushCacheToTFile(SIndex* sIdx, void* cache, bool quit) {
tfileIteratorDestroy(tfileIter);
tfileReaderUnRef(pReader);
indexCacheUnRef(pCache);
idxCacheUnRef(pCache);
int64_t cost = taosGetTimestampUs() - st;
if (ret != 0) {
@ -542,9 +542,9 @@ int idxFlushCacheToTFile(SIndex* sIdx, void* cache, bool quit) {
}
atomic_store_32(&pCache->merging, 0);
if (quit) {
indexPost(sIdx);
idxPost(sIdx);
}
indexReleaseRef(sIdx->refId);
idxReleaseRef(sIdx->refId);
return ret;
}
@ -561,7 +561,7 @@ void iterateValueDestroy(IterateValue* value, bool destroy) {
value->colVal = NULL;
}
static int64_t indexGetAvaialbleVer(SIndex* sIdx, IndexCache* cache) {
static int64_t idxGetAvailableVer(SIndex* sIdx, IndexCache* cache) {
ICacheKey key = {.suid = cache->suid, .colName = cache->colName, .nColName = strlen(cache->colName)};
int64_t ver = CACHE_VERSION(cache);
@ -579,7 +579,7 @@ static int64_t indexGetAvaialbleVer(SIndex* sIdx, IndexCache* cache) {
return ver;
}
static int idxGenTFile(SIndex* sIdx, IndexCache* cache, SArray* batch) {
int64_t version = indexGetAvaialbleVer(sIdx, cache);
int64_t version = idxGetAvailableVer(sIdx, cache);
indexInfo("file name version: %" PRId64 "", version);
uint8_t colType = cache->type;
@ -620,8 +620,8 @@ END:
return -1;
}
int32_t indexSerialCacheKey(ICacheKey* key, char* buf) {
bool hasJson = INDEX_TYPE_CONTAIN_EXTERN_TYPE(key->colType, TSDB_DATA_TYPE_JSON);
int32_t idxSerialCacheKey(ICacheKey* key, char* buf) {
bool hasJson = IDX_TYPE_CONTAIN_EXTERN_TYPE(key->colType, TSDB_DATA_TYPE_JSON);
char* p = buf;
char tbuf[65] = {0};

104
source/libs/index/src/indexCache.c
View File

@ -68,7 +68,7 @@ static int32_t (*cacheSearch[][QUERY_MAX])(void* cache, SIndexTerm* ct, SIdxTRsl
cacheSearchLessThan_JSON, cacheSearchLessEqual_JSON, cacheSearchGreaterThan_JSON, cacheSearchGreaterEqual_JSON,
cacheSearchRange_JSON}};
static void doMergeWork(SSchedMsg* msg);
static void idxDoMergeWork(SSchedMsg* msg);
static bool idxCacheIteratorNext(Iterate* itera);
static int32_t cacheSearchTerm(void* cache, SIndexTerm* term, SIdxTRslt* tr, STermValueType* s) {
@ -127,7 +127,7 @@ static int32_t cacheSearchCompareFunc(void* cache, SIndexTerm* term, SIdxTRslt*
MemTable* mem = cache;
IndexCache* pCache = mem->pCache;
_cache_range_compare cmpFn = indexGetCompare(type);
_cache_range_compare cmpFn = idxGetCompare(type);
CacheTerm* pCt = taosMemoryCalloc(1, sizeof(CacheTerm));
pCt->colVal = term->colVal;
@ -187,7 +187,7 @@ static int32_t cacheSearchTerm_JSON(void* cache, SIndexTerm* term, SIdxTRslt* tr
pCt->version = atomic_load_64(&pCache->version);
char* exBuf = NULL;
if (INDEX_TYPE_CONTAIN_EXTERN_TYPE(term->colType, TSDB_DATA_TYPE_JSON)) {
if (IDX_TYPE_CONTAIN_EXTERN_TYPE(term->colType, TSDB_DATA_TYPE_JSON)) {
exBuf = idxPackJsonData(term);
pCt->colVal = exBuf;
}
@ -257,7 +257,7 @@ static int32_t cacheSearchCompareFunc_JSON(void* cache, SIndexTerm* term, SIdxTR
if (cache == NULL) {
return 0;
}
_cache_range_compare cmpFn = indexGetCompare(type);
_cache_range_compare cmpFn = idxGetCompare(type);
MemTable* mem = cache;
IndexCache* pCache = mem->pCache;
@ -266,7 +266,7 @@ static int32_t cacheSearchCompareFunc_JSON(void* cache, SIndexTerm* term, SIdxTR
pCt->colVal = term->colVal;
pCt->version = atomic_load_64(&pCache->version);
int8_t dType = INDEX_TYPE_GET_TYPE(term->colType);
int8_t dType = IDX_TYPE_GET_TYPE(term->colType);
int skip = 0;
char* exBuf = NULL;
if (type == CONTAINS) {
@ -331,9 +331,9 @@ static int32_t cacheSearchRange(void* cache, SIndexTerm* term, SIdxTRslt* tr, ST
// impl later
return 0;
}
static IterateValue* indexCacheIteratorGetValue(Iterate* iter);
static IterateValue* idxCacheIteratorGetValue(Iterate* iter);
IndexCache* indexCacheCreate(SIndex* idx, uint64_t suid, const char* colName, int8_t type) {
IndexCache* idxCacheCreate(SIndex* idx, uint64_t suid, const char* colName, int8_t type) {
IndexCache* cache = taosMemoryCalloc(1, sizeof(IndexCache));
if (cache == NULL) {
indexError("failed to create index cache");
@ -342,7 +342,7 @@ IndexCache* indexCacheCreate(SIndex* idx, uint64_t suid, const char* colName, in
cache->mem = idxInternalCacheCreate(type);
cache->mem->pCache = cache;
cache->colName = INDEX_TYPE_CONTAIN_EXTERN_TYPE(type, TSDB_DATA_TYPE_JSON) ? tstrdup(JSON_COLUMN) : tstrdup(colName);
cache->colName = IDX_TYPE_CONTAIN_EXTERN_TYPE(type, TSDB_DATA_TYPE_JSON) ? tstrdup(JSON_COLUMN) : tstrdup(colName);
cache->type = type;
cache->index = idx;
cache->version = 0;
@ -352,13 +352,13 @@ IndexCache* indexCacheCreate(SIndex* idx, uint64_t suid, const char* colName, in
taosThreadMutexInit(&cache->mtx, NULL);
taosThreadCondInit(&cache->finished, NULL);
indexCacheRef(cache);
idxCacheRef(cache);
if (idx != NULL) {
indexAcquireRef(idx->refId);
idxAcquireRef(idx->refId);
}
return cache;
}
void indexCacheDebug(IndexCache* cache) {
void idxCacheDebug(IndexCache* cache) {
MemTable* tbl = NULL;
taosThreadMutexLock(&cache->mtx);
@ -405,7 +405,7 @@ void indexCacheDebug(IndexCache* cache) {
}
}
void indexCacheDestroySkiplist(SSkipList* slt) {
void idxCacheDestroySkiplist(SSkipList* slt) {
SSkipListIterator* iter = tSkipListCreateIter(slt);
while (iter != NULL && tSkipListIterNext(iter)) {
SSkipListNode* node = tSkipListIterGet(iter);
@ -418,11 +418,11 @@ void indexCacheDestroySkiplist(SSkipList* slt) {
tSkipListDestroyIter(iter);
tSkipListDestroy(slt);
}
void indexCacheBroadcast(void* cache) {
void idxCacheBroadcast(void* cache) {
IndexCache* pCache = cache;
taosThreadCondBroadcast(&pCache->finished);
}
void indexCacheWait(void* cache) {
void idxCacheWait(void* cache) {
IndexCache* pCache = cache;
taosThreadCondWait(&pCache->finished, &pCache->mtx);
}
@ -435,14 +435,14 @@ void idxCacheDestroyImm(IndexCache* cache) {
tbl = cache->imm;
cache->imm = NULL; // or throw int bg thread
indexCacheBroadcast(cache);
idxCacheBroadcast(cache);
taosThreadMutexUnlock(&cache->mtx);
idxMemUnRef(tbl);
idxMemUnRef(tbl);
}
void indexCacheDestroy(void* cache) {
void idxCacheDestroy(void* cache) {
IndexCache* pCache = cache;
if (pCache == NULL) {
return;
@ -455,12 +455,12 @@ void indexCacheDestroy(void* cache) {
taosThreadMutexDestroy(&pCache->mtx);
taosThreadCondDestroy(&pCache->finished);
if (pCache->index != NULL) {
indexReleaseRef(((SIndex*)pCache->index)->refId);
idxReleaseRef(((SIndex*)pCache->index)->refId);
}
taosMemoryFree(pCache);
}
Iterate* indexCacheIteratorCreate(IndexCache* cache) {
Iterate* idxCacheIteratorCreate(IndexCache* cache) {
if (cache->imm == NULL) {
return NULL;
}
@ -477,7 +477,7 @@ Iterate* indexCacheIteratorCreate(IndexCache* cache) {
iiter->val.colVal = NULL;
iiter->iter = tbl != NULL ? tSkipListCreateIter(tbl->mem) : NULL;
iiter->next = idxCacheIteratorNext;
iiter->getValue = indexCacheIteratorGetValue;
iiter->getValue = idxCacheIteratorGetValue;
taosThreadMutexUnlock(&cache->mtx);
@ -492,30 +492,30 @@ void idxCacheIteratorDestroy(Iterate* iter) {
taosMemoryFree(iter);
}
int indexCacheSchedToMerge(IndexCache* pCache, bool notify) {
int idxCacheSchedToMerge(IndexCache* pCache, bool notify) {
SSchedMsg schedMsg = {0};
schedMsg.fp = doMergeWork;
schedMsg.fp = idxDoMergeWork;
schedMsg.ahandle = pCache;
if (notify) {
schedMsg.thandle = taosMemoryMalloc(1);
}
schedMsg.msg = NULL;
indexAcquireRef(pCache->index->refId);
idxAcquireRef(pCache->index->refId);
taosScheduleTask(indexQhandle, &schedMsg);
return 0;
}
static void indexCacheMakeRoomForWrite(IndexCache* cache) {
static void idxCacheMakeRoomForWrite(IndexCache* cache) {
while (true) {
if (cache->occupiedMem * MEM_ESTIMATE_RADIO < MEM_THRESHOLD) {
break;
} else if (cache->imm != NULL) {
// TODO: wake up by condition variable
indexCacheWait(cache);
idxCacheWait(cache);
} else {
bool quit = cache->occupiedMem >= MEM_SIGNAL_QUIT ? true : false;
indexCacheRef(cache);
idxCacheRef(cache);
cache->imm = cache->mem;
cache->mem = idxInternalCacheCreate(cache->type);
cache->mem->pCache = cache;
@ -525,18 +525,18 @@ static void indexCacheMakeRoomForWrite(IndexCache* cache) {
}
// sched to merge
// unref cache in bgwork
indexCacheSchedToMerge(cache, quit);
idxCacheSchedToMerge(cache, quit);
}
}
}
int indexCachePut(void* cache, SIndexTerm* term, uint64_t uid) {
int idxCachePut(void* cache, SIndexTerm* term, uint64_t uid) {
if (cache == NULL) {
return -1;
}
bool hasJson = INDEX_TYPE_CONTAIN_EXTERN_TYPE(term->colType, TSDB_DATA_TYPE_JSON);
bool hasJson = IDX_TYPE_CONTAIN_EXTERN_TYPE(term->colType, TSDB_DATA_TYPE_JSON);
IndexCache* pCache = cache;
indexCacheRef(pCache);
idxCacheRef(pCache);
// encode data
CacheTerm* ct = taosMemoryCalloc(1, sizeof(CacheTerm));
if (cache == NULL) {
@ -559,7 +559,7 @@ int indexCachePut(void* cache, SIndexTerm* term, uint64_t uid) {
taosThreadMutexLock(&pCache->mtx);
pCache->occupiedMem += estimate;
indexCacheMakeRoomForWrite(pCache);
idxCacheMakeRoomForWrite(pCache);
MemTable* tbl = pCache->mem;
idxMemRef(tbl);
tSkipListPut(tbl->mem, (char*)ct);
@ -567,29 +567,29 @@ int indexCachePut(void* cache, SIndexTerm* term, uint64_t uid) {
taosThreadMutexUnlock(&pCache->mtx);
indexCacheUnRef(pCache);
idxCacheUnRef(pCache);
return 0;
// encode end
}
void indexCacheForceToMerge(void* cache) {
void idxCacheForceToMerge(void* cache) {
IndexCache* pCache = cache;
indexCacheRef(pCache);
idxCacheRef(pCache);
taosThreadMutexLock(&pCache->mtx);
indexInfo("%p is forced to merge into tfile", pCache);
pCache->occupiedMem += MEM_SIGNAL_QUIT;
indexCacheMakeRoomForWrite(pCache);
idxCacheMakeRoomForWrite(pCache);
taosThreadMutexUnlock(&pCache->mtx);
indexCacheUnRef(pCache);
idxCacheUnRef(pCache);
return;
}
int indexCacheDel(void* cache, const char* fieldValue, int32_t fvlen, uint64_t uid, int8_t operType) {
int idxCacheDel(void* cache, const char* fieldValue, int32_t fvlen, uint64_t uid, int8_t operType) {
IndexCache* pCache = cache;
return 0;
}
static int32_t indexQueryMem(MemTable* mem, SIndexTermQuery* query, SIdxTRslt* tr, STermValueType* s) {
static int32_t idxQueryMem(MemTable* mem, SIndexTermQuery* query, SIdxTRslt* tr, STermValueType* s) {
if (mem == NULL) {
return 0;
}
@ -597,13 +597,13 @@ static int32_t indexQueryMem(MemTable* mem, SIndexTermQuery* query, SIdxTRslt* t
SIndexTerm* term = query->term;
EIndexQueryType qtype = query->qType;
if (INDEX_TYPE_CONTAIN_EXTERN_TYPE(term->colType, TSDB_DATA_TYPE_JSON)) {
if (IDX_TYPE_CONTAIN_EXTERN_TYPE(term->colType, TSDB_DATA_TYPE_JSON)) {
return cacheSearch[1][qtype](mem, term, tr, s);
} else {
return cacheSearch[0][qtype](mem, term, tr, s);
}
}
int indexCacheSearch(void* cache, SIndexTermQuery* query, SIdxTRslt* result, STermValueType* s) {
int idxCacheSearch(void* cache, SIndexTermQuery* query, SIdxTRslt* result, STermValueType* s) {
int64_t st = taosGetTimestampUs();
if (cache == NULL) {
return 0;
@ -618,10 +618,10 @@ int indexCacheSearch(void* cache, SIndexTermQuery* query, SIdxTRslt* result, STe
idxMemRef(imm);
taosThreadMutexUnlock(&pCache->mtx);
int ret = (mem && mem->mem) ? indexQueryMem(mem, query, result, s) : 0;
int ret = (mem && mem->mem) ? idxQueryMem(mem, query, result, s) : 0;
if (ret == 0 && *s != kTypeDeletion) {
// continue search in imm
ret = (imm && imm->mem) ? indexQueryMem(imm, query, result, s) : 0;
ret = (imm && imm->mem) ? idxQueryMem(imm, query, result, s) : 0;
}
idxMemUnRef(mem);
@ -631,20 +631,20 @@ int indexCacheSearch(void* cache, SIndexTermQuery* query, SIdxTRslt* result, STe
return ret;
}
void indexCacheRef(IndexCache* cache) {
void idxCacheRef(IndexCache* cache) {
if (cache == NULL) {
return;
}
int ref = T_REF_INC(cache);
UNUSED(ref);
}
void indexCacheUnRef(IndexCache* cache) {
void idxCacheUnRef(IndexCache* cache) {
if (cache == NULL) {
return;
}
int ref = T_REF_DEC(cache);
if (ref == 0) {
indexCacheDestroy(cache);
idxCacheDestroy(cache);
}
}
@ -662,7 +662,7 @@ void idxMemUnRef(MemTable* tbl) {
int ref = T_REF_DEC(tbl);
if (ref == 0) {
SSkipList* slt = tbl->mem;
indexCacheDestroySkiplist(slt);
idxCacheDestroySkiplist(slt);
taosMemoryFree(tbl);
}
}
@ -693,15 +693,15 @@ static int32_t idxCacheTermCompare(const void* l, const void* r) {
return cmp;
}
static int indexFindCh(char* a, char c) {
static int idxFindCh(char* a, char c) {
char* p = a;
while (*p != 0 && *p++ != c) {
}
return p - a;
}
static int idxCacheJsonTermCompareImpl(char* a, char* b) {
// int alen = indexFindCh(a, '&');
// int blen = indexFindCh(b, '&');
// int alen = idxFindCh(a, '&');
// int blen = idxFindCh(b, '&');
// int cmp = strncmp(a, b, MIN(alen, blen));
// if (cmp == 0) {
@ -730,9 +730,9 @@ static int32_t idxCacheJsonTermCompare(const void* l, const void* r) {
return cmp;
}
static MemTable* idxInternalCacheCreate(int8_t type) {
int ttype = INDEX_TYPE_CONTAIN_EXTERN_TYPE(type, TSDB_DATA_TYPE_JSON) ? TSDB_DATA_TYPE_BINARY : TSDB_DATA_TYPE_BINARY;
int ttype = IDX_TYPE_CONTAIN_EXTERN_TYPE(type, TSDB_DATA_TYPE_JSON) ? TSDB_DATA_TYPE_BINARY : TSDB_DATA_TYPE_BINARY;
int32_t (*cmpFn)(const void* l, const void* r) =
INDEX_TYPE_CONTAIN_EXTERN_TYPE(type, TSDB_DATA_TYPE_JSON) ? idxCacheJsonTermCompare : idxCacheTermCompare;
IDX_TYPE_CONTAIN_EXTERN_TYPE(type, TSDB_DATA_TYPE_JSON) ? idxCacheJsonTermCompare : idxCacheTermCompare;
MemTable* tbl = taosMemoryCalloc(1, sizeof(MemTable));
idxMemRef(tbl);
@ -742,7 +742,7 @@ static MemTable* idxInternalCacheCreate(int8_t type) {
return tbl;
}
static void doMergeWork(SSchedMsg* msg) {
static void idxDoMergeWork(SSchedMsg* msg) {
IndexCache* pCache = msg->ahandle;
SIndex* sidx = (SIndex*)pCache->index;
@ -771,7 +771,7 @@ static bool idxCacheIteratorNext(Iterate* itera) {
return next;
}
static IterateValue* indexCacheIteratorGetValue(Iterate* iter) {
static IterateValue* idxCacheIteratorGetValue(Iterate* iter) {
// opt later
return &iter->val;
}

22
source/libs/index/src/indexComm.c
View File

@ -75,35 +75,35 @@ char* idxInt2str(int64_t val, char* dst, int radix) {
;
return dst - 1;
}
__compar_fn_t indexGetCompar(int8_t type) {
__compar_fn_t idxGetCompar(int8_t type) {
if (type == TSDB_DATA_TYPE_BINARY || type == TSDB_DATA_TYPE_NCHAR) {
return (__compar_fn_t)strcmp;
}
return getComparFunc(type, 0);
}
static TExeCond tCompareLessThan(void* a, void* b, int8_t type) {
__compar_fn_t func = indexGetCompar(type);
__compar_fn_t func = idxGetCompar(type);
return tCompare(func, QUERY_LESS_THAN, a, b, type);
}
static TExeCond tCompareLessEqual(void* a, void* b, int8_t type) {
__compar_fn_t func = indexGetCompar(type);
__compar_fn_t func = idxGetCompar(type);
return tCompare(func, QUERY_LESS_EQUAL, a, b, type);
}
static TExeCond tCompareGreaterThan(void* a, void* b, int8_t type) {
__compar_fn_t func = indexGetCompar(type);
__compar_fn_t func = idxGetCompar(type);
return tCompare(func, QUERY_GREATER_THAN, a, b, type);
}
static TExeCond tCompareGreaterEqual(void* a, void* b, int8_t type) {
__compar_fn_t func = indexGetCompar(type);
__compar_fn_t func = idxGetCompar(type);
return tCompare(func, QUERY_GREATER_EQUAL, a, b, type);
}
static TExeCond tCompareContains(void* a, void* b, int8_t type) {
__compar_fn_t func = indexGetCompar(type);
__compar_fn_t func = idxGetCompar(type);
return tCompare(func, QUERY_TERM, a, b, type);
}
static TExeCond tCompareEqual(void* a, void* b, int8_t type) {
__compar_fn_t func = indexGetCompar(type);
__compar_fn_t func = idxGetCompar(type);
return tCompare(func, QUERY_TERM, a, b, type);
}
TExeCond tCompare(__compar_fn_t func, int8_t cmptype, void* a, void* b, int8_t dtype) {
@ -205,14 +205,14 @@ TExeCond tDoCompare(__compar_fn_t func, int8_t comparType, void* a, void* b) {
static TExeCond (*rangeCompare[])(void* a, void* b, int8_t type) = {
tCompareLessThan, tCompareLessEqual, tCompareGreaterThan, tCompareGreaterEqual, tCompareContains, tCompareEqual};
_cache_range_compare indexGetCompare(RangeType ty) { return rangeCompare[ty]; }
_cache_range_compare idxGetCompare(RangeType ty) { return rangeCompare[ty]; }
char* idxPackJsonData(SIndexTerm* itm) {
/*
* |<-----colname---->|<-----dataType---->|<--------colVal---------->|
* |<-----string----->|<-----uint8_t----->|<----depend on dataType-->|
*/
uint8_t ty = INDEX_TYPE_GET_TYPE(itm->colType);
uint8_t ty = IDX_TYPE_GET_TYPE(itm->colType);
int32_t sz = itm->nColName + itm->nColVal + sizeof(uint8_t) + sizeof(JSON_VALUE_DELIM) * 2 + 1;
char* buf = (char*)taosMemoryCalloc(1, sz);
@ -240,7 +240,7 @@ char* idxPackJsonDataPrefix(SIndexTerm* itm, int32_t* skip) {
* |<-----colname---->|<-----dataType---->|<--------colVal---------->|
* |<-----string----->|<-----uint8_t----->|<----depend on dataType-->|
*/
uint8_t ty = INDEX_TYPE_GET_TYPE(itm->colType);
uint8_t ty = IDX_TYPE_GET_TYPE(itm->colType);
int32_t sz = itm->nColName + itm->nColVal + sizeof(uint8_t) + sizeof(JSON_VALUE_DELIM) * 2 + 1;
char* buf = (char*)taosMemoryCalloc(1, sz);
@ -267,7 +267,7 @@ char* idxPackJsonDataPrefixNoType(SIndexTerm* itm, int32_t* skip) {
* |<-----colname---->|<-----dataType---->|<--------colVal---------->|
* |<-----string----->|<-----uint8_t----->|<----depend on dataType-->|
*/
uint8_t ty = INDEX_TYPE_GET_TYPE(itm->colType);
uint8_t ty = IDX_TYPE_GET_TYPE(itm->colType);
int32_t sz = itm->nColName + itm->nColVal + sizeof(uint8_t) + sizeof(JSON_VALUE_DELIM) * 2 + 1;
char* buf = (char*)taosMemoryCalloc(1, sz);

2
source/libs/index/src/indexFilter.c
View File

@ -318,7 +318,7 @@ int sifLessThan(void *a, void *b, int16_t dtype) {
}
int sifEqual(void *a, void *b, int16_t dtype) {
__compar_fn_t func = getComparFunc(dtype, 0);
//__compar_fn_t func = indexGetCompar(dtype);
//__compar_fn_t func = idxGetCompar(dtype);
return (int)tDoCompare(func, QUERY_TERM, a, b);
}
static Filter sifGetFilterFunc(EIndexQueryType type, bool *reverse) {

4
source/libs/index/src/indexJson.c
View File

@ -30,7 +30,7 @@ int indexJsonPut(SIndexJson *index, SIndexJsonMultiTerm *terms, uint64_t uid) {
} else {
p->colType = TSDB_DATA_TYPE_DOUBLE;
}
INDEX_TYPE_ADD_EXTERN_TYPE(p->colType, TSDB_DATA_TYPE_JSON);
IDX_TYPE_ADD_EXTERN_TYPE(p->colType, TSDB_DATA_TYPE_JSON);
}
// handle put
return indexPut(index, terms, uid);
@ -48,7 +48,7 @@ int indexJsonSearch(SIndexJson *index, SIndexJsonMultiTermQuery *tq, SArray *res
} else {
p->colType = TSDB_DATA_TYPE_DOUBLE;
}
INDEX_TYPE_ADD_EXTERN_TYPE(p->colType, TSDB_DATA_TYPE_JSON);
IDX_TYPE_ADD_EXTERN_TYPE(p->colType, TSDB_DATA_TYPE_JSON);
}
// handle search
return indexSearch(index, tq, result);

28
source/libs/index/src/indexTfile.c
View File

@ -118,7 +118,7 @@ TFileCache* tfileCacheCreate(const char* path) {
ICacheKey key = {.suid = header->suid, .colName = header->colName, .nColName = (int32_t)strlen(header->colName)};
char buf[128] = {0};
int32_t sz = indexSerialCacheKey(&key, buf);
int32_t sz = idxSerialCacheKey(&key, buf);
assert(sz < sizeof(buf));
taosHashPut(tcache->tableCache, buf, sz, &reader, sizeof(void*));
tfileReaderRef(reader);
@ -149,7 +149,7 @@ void tfileCacheDestroy(TFileCache* tcache) {
TFileReader* tfileCacheGet(TFileCache* tcache, ICacheKey* key) {
char buf[128] = {0};
int32_t sz = indexSerialCacheKey(key, buf);
int32_t sz = idxSerialCacheKey(key, buf);
assert(sz < sizeof(buf));
TFileReader** reader = taosHashGet(tcache->tableCache, buf, sz);
if (reader == NULL || *reader == NULL) {
@ -161,7 +161,7 @@ TFileReader* tfileCacheGet(TFileCache* tcache, ICacheKey* key) {
}
void tfileCachePut(TFileCache* tcache, ICacheKey* key, TFileReader* reader) {
char buf[128] = {0};
int32_t sz = indexSerialCacheKey(key, buf);
int32_t sz = idxSerialCacheKey(key, buf);
// remove last version index reader
TFileReader** p = taosHashGet(tcache->tableCache, buf, sz);
if (p != NULL && *p != NULL) {
@ -281,7 +281,7 @@ static int32_t tfSearchSuffix(void* reader, SIndexTerm* tem, SIdxTRslt* tr) {
return 0;
}
static int32_t tfSearchRegex(void* reader, SIndexTerm* tem, SIdxTRslt* tr) {
bool hasJson = INDEX_TYPE_CONTAIN_EXTERN_TYPE(tem->colType, TSDB_DATA_TYPE_JSON);
bool hasJson = IDX_TYPE_CONTAIN_EXTERN_TYPE(tem->colType, TSDB_DATA_TYPE_JSON);
int ret = 0;
char* p = tem->colVal;
@ -305,7 +305,7 @@ static int32_t tfSearchCompareFunc(void* reader, SIndexTerm* tem, SIdxTRslt* tr,
int ret = 0;
char* p = tem->colVal;
int skip = 0;
_cache_range_compare cmpFn = indexGetCompare(type);
_cache_range_compare cmpFn = idxGetCompare(type);
SArray* offsets = taosArrayInit(16, sizeof(uint64_t));
@ -431,7 +431,7 @@ static int32_t tfSearchCompareFunc_JSON(void* reader, SIndexTerm* tem, SIdxTRslt
p = idxPackJsonDataPrefix(tem, &skip);
}
_cache_range_compare cmpFn = indexGetCompare(ctype);
_cache_range_compare cmpFn = idxGetCompare(ctype);
SArray* offsets = taosArrayInit(16, sizeof(uint64_t));
@ -457,7 +457,7 @@ static int32_t tfSearchCompareFunc_JSON(void* reader, SIndexTerm* tem, SIdxTRslt
} else if (0 != strncmp(ch, p, skip)) {
continue;
}
cond = cmpFn(ch + skip, tem->colVal, INDEX_TYPE_GET_TYPE(tem->colType));
cond = cmpFn(ch + skip, tem->colVal, IDX_TYPE_GET_TYPE(tem->colType));
}
if (MATCH == cond) {
tfileReaderLoadTableIds((TFileReader*)reader, rt->out.out, tr->total);
@ -476,7 +476,7 @@ int tfileReaderSearch(TFileReader* reader, SIndexTermQuery* query, SIdxTRslt* tr
SIndexTerm* term = query->term;
EIndexQueryType qtype = query->qType;
int ret = 0;
if (INDEX_TYPE_CONTAIN_EXTERN_TYPE(term->colType, TSDB_DATA_TYPE_JSON)) {
if (IDX_TYPE_CONTAIN_EXTERN_TYPE(term->colType, TSDB_DATA_TYPE_JSON)) {
ret = tfSearch[1][qtype](reader, term, tr);
} else {
ret = tfSearch[0][qtype](reader, term, tr);
@ -536,7 +536,7 @@ int tfileWriterPut(TFileWriter* tw, void* data, bool order) {
__compar_fn_t fn;
int8_t colType = tw->header.colType;
colType = INDEX_TYPE_GET_TYPE(colType);
colType = IDX_TYPE_GET_TYPE(colType);
if (colType == TSDB_DATA_TYPE_BINARY || colType == TSDB_DATA_TYPE_NCHAR) {
fn = tfileStrCompare;
} else {
@ -620,7 +620,7 @@ void tfileWriterDestroy(TFileWriter* tw) {
taosMemoryFree(tw);
}
IndexTFile* indexTFileCreate(const char* path) {
IndexTFile* idxTFileCreate(const char* path) {
TFileCache* cache = tfileCacheCreate(path);
if (cache == NULL) {
return NULL;
@ -635,7 +635,7 @@ IndexTFile* indexTFileCreate(const char* path) {
tfile->cache = cache;
return tfile;
}
void indexTFileDestroy(IndexTFile* tfile) {
void idxTFileDestroy(IndexTFile* tfile) {
if (tfile == NULL) {
return;
}
@ -644,7 +644,7 @@ void indexTFileDestroy(IndexTFile* tfile) {
taosMemoryFree(tfile);
}
int indexTFileSearch(void* tfile, SIndexTermQuery* query, SIdxTRslt* result) {
int idxTFileSearch(void* tfile, SIndexTermQuery* query, SIdxTRslt* result) {
int ret = -1;
if (tfile == NULL) {
return ret;
@ -667,7 +667,7 @@ int indexTFileSearch(void* tfile, SIndexTermQuery* query, SIdxTRslt* result) {
return tfileReaderSearch(reader, query, result);
}
int indexTFilePut(void* tfile, SIndexTerm* term, uint64_t uid) {
int idxTFilePut(void* tfile, SIndexTerm* term, uint64_t uid) {
// TFileWriterOpt wOpt = {.suid = term->suid, .colType = term->colType, .colName = term->colName, .nColName =
// term->nColName, .version = 1};
@ -845,7 +845,7 @@ static int tfileWriteData(TFileWriter* write, TFileValue* tval) {
TFileHeader* header = &write->header;
uint8_t colType = header->colType;
colType = INDEX_TYPE_GET_TYPE(colType);
colType = IDX_TYPE_GET_TYPE(colType);
FstSlice key = fstSliceCreate((uint8_t*)(tval->colVal), (size_t)strlen(tval->colVal));
if (fstBuilderInsert(write->fb, key, tval->offset)) {
fstSliceDestroy(&key);

10
source/libs/index/test/indexTests.cc
View File

@ -521,10 +521,10 @@ class CacheObj {
public:
CacheObj() {
// TODO
cache = indexCacheCreate(NULL, 0, "voltage", TSDB_DATA_TYPE_BINARY);
cache = idxCacheCreate(NULL, 0, "voltage", TSDB_DATA_TYPE_BINARY);
}
int Put(SIndexTerm* term, int16_t colId, int32_t version, uint64_t uid) {
int ret = indexCachePut(cache, term, uid);
int ret = idxCachePut(cache, term, uid);
if (ret != 0) {
//
std::cout << "failed to put into cache: " << ret << std::endl;
@ -533,12 +533,12 @@ class CacheObj {
}
void Debug() {
//
indexCacheDebug(cache);
idxCacheDebug(cache);
}
int Get(SIndexTermQuery* query, int16_t colId, int32_t version, SArray* result, STermValueType* s) {
SIdxTRslt* tr = idxTRsltCreate();
int ret = indexCacheSearch(cache, query, tr, s);
int ret = idxCacheSearch(cache, query, tr, s);
idxTRsltMergeTo(tr, result);
idxTRsltDestroy(tr);
@ -549,7 +549,7 @@ class CacheObj {
}
~CacheObj() {
// TODO
indexCacheDestroy(cache);
idxCacheDestroy(cache);
}
private:

1
source/libs/planner/src/planPhysiCreater.c
View File

@ -1186,6 +1186,7 @@ static int32_t createExchangePhysiNodeByMerge(SMergePhysiNode* pMerge) {
return TSDB_CODE_OUT_OF_MEMORY;
}
pExchange->srcGroupId = pMerge->srcGroupId;
pExchange->singleChannel = true;
pExchange->node.pParent = (SPhysiNode*)pMerge;
pExchange->node.pOutputDataBlockDesc = nodesCloneNode(pMerge->node.pOutputDataBlockDesc);
if (NULL == pExchange->node.pOutputDataBlockDesc) {

2
source/libs/scalar/src/sclvector.c
View File

@ -348,7 +348,7 @@ static FORCE_INLINE void varToNchar(char* buf, SScalarParam* pOut, int32_t rowIn
int32_t outputMaxLen = (inputLen + 1) * TSDB_NCHAR_SIZE + VARSTR_HEADER_SIZE;
char* t = taosMemoryCalloc(1, outputMaxLen);
/*int32_t resLen = */taosMbsToUcs4(varDataVal(buf), inputLen, (TdUcs4*) varDataVal(t), outputMaxLen, &len);
/*int32_t resLen = */taosMbsToUcs4(varDataVal(buf), inputLen, (TdUcs4*) varDataVal(t), outputMaxLen - VARSTR_HEADER_SIZE, &len);
varDataSetLen(t, len);
colDataAppend(pOut->columnData, rowIndex, t, false);

1
source/libs/stream/src/streamData.c
View File

@ -53,6 +53,7 @@ int32_t streamDispatchReqToData(const SStreamDispatchReq* pReq, SStreamDataBlock
SSDataBlock* pDataBlock = taosArrayGet(pArray, i);
blockCompressDecode(pDataBlock, htonl(pRetrieve->numOfCols), htonl(pRetrieve->numOfRows), pRetrieve->data);
// TODO: refactor
pDataBlock->info.type = pRetrieve->streamBlockType;
pDataBlock->info.childId = pReq->sourceChildId;
}
pData->blocks = pArray;

1
source/libs/stream/src/streamDispatch.c
View File

@ -72,6 +72,7 @@ static int32_t streamAddBlockToDispatchMsg(const SSDataBlock* pBlock, SStreamDis
pRetrieve->precision = TSDB_DEFAULT_PRECISION;
pRetrieve->compressed = 0;
pRetrieve->completed = 1;
pRetrieve->streamBlockType = pBlock->info.type;
pRetrieve->numOfRows = htonl(pBlock->info.rows);
pRetrieve->numOfCols = htonl(pBlock->info.numOfCols);

1
source/libs/sync/inc/syncAppendEntries.h
View File

@ -93,6 +93,7 @@ extern "C" {
// /\ UNCHANGED <<candidateVars, leaderVars>>
//
int32_t syncNodeOnAppendEntriesCb(SSyncNode* ths, SyncAppendEntries* pMsg);
int32_t syncNodeOnAppendEntriesSnapshotCb(SSyncNode* ths, SyncAppendEntries* pMsg);
#ifdef __cplusplus
}

1
source/libs/sync/inc/syncAppendEntriesReply.h
View File

@ -41,6 +41,7 @@ extern "C" {
// /\ UNCHANGED <<serverVars, candidateVars, logVars, elections>>
//
int32_t syncNodeOnAppendEntriesReplyCb(SSyncNode* ths, SyncAppendEntriesReply* pMsg);
int32_t syncNodeOnAppendEntriesReplySnapshotCb(SSyncNode* ths, SyncAppendEntriesReply* pMsg);
#ifdef __cplusplus
}

2
source/libs/sync/inc/syncElection.h
View File

@ -39,6 +39,8 @@ extern "C" {
// /\ UNCHANGED <<serverVars, candidateVars, leaderVars, logVars>>
//
int32_t syncNodeRequestVotePeers(SSyncNode* pSyncNode);
int32_t syncNodeRequestVotePeersSnapshot(SSyncNode* pSyncNode);
int32_t syncNodeElect(SSyncNode* pSyncNode);
int32_t syncNodeRequestVote(SSyncNode* pSyncNode, const SRaftId* destRaftId, const SyncRequestVote* pMsg);

3
source/libs/sync/inc/syncIO.h
View File

@ -57,6 +57,9 @@ typedef struct SSyncIO {
int32_t (*FpOnSyncAppendEntriesReply)(SSyncNode *pSyncNode, SyncAppendEntriesReply *pMsg);
int32_t (*FpOnSyncTimeout)(SSyncNode *pSyncNode, SyncTimeout *pMsg);
int32_t (*FpOnSyncSnapshotSend)(SSyncNode *pSyncNode, SyncSnapshotSend *pMsg);
int32_t (*FpOnSyncSnapshotRsp)(SSyncNode *pSyncNode, SyncSnapshotRsp *pMsg);
int8_t isStart;
} SSyncIO;

4
source/libs/sync/inc/syncIndexMgr.h
View File

@ -30,6 +30,7 @@ extern "C" {
typedef struct SSyncIndexMgr {
SRaftId (*replicas)[TSDB_MAX_REPLICA];
SyncIndex index[TSDB_MAX_REPLICA];
SyncTerm privateTerm[TSDB_MAX_REPLICA]; // for advanced function
int32_t replicaNum;
SSyncNode *pSyncNode;
} SSyncIndexMgr;
@ -43,6 +44,9 @@ SyncIndex syncIndexMgrGetIndex(SSyncIndexMgr *pSyncIndexMgr, const SRaftId
cJSON * syncIndexMgr2Json(SSyncIndexMgr *pSyncIndexMgr);
char * syncIndexMgr2Str(SSyncIndexMgr *pSyncIndexMgr);
// void syncIndexMgrSetTerm(SSyncIndexMgr *pSyncIndexMgr, const SRaftId *pRaftId, SyncTerm term);
// SyncTerm syncIndexMgrGetTerm(SSyncIndexMgr *pSyncIndexMgr, const SRaftId *pRaftId);
// for debug -------------------
void syncIndexMgrPrint(SSyncIndexMgr *pObj);
void syncIndexMgrPrint2(char *s, SSyncIndexMgr *pObj);

53
source/libs/sync/inc/syncInt.h
View File

@ -58,6 +58,8 @@ typedef struct SSyncRespMgr SSyncRespMgr;
typedef struct SSyncSnapshotSender SSyncSnapshotSender;
typedef struct SSyncSnapshotReceiver SSyncSnapshotReceiver;
extern bool gRaftDetailLog;
typedef struct SSyncNode {
// init by SSyncInfo
SyncGroupId vgId;
@ -137,24 +139,27 @@ typedef struct SSyncNode {
uint64_t heartbeatTimerCounter;
// callback
int32_t (*FpOnPing)(SSyncNode* ths, SyncPing* pMsg);
int32_t (*FpOnPingReply)(SSyncNode* ths, SyncPingReply* pMsg);
int32_t (*FpOnClientRequest)(SSyncNode* ths, SyncClientRequest* pMsg);
int32_t (*FpOnRequestVote)(SSyncNode* ths, SyncRequestVote* pMsg);
int32_t (*FpOnRequestVoteReply)(SSyncNode* ths, SyncRequestVoteReply* pMsg);
int32_t (*FpOnAppendEntries)(SSyncNode* ths, SyncAppendEntries* pMsg);
int32_t (*FpOnAppendEntriesReply)(SSyncNode* ths, SyncAppendEntriesReply* pMsg);
int32_t (*FpOnTimeout)(SSyncNode* pSyncNode, SyncTimeout* pMsg);
FpOnPingCb FpOnPing;
FpOnPingReplyCb FpOnPingReply;
FpOnClientRequestCb FpOnClientRequest;
FpOnTimeoutCb FpOnTimeout;
FpOnRequestVoteCb FpOnRequestVote;
FpOnRequestVoteReplyCb FpOnRequestVoteReply;
FpOnAppendEntriesCb FpOnAppendEntries;
FpOnAppendEntriesReplyCb FpOnAppendEntriesReply;
FpOnSnapshotSendCb FpOnSnapshotSend;
FpOnSnapshotRspCb FpOnSnapshotRsp;
// tools
SSyncRespMgr* pSyncRespMgr;
// restore state
// sem_t restoreSem;
bool restoreFinish;
SSnapshot* pSnapshot;
SSyncSnapshotSender* pSender;
SSyncSnapshotReceiver* pReceiver;
bool restoreFinish;
// SSnapshot* pSnapshot;
SSyncSnapshotSender* senders[TSDB_MAX_REPLICA];
SSyncSnapshotReceiver* pNewNodeReceiver;
SSnapshotMeta sMeta;
} SSyncNode;
@ -164,6 +169,9 @@ void syncNodeStart(SSyncNode* pSyncNode);
void syncNodeStartStandBy(SSyncNode* pSyncNode);
void syncNodeClose(SSyncNode* pSyncNode);
// option
bool syncNodeSnapshotEnable(SSyncNode* pSyncNode);
// ping --------------
int32_t syncNodePing(SSyncNode* pSyncNode, const SRaftId* destRaftId, SyncPing* pMsg);
int32_t syncNodePingSelf(SSyncNode* pSyncNode);
@ -205,6 +213,25 @@ void syncNodeCandidate2Follower(SSyncNode* pSyncNode);
void syncNodeVoteForTerm(SSyncNode* pSyncNode, SyncTerm term, SRaftId* pRaftId);
void syncNodeVoteForSelf(SSyncNode* pSyncNode);
// snapshot --------------
bool syncNodeHasSnapshot(SSyncNode* pSyncNode);
bool syncNodeIsIndexInSnapshot(SSyncNode* pSyncNode, SyncIndex index);
SyncIndex syncNodeGetLastIndex(SSyncNode* pSyncNode);
SyncTerm syncNodeGetLastTerm(SSyncNode* pSyncNode);
int32_t syncNodeGetLastIndexTerm(SSyncNode* pSyncNode, SyncIndex* pLastIndex, SyncTerm* pLastTerm);
SyncIndex syncNodeSyncStartIndex(SSyncNode* pSyncNode);
SyncIndex syncNodeGetPreIndex(SSyncNode* pSyncNode, SyncIndex index);
SyncTerm syncNodeGetPreTerm(SSyncNode* pSyncNode, SyncIndex index);
int32_t syncNodeGetPreIndexTerm(SSyncNode* pSyncNode, SyncIndex index, SyncIndex* pPreIndex, SyncTerm* pPreTerm);
int32_t syncNodeCommit(SSyncNode* ths, SyncIndex beginIndex, SyncIndex endIndex, uint64_t flag);
bool syncNodeInRaftGroup(SSyncNode* ths, SRaftId* pRaftId);
SSyncSnapshotSender* syncNodeGetSnapshotSender(SSyncNode* ths, SRaftId* pDestId);
// for debug --------------
void syncNodePrint(SSyncNode* pObj);
void syncNodePrint2(char* s, SSyncNode* pObj);

8
source/libs/sync/inc/syncRaftCfg.h
View File

@ -34,6 +34,7 @@ typedef struct SRaftCfg {
TdFilePtr pFile;
char path[TSDB_FILENAME_LEN * 2];
int8_t isStandBy;
int8_t snapshotEnable;
} SRaftCfg;
SRaftCfg *raftCfgOpen(const char *path);
@ -50,7 +51,12 @@ char * raftCfg2Str(SRaftCfg *pRaftCfg);
int32_t raftCfgFromJson(const cJSON *pRoot, SRaftCfg *pRaftCfg);
int32_t raftCfgFromStr(const char *s, SRaftCfg *pRaftCfg);
int32_t raftCfgCreateFile(SSyncCfg *pCfg, int8_t isStandBy, const char *path);
typedef struct SRaftCfgMeta {
int8_t isStandBy;
int8_t snapshotEnable;
} SRaftCfgMeta;
int32_t raftCfgCreateFile(SSyncCfg *pCfg, SRaftCfgMeta meta, const char *path);
// for debug ----------------------
void syncCfgPrint(SSyncCfg *pCfg);

10
source/libs/sync/inc/syncRaftLog.h
View File

@ -30,6 +30,7 @@ extern "C" {
typedef struct SSyncLogStoreData {
SSyncNode* pSyncNode;
SWal* pWal;
SyncIndex beginIndex; // valid begin index, default 0, may be set beginIndex > 0
} SSyncLogStoreData;
SSyncLogStore* logStoreCreate(SSyncNode* pSyncNode);
@ -39,14 +40,7 @@ char* logStore2Str(SSyncLogStore* pLogStore);
cJSON* logStoreSimple2Json(SSyncLogStore* pLogStore);
char* logStoreSimple2Str(SSyncLogStore* pLogStore);
// SSyncRaftEntry* logStoreGetLastEntry(SSyncLogStore* pLogStore);
// SyncIndex logStoreLastIndex(SSyncLogStore* pLogStore);
// SyncTerm logStoreLastTerm(SSyncLogStore* pLogStore);
// SSyncRaftEntry* logStoreGetEntry(SSyncLogStore* pLogStore, SyncIndex index);
// int32_t logStoreAppendEntry(SSyncLogStore* pLogStore, SSyncRaftEntry* pEntry);
// int32_t logStoreTruncate(SSyncLogStore* pLogStore, SyncIndex fromIndex);
// int32_t logStoreUpdateCommitIndex(SSyncLogStore* pLogStore, SyncIndex index);
// SyncIndex logStoreGetCommitIndex(SSyncLogStore* pLogStore);
SyncIndex logStoreFirstIndex(SSyncLogStore* pLogStore);
// for debug
void logStorePrint(SSyncLogStore* pLogStore);

4
source/libs/sync/inc/syncRaftStore.h
View File

@ -49,8 +49,8 @@ void raftStoreClearVote(SRaftStore *pRaftStore);
void raftStoreNextTerm(SRaftStore *pRaftStore);
void raftStoreSetTerm(SRaftStore *pRaftStore, SyncTerm term);
int32_t raftStoreFromJson(SRaftStore *pRaftStore, cJSON *pJson);
cJSON * raftStore2Json(SRaftStore *pRaftStore);
char * raftStore2Str(SRaftStore *pRaftStore);
cJSON *raftStore2Json(SRaftStore *pRaftStore);
char *raftStore2Str(SRaftStore *pRaftStore);
// for debug -------------------
void raftStorePrint(SRaftStore *pObj);

1
source/libs/sync/inc/syncReplication.h
View File

@ -52,6 +52,7 @@ extern "C" {
// /\ UNCHANGED <<serverVars, candidateVars, leaderVars, logVars>>
//
int32_t syncNodeAppendEntriesPeers(SSyncNode* pSyncNode);
int32_t syncNodeAppendEntriesPeersSnapshot(SSyncNode* pSyncNode);
int32_t syncNodeReplicate(SSyncNode* pSyncNode);
int32_t syncNodeAppendEntries(SSyncNode* pSyncNode, const SRaftId* destRaftId, const SyncAppendEntries* pMsg);

1
source/libs/sync/inc/syncRequestVote.h
View File

@ -50,6 +50,7 @@ extern "C" {
// /\ UNCHANGED <<state, currentTerm, candidateVars, leaderVars, logVars>>
//
int32_t syncNodeOnRequestVoteCb(SSyncNode* ths, SyncRequestVote* pMsg);
int32_t syncNodeOnRequestVoteSnapshotCb(SSyncNode* ths, SyncRequestVote* pMsg);
#ifdef __cplusplus
}

1
source/libs/sync/inc/syncRequestVoteReply.h
View File

@ -45,6 +45,7 @@ extern "C" {
// /\ UNCHANGED <<serverVars, votedFor, leaderVars, logVars>>
//
int32_t syncNodeOnRequestVoteReplyCb(SSyncNode* ths, SyncRequestVoteReply* pMsg);
int32_t syncNodeOnRequestVoteReplySnapshotCb(SSyncNode* ths, SyncRequestVoteReply* pMsg);
#ifdef __cplusplus
}

60
source/libs/sync/inc/syncSnapshot.h
View File

@ -25,40 +25,64 @@ extern "C" {
#include <stdlib.h>
#include "cJSON.h"
#include "syncInt.h"
#include "syncMessage.h"
#include "taosdef.h"
#define SYNC_SNAPSHOT_SEQ_INVALID -1
#define SYNC_SNAPSHOT_SEQ_FORCE_CLOSE -2
#define SYNC_SNAPSHOT_SEQ_BEGIN 0
#define SYNC_SNAPSHOT_SEQ_END 0x7FFFFFFF
#define SYNC_SNAPSHOT_RETRY_MS 5000
typedef struct SSyncSnapshotSender {
int32_t sending;
int32_t received;
bool finish;
void * pCurrentBlock;
bool start;
int32_t seq;
int32_t ack;
void *pReader;
void *pCurrentBlock;
int32_t blockLen;
SSnapshot snapshot;
int64_t sendingMS;
SSnapshot *pSnapshot;
SSyncNode *pSyncNode;
int32_t replicaIndex;
SyncTerm term;
SyncTerm privateTerm;
bool finish;
} SSyncSnapshotSender;
SSyncSnapshotSender *snapshotSenderCreate(SSyncNode *pSyncNode);
SSyncSnapshotSender *snapshotSenderCreate(SSyncNode *pSyncNode, int32_t replicaIndex);
void snapshotSenderDestroy(SSyncSnapshotSender *pSender);
bool snapshotSenderIsStart(SSyncSnapshotSender *pSender);
void snapshotSenderStart(SSyncSnapshotSender *pSender);
void snapshotSenderStop(SSyncSnapshotSender *pSender);
int32_t snapshotSend(SSyncSnapshotSender *pSender);
cJSON * snapshotSender2Json(SSyncSnapshotSender *pSender);
char * snapshotSender2Str(SSyncSnapshotSender *pSender);
int32_t snapshotReSend(SSyncSnapshotSender *pSender);
cJSON *snapshotSender2Json(SSyncSnapshotSender *pSender);
char *snapshotSender2Str(SSyncSnapshotSender *pSender);
typedef struct SSyncSnapshotReceiver {
bool start;
int32_t received;
int32_t progressIndex;
void * pCurrentBlock;
int32_t len;
SSnapshot *pSnapshot;
bool start;
int32_t ack;
void *pWriter;
SyncTerm term;
SyncTerm privateTerm;
SSyncNode *pSyncNode;
int32_t replicaIndex;
} SSyncSnapshotReceiver;
SSyncSnapshotReceiver *snapshotReceiverCreate(SSyncNode *pSyncNode);
SSyncSnapshotReceiver *snapshotReceiverCreate(SSyncNode *pSyncNode, int32_t replicaIndex);
void snapshotReceiverDestroy(SSyncSnapshotReceiver *pReceiver);
int32_t snapshotReceive(SSyncSnapshotReceiver *pReceiver);
cJSON * snapshotReceiver2Json(SSyncSnapshotReceiver *pReceiver);
char * snapshotReceiver2Str(SSyncSnapshotReceiver *pReceiver);
void snapshotReceiverStart(SSyncSnapshotReceiver *pReceiver, SyncTerm privateTerm);
bool snapshotReceiverIsStart(SSyncSnapshotReceiver *pReceiver);
void snapshotReceiverStop(SSyncSnapshotReceiver *pReceiver, bool apply);
cJSON *snapshotReceiver2Json(SSyncSnapshotReceiver *pReceiver);
char *snapshotReceiver2Str(SSyncSnapshotReceiver *pReceiver);
int32_t syncNodeOnSnapshotSendCb(SSyncNode *ths, SyncSnapshotSend *pMsg);
int32_t syncNodeOnSnapshotRspCb(SSyncNode *ths, SyncSnapshotRsp *pMsg);
#ifdef __cplusplus
}

1
source/libs/sync/inc/syncUtil.h
View File

@ -61,6 +61,7 @@ bool syncUtilIsData(tmsg_t msgType);
bool syncUtilUserPreCommit(tmsg_t msgType);
bool syncUtilUserCommit(tmsg_t msgType);
bool syncUtilUserRollback(tmsg_t msgType);
void syncUtilJson2Line(char* jsonStr);
#ifdef __cplusplus
}

337
source/libs/sync/src/syncAppendEntries.c
View File

@ -18,8 +18,10 @@
#include "syncRaftCfg.h"
#include "syncRaftLog.h"
#include "syncRaftStore.h"
#include "syncSnapshot.h"
#include "syncUtil.h"
#include "syncVoteMgr.h"
#include "wal.h"
// TLA+ Spec
// HandleAppendEntriesRequest(i, j, m) ==
@ -335,8 +337,12 @@ int32_t syncNodeOnAppendEntriesCb(SSyncNode* ths, SyncAppendEntries* pMsg) {
cbMeta.currentTerm = ths->pRaftStore->currentTerm;
cbMeta.flag = 0x11;
SSnapshot snapshot;
ASSERT(ths->pFsm->FpGetSnapshot != NULL);
ths->pFsm->FpGetSnapshot(ths->pFsm, &snapshot);
bool needExecute = true;
if (ths->pSnapshot != NULL && cbMeta.index <= ths->pSnapshot->lastApplyIndex) {
if (cbMeta.index <= snapshot.lastApplyIndex) {
needExecute = false;
}
@ -427,3 +433,332 @@ int32_t syncNodeOnAppendEntriesCb(SSyncNode* ths, SyncAppendEntries* pMsg) {
return ret;
}
static int32_t syncNodeMakeLogSame(SSyncNode* ths, SyncAppendEntries* pMsg) {
int32_t code;
SyncIndex delBegin = pMsg->prevLogIndex + 1;
SyncIndex delEnd = ths->pLogStore->syncLogLastIndex(ths->pLogStore);
// invert roll back!
for (SyncIndex index = delEnd; index >= delBegin; --index) {
if (ths->pFsm->FpRollBackCb != NULL) {
SSyncRaftEntry* pRollBackEntry;
code = ths->pLogStore->syncLogGetEntry(ths->pLogStore, index, &pRollBackEntry);
ASSERT(code == 0);
ASSERT(pRollBackEntry != NULL);
if (syncUtilUserRollback(pRollBackEntry->msgType)) {
SRpcMsg rpcMsg;
syncEntry2OriginalRpc(pRollBackEntry, &rpcMsg);
SFsmCbMeta cbMeta;
cbMeta.index = pRollBackEntry->index;
cbMeta.isWeak = pRollBackEntry->isWeak;
cbMeta.code = 0;
cbMeta.state = ths->state;
cbMeta.seqNum = pRollBackEntry->seqNum;
ths->pFsm->FpRollBackCb(ths->pFsm, &rpcMsg, cbMeta);
rpcFreeCont(rpcMsg.pCont);
}
syncEntryDestory(pRollBackEntry);
}
}
// delete confict entries
code = ths->pLogStore->syncLogTruncate(ths->pLogStore, delBegin);
ASSERT(code == 0);
sInfo("sync event log truncate, from %ld to %ld", delBegin, delEnd);
logStoreSimpleLog2("after syncNodeMakeLogSame", ths->pLogStore);
return code;
}
static int32_t syncNodePreCommit(SSyncNode* ths, SSyncRaftEntry* pEntry) {
SRpcMsg rpcMsg;
syncEntry2OriginalRpc(pEntry, &rpcMsg);
if (ths->pFsm != NULL) {
if (ths->pFsm->FpPreCommitCb != NULL && syncUtilUserPreCommit(pEntry->originalRpcType)) {
SFsmCbMeta cbMeta;
cbMeta.index = pEntry->index;
cbMeta.isWeak = pEntry->isWeak;
cbMeta.code = 2;
cbMeta.state = ths->state;
cbMeta.seqNum = pEntry->seqNum;
ths->pFsm->FpPreCommitCb(ths->pFsm, &rpcMsg, cbMeta);
}
}
rpcFreeCont(rpcMsg.pCont);
return 0;
}
// really pre log match
// prevLogIndex == -1
static bool syncNodeOnAppendEntriesLogOK(SSyncNode* pSyncNode, SyncAppendEntries* pMsg) {
if (pMsg->prevLogIndex == SYNC_INDEX_INVALID) {
if (gRaftDetailLog) {
sTrace("syncNodeOnAppendEntriesLogOK true, pMsg->prevLogIndex:%ld", pMsg->prevLogIndex);
}
return true;
}
SyncIndex myLastIndex = syncNodeGetLastIndex(pSyncNode);
if (pMsg->prevLogIndex > myLastIndex) {
if (gRaftDetailLog) {
sTrace("syncNodeOnAppendEntriesLogOK false, pMsg->prevLogIndex:%ld, myLastIndex:%ld", pMsg->prevLogIndex,
myLastIndex);
}
return false;
}
SyncTerm myPreLogTerm = syncNodeGetPreTerm(pSyncNode, pMsg->prevLogIndex + 1);
if (pMsg->prevLogIndex <= myLastIndex && pMsg->prevLogTerm == myPreLogTerm) {
if (gRaftDetailLog) {
sTrace(
"syncNodeOnAppendEntriesLogOK true, pMsg->prevLogIndex:%ld, myLastIndex:%ld, pMsg->prevLogTerm:%lu, "
"myPreLogTerm:%lu",
pMsg->prevLogIndex, myLastIndex, pMsg->prevLogTerm, myPreLogTerm);
}
return true;
}
if (gRaftDetailLog) {
sTrace(
"syncNodeOnAppendEntriesLogOK false, pMsg->prevLogIndex:%ld, myLastIndex:%ld, pMsg->prevLogTerm:%lu, "
"myPreLogTerm:%lu",
pMsg->prevLogIndex, myLastIndex, pMsg->prevLogTerm, myPreLogTerm);
}
return false;
}
int32_t syncNodeOnAppendEntriesSnapshotCb(SSyncNode* ths, SyncAppendEntries* pMsg) {
int32_t ret = 0;
int32_t code = 0;
// print log
char logBuf[128] = {0};
snprintf(logBuf, sizeof(logBuf), "recv SyncAppendEntries, vgId:%d, term:%lu", ths->vgId,
ths->pRaftStore->currentTerm);
syncAppendEntriesLog2(logBuf, pMsg);
// if already drop replica, do not process
if (!syncNodeInRaftGroup(ths, &(pMsg->srcId)) && !ths->pRaftCfg->isStandBy) {
sInfo("recv SyncAppendEntries maybe replica already dropped");
return ret;
}
// maybe update term
if (pMsg->term > ths->pRaftStore->currentTerm) {
syncNodeUpdateTerm(ths, pMsg->term);
}
ASSERT(pMsg->term <= ths->pRaftStore->currentTerm);
// reset elect timer
if (pMsg->term == ths->pRaftStore->currentTerm) {
ths->leaderCache = pMsg->srcId;
syncNodeResetElectTimer(ths);
}
ASSERT(pMsg->dataLen >= 0);
// candidate to follower
//
// operation:
// to follower
do {
bool condition = pMsg->term == ths->pRaftStore->currentTerm && ths->state == TAOS_SYNC_STATE_CANDIDATE;
if (condition) {
sTrace("recv SyncAppendEntries, candidate to follower");
syncNodeBecomeFollower(ths);
// do not reply?
return ret;
}
} while (0);
// fake match
//
// condition1:
// I have snapshot, no log, preIndex > myLastIndex
//
// condition2:
// I have snapshot, have log, log <= snapshot, preIndex > myLastIndex
//
// condition3:
// I have snapshot, preIndex < snapshot.lastApplyIndex
//
// condition4:
// I have snapshot, preIndex == snapshot.lastApplyIndex, no data
//
// operation:
// match snapshot.lastApplyIndex - 1;
// no operation on log
do {
SyncIndex myLastIndex = syncNodeGetLastIndex(ths);
SSnapshot snapshot;
ths->pFsm->FpGetSnapshot(ths->pFsm, &snapshot);
bool condition0 = (pMsg->term == ths->pRaftStore->currentTerm) && (ths->state == TAOS_SYNC_STATE_FOLLOWER) &&
syncNodeHasSnapshot(ths);
bool condition1 =
condition0 && (ths->pLogStore->syncLogEntryCount(ths->pLogStore) == 0) && (pMsg->prevLogIndex > myLastIndex);
bool condition2 = condition0 && (ths->pLogStore->syncLogLastIndex(ths->pLogStore) <= snapshot.lastApplyIndex) &&
(pMsg->prevLogIndex > myLastIndex);
bool condition3 = condition0 && (pMsg->prevLogIndex < snapshot.lastApplyIndex);
bool condition4 = condition0 && (pMsg->prevLogIndex == snapshot.lastApplyIndex) && (pMsg->dataLen == 0);
bool condition = condition1 || condition2 || condition3 || condition4;
if (condition) {
sTrace(
"recv SyncAppendEntries, fake match, myLastIndex:%ld, syncLogBeginIndex:%ld, syncLogEndIndex:%ld, "
"condition1:%d, condition2:%d, condition3:%d, condition4:%d",
myLastIndex, ths->pLogStore->syncLogBeginIndex(ths->pLogStore),
ths->pLogStore->syncLogEndIndex(ths->pLogStore), condition1, condition2, condition3, condition4);
// prepare response msg
SyncAppendEntriesReply* pReply = syncAppendEntriesReplyBuild(ths->vgId);
pReply->srcId = ths->myRaftId;
pReply->destId = pMsg->srcId;
pReply->term = ths->pRaftStore->currentTerm;
pReply->privateTerm = ths->pNewNodeReceiver->privateTerm;
pReply->success = true;
pReply->matchIndex = snapshot.lastApplyIndex;
// send response
SRpcMsg rpcMsg;
syncAppendEntriesReply2RpcMsg(pReply, &rpcMsg);
syncNodeSendMsgById(&pReply->destId, ths, &rpcMsg);
syncAppendEntriesReplyDestroy(pReply);
return ret;
}
} while (0);
// calculate logOK here, before will coredump, due to fake match
bool logOK = syncNodeOnAppendEntriesLogOK(ths, pMsg);
// not match
//
// condition1:
// term < myTerm
//
// condition2:
// !logOK
//
// operation:
// not match
// no operation on log
do {
bool condition1 = pMsg->term < ths->pRaftStore->currentTerm;
bool condition2 =
(pMsg->term == ths->pRaftStore->currentTerm) && (ths->state == TAOS_SYNC_STATE_FOLLOWER) && !logOK;
bool condition = condition1 || condition2;
if (condition) {
sTrace(
"recv SyncAppendEntries, not match, syncLogBeginIndex:%ld, syncLogEndIndex:%ld, condition1:%d, "
"condition2:%d, logOK:%d",
ths->pLogStore->syncLogBeginIndex(ths->pLogStore), ths->pLogStore->syncLogEndIndex(ths->pLogStore),
condition1, condition2, logOK);
// prepare response msg
SyncAppendEntriesReply* pReply = syncAppendEntriesReplyBuild(ths->vgId);
pReply->srcId = ths->myRaftId;
pReply->destId = pMsg->srcId;
pReply->term = ths->pRaftStore->currentTerm;
pReply->privateTerm = ths->pNewNodeReceiver->privateTerm;
pReply->success = false;
pReply->matchIndex = SYNC_INDEX_INVALID;
// send response
SRpcMsg rpcMsg;
syncAppendEntriesReply2RpcMsg(pReply, &rpcMsg);
syncNodeSendMsgById(&pReply->destId, ths, &rpcMsg);
syncAppendEntriesReplyDestroy(pReply);
return ret;
}
} while (0);
// really match
//
// condition:
// logOK
//
// operation:
// match
// make log same
do {
bool condition = (pMsg->term == ths->pRaftStore->currentTerm) && (ths->state == TAOS_SYNC_STATE_FOLLOWER) && logOK;
if (condition) {
// has extra entries (> preIndex) in local log
SyncIndex myLastIndex = syncNodeGetLastIndex(ths);
bool hasExtraEntries = myLastIndex > pMsg->prevLogIndex;
// has entries in SyncAppendEntries msg
bool hasAppendEntries = pMsg->dataLen > 0;
sTrace("recv SyncAppendEntries, match, myLastIndex:%ld, hasExtraEntries:%d, hasAppendEntries:%d", myLastIndex,
hasExtraEntries, hasAppendEntries);
if (hasExtraEntries) {
// make log same, rollback deleted entries
code = syncNodeMakeLogSame(ths, pMsg);
ASSERT(code == 0);
}
if (hasAppendEntries) {
// append entry
SSyncRaftEntry* pAppendEntry = syncEntryDeserialize(pMsg->data, pMsg->dataLen);
ASSERT(pAppendEntry != NULL);
code = ths->pLogStore->syncLogAppendEntry(ths->pLogStore, pAppendEntry);
ASSERT(code == 0);
// pre commit
code = syncNodePreCommit(ths, pAppendEntry);
ASSERT(code == 0);
syncEntryDestory(pAppendEntry);
}
// prepare response msg
SyncAppendEntriesReply* pReply = syncAppendEntriesReplyBuild(ths->vgId);
pReply->srcId = ths->myRaftId;
pReply->destId = pMsg->srcId;
pReply->term = ths->pRaftStore->currentTerm;
pReply->privateTerm = ths->pNewNodeReceiver->privateTerm;
pReply->success = true;
pReply->matchIndex = hasAppendEntries ? pMsg->prevLogIndex + 1 : pMsg->prevLogIndex;
// send response
SRpcMsg rpcMsg;
syncAppendEntriesReply2RpcMsg(pReply, &rpcMsg);
syncNodeSendMsgById(&pReply->destId, ths, &rpcMsg);
syncAppendEntriesReplyDestroy(pReply);
// maybe update commit index, leader notice me
if (pMsg->commitIndex > ths->commitIndex) {
// has commit entry in local
if (pMsg->commitIndex <= ths->pLogStore->syncLogLastIndex(ths->pLogStore)) {
SyncIndex beginIndex = ths->commitIndex + 1;
SyncIndex endIndex = pMsg->commitIndex;
// update commit index
ths->commitIndex = pMsg->commitIndex;
// call back Wal
code = ths->pLogStore->updateCommitIndex(ths->pLogStore, ths->commitIndex);
ASSERT(code == 0);
code = syncNodeCommit(ths, beginIndex, endIndex, ths->state);
ASSERT(code == 0);
}
}
return ret;
}
} while (0);
return ret;
}

115
source/libs/sync/src/syncAppendEntriesReply.c
View File

@ -17,8 +17,10 @@
#include "syncCommit.h"
#include "syncIndexMgr.h"
#include "syncInt.h"
#include "syncRaftCfg.h"
#include "syncRaftLog.h"
#include "syncRaftStore.h"
#include "syncSnapshot.h"
#include "syncUtil.h"
#include "syncVoteMgr.h"
@ -94,3 +96,116 @@ int32_t syncNodeOnAppendEntriesReplyCb(SSyncNode* ths, SyncAppendEntriesReply* p
return ret;
}
int32_t syncNodeOnAppendEntriesReplySnapshotCb(SSyncNode* ths, SyncAppendEntriesReply* pMsg) {
int32_t ret = 0;
// print log
char logBuf[128] = {0};
snprintf(logBuf, sizeof(logBuf), "recv SyncAppendEntriesReply, vgId:%d, term:%lu", ths->vgId,
ths->pRaftStore->currentTerm);
syncAppendEntriesReplyLog2(logBuf, pMsg);
// if already drop replica, do not process
if (!syncNodeInRaftGroup(ths, &(pMsg->srcId)) && !ths->pRaftCfg->isStandBy) {
sInfo("recv SyncAppendEntriesReply, maybe replica already dropped");
return ret;
}
// drop stale response
if (pMsg->term < ths->pRaftStore->currentTerm) {
sTrace("recv SyncAppendEntriesReply, drop stale response, receive_term:%lu current_term:%lu", pMsg->term,
ths->pRaftStore->currentTerm);
return ret;
}
syncIndexMgrLog2("recv SyncAppendEntriesReply, before pNextIndex:", ths->pNextIndex);
syncIndexMgrLog2("recv SyncAppendEntriesReply, before pMatchIndex:", ths->pMatchIndex);
{
SSnapshot snapshot;
ths->pFsm->FpGetSnapshot(ths->pFsm, &snapshot);
sTrace("recv SyncAppendEntriesReply, before snapshot.lastApplyIndex:%ld, snapshot.lastApplyTerm:%lu",
snapshot.lastApplyIndex, snapshot.lastApplyTerm);
}
// no need this code, because if I receive reply.term, then I must have sent for that term.
// if (pMsg->term > ths->pRaftStore->currentTerm) {
// syncNodeUpdateTerm(ths, pMsg->term);
// }
if (pMsg->term > ths->pRaftStore->currentTerm) {
char logBuf[128] = {0};
snprintf(logBuf, sizeof(logBuf), "recv SyncAppendEntriesReply, error term, receive_term:%lu current_term:%lu",
pMsg->term, ths->pRaftStore->currentTerm);
syncNodeLog2(logBuf, ths);
sError("%s", logBuf);
return ret;
}
ASSERT(pMsg->term == ths->pRaftStore->currentTerm);
if (pMsg->success) {
// nextIndex' = [nextIndex EXCEPT ![i][j] = m.mmatchIndex + 1]
syncIndexMgrSetIndex(ths->pNextIndex, &(pMsg->srcId), pMsg->matchIndex + 1);
sTrace("update next match, index:%ld, success:%d", pMsg->matchIndex + 1, pMsg->success);
// matchIndex' = [matchIndex EXCEPT ![i][j] = m.mmatchIndex]
syncIndexMgrSetIndex(ths->pMatchIndex, &(pMsg->srcId), pMsg->matchIndex);
// maybe commit
if (ths->state == TAOS_SYNC_STATE_LEADER) {
syncMaybeAdvanceCommitIndex(ths);
}
} else {
SyncIndex nextIndex = syncIndexMgrGetIndex(ths->pNextIndex, &(pMsg->srcId));
sTrace("update next not match, begin, index:%ld, success:%d", nextIndex, pMsg->success);
// notice! int64, uint64
if (nextIndex > SYNC_INDEX_BEGIN) {
--nextIndex;
// get sender
SSyncSnapshotSender* pSender = syncNodeGetSnapshotSender(ths, &(pMsg->srcId));
ASSERT(pSender != NULL);
bool hasSnapshot = syncNodeHasSnapshot(ths);
SSnapshot snapshot;
ths->pFsm->FpGetSnapshot(ths->pFsm, &snapshot);
// start sending snapshot first time
// start here, stop by receiver
if (hasSnapshot && nextIndex <= snapshot.lastApplyIndex + 1 && !snapshotSenderIsStart(pSender) &&
pMsg->privateTerm < pSender->privateTerm) {
snapshotSenderStart(pSender);
char* s = snapshotSender2Str(pSender);
sInfo("sync event snapshot send start sender first time, sender:%s", s);
taosMemoryFree(s);
}
SyncIndex sentryIndex = pSender->snapshot.lastApplyIndex + 1;
// update nextIndex to sentryIndex
if (nextIndex <= sentryIndex) {
nextIndex = sentryIndex;
}
} else {
nextIndex = SYNC_INDEX_BEGIN;
}
syncIndexMgrSetIndex(ths->pNextIndex, &(pMsg->srcId), nextIndex);
sTrace("update next not match, end, index:%ld, success:%d", nextIndex, pMsg->success);
}
syncIndexMgrLog2("recv SyncAppendEntriesReply, after pNextIndex:", ths->pNextIndex);
syncIndexMgrLog2("recv SyncAppendEntriesReply, after pMatchIndex:", ths->pMatchIndex);
{
SSnapshot snapshot;
ths->pFsm->FpGetSnapshot(ths->pFsm, &snapshot);
sTrace("recv SyncAppendEntriesReply, after snapshot.lastApplyIndex:%ld, snapshot.lastApplyTerm:%lu",
snapshot.lastApplyIndex, snapshot.lastApplyTerm);
}
return ret;
}

105
source/libs/sync/src/syncCommit.c
View File

@ -94,109 +94,8 @@ void syncMaybeAdvanceCommitIndex(SSyncNode* pSyncNode) {
// execute fsm
if (pSyncNode->pFsm != NULL) {
for (SyncIndex i = beginIndex; i <= endIndex; ++i) {
if (i != SYNC_INDEX_INVALID) {
SSyncRaftEntry* pEntry = pSyncNode->pLogStore->getEntry(pSyncNode->pLogStore, i);
assert(pEntry != NULL);
SRpcMsg rpcMsg;
syncEntry2OriginalRpc(pEntry, &rpcMsg);
if (pSyncNode->pFsm->FpCommitCb != NULL && syncUtilUserCommit(pEntry->originalRpcType)) {
SFsmCbMeta cbMeta;
cbMeta.index = pEntry->index;
cbMeta.isWeak = pEntry->isWeak;
cbMeta.code = 0;
cbMeta.state = pSyncNode->state;
cbMeta.seqNum = pEntry->seqNum;
cbMeta.term = pEntry->term;
cbMeta.currentTerm = pSyncNode->pRaftStore->currentTerm;
cbMeta.flag = 0x1;
bool needExecute = true;
if (pSyncNode->pSnapshot != NULL && cbMeta.index <= pSyncNode->pSnapshot->lastApplyIndex) {
needExecute = false;
}
if (needExecute) {
pSyncNode->pFsm->FpCommitCb(pSyncNode->pFsm, &rpcMsg, cbMeta);
}
}
// config change
if (pEntry->originalRpcType == TDMT_SYNC_CONFIG_CHANGE) {
SSyncCfg oldSyncCfg = pSyncNode->pRaftCfg->cfg;
SSyncCfg newSyncCfg;
int32_t ret = syncCfgFromStr(rpcMsg.pCont, &newSyncCfg);
ASSERT(ret == 0);
// update new config myIndex
bool hit = false;
for (int i = 0; i < newSyncCfg.replicaNum; ++i) {
if (strcmp(pSyncNode->myNodeInfo.nodeFqdn, (newSyncCfg.nodeInfo)[i].nodeFqdn) == 0 &&
pSyncNode->myNodeInfo.nodePort == (newSyncCfg.nodeInfo)[i].nodePort) {
newSyncCfg.myIndex = i;
hit = true;
break;
}
}
if (pSyncNode->state == TAOS_SYNC_STATE_LEADER) {
ASSERT(hit == true);
}
bool isDrop;
syncNodeUpdateConfig(pSyncNode, &newSyncCfg, &isDrop);
// change isStandBy to normal
if (!isDrop) {
if (pSyncNode->state == TAOS_SYNC_STATE_LEADER) {
syncNodeBecomeLeader(pSyncNode);
} else {
syncNodeBecomeFollower(pSyncNode);
}
}
char* sOld = syncCfg2Str(&oldSyncCfg);
char* sNew = syncCfg2Str(&newSyncCfg);
sInfo("==config change== 0x1 old:%s new:%s isDrop:%d \n", sOld, sNew, isDrop);
taosMemoryFree(sOld);
taosMemoryFree(sNew);
if (pSyncNode->pFsm->FpReConfigCb != NULL) {
SReConfigCbMeta cbMeta = {0};
cbMeta.code = 0;
cbMeta.currentTerm = pSyncNode->pRaftStore->currentTerm;
cbMeta.index = pEntry->index;
cbMeta.term = pEntry->term;
cbMeta.oldCfg = oldSyncCfg;
cbMeta.flag = 0x1;
cbMeta.isDrop = isDrop;
pSyncNode->pFsm->FpReConfigCb(pSyncNode->pFsm, newSyncCfg, cbMeta);
}
}
// restore finish
if (pEntry->index == pSyncNode->pLogStore->getLastIndex(pSyncNode->pLogStore)) {
if (pSyncNode->restoreFinish == false) {
if (pSyncNode->pFsm->FpRestoreFinishCb != NULL) {
pSyncNode->pFsm->FpRestoreFinishCb(pSyncNode->pFsm);
}
pSyncNode->restoreFinish = true;
sInfo("==syncMaybeAdvanceCommitIndex== restoreFinish set true %p vgId:%d", pSyncNode, pSyncNode->vgId);
/*
tsem_post(&pSyncNode->restoreSem);
sInfo("==syncMaybeAdvanceCommitIndex== RestoreFinish tsem_post %p", pSyncNode);
*/
}
}
rpcFreeCont(rpcMsg.pCont);
syncEntryDestory(pEntry);
}
}
int32_t code = syncNodeCommit(pSyncNode, beginIndex, endIndex, pSyncNode->state);
ASSERT(code == 0);
}
}
}

28
source/libs/sync/src/syncElection.c
View File

@ -15,6 +15,7 @@
#include "syncElection.h"
#include "syncMessage.h"
#include "syncRaftCfg.h"
#include "syncRaftStore.h"
#include "syncVoteMgr.h"
@ -49,6 +50,26 @@ int32_t syncNodeRequestVotePeers(SSyncNode* pSyncNode) {
return ret;
}
int32_t syncNodeRequestVotePeersSnapshot(SSyncNode* pSyncNode) {
ASSERT(pSyncNode->state == TAOS_SYNC_STATE_CANDIDATE);
int32_t ret = 0;
for (int i = 0; i < pSyncNode->peersNum; ++i) {
SyncRequestVote* pMsg = syncRequestVoteBuild(pSyncNode->vgId);
pMsg->srcId = pSyncNode->myRaftId;
pMsg->destId = pSyncNode->peersId[i];
pMsg->term = pSyncNode->pRaftStore->currentTerm;
ret = syncNodeGetLastIndexTerm(pSyncNode, &(pMsg->lastLogIndex), &(pMsg->lastLogTerm));
ASSERT(ret == 0);
ret = syncNodeRequestVote(pSyncNode, &pSyncNode->peersId[i], pMsg);
ASSERT(ret == 0);
syncRequestVoteDestroy(pMsg);
}
return ret;
}
int32_t syncNodeElect(SSyncNode* pSyncNode) {
int32_t ret = 0;
if (pSyncNode->state == TAOS_SYNC_STATE_FOLLOWER) {
@ -71,7 +92,12 @@ int32_t syncNodeElect(SSyncNode* pSyncNode) {
return ret;
}
ret = syncNodeRequestVotePeers(pSyncNode);
if (pSyncNode->pRaftCfg->snapshotEnable) {
ret = syncNodeRequestVotePeersSnapshot(pSyncNode);
} else {
ret = syncNodeRequestVotePeers(pSyncNode);
}
assert(ret == 0);
syncNodeResetElectTimer(pSyncNode);

30
source/libs/sync/src/syncIO.c
View File

@ -75,7 +75,8 @@ int32_t syncIOSendMsg(const SEpSet *pEpSet, SRpcMsg *pMsg) {
syncUtilMsgNtoH(pMsg->pCont);
char logBuf[256] = {0};
snprintf(logBuf, sizeof(logBuf), "==syncIOSendMsg== %s:%d", pEpSet->eps[0].fqdn, pEpSet->eps[0].port);
snprintf(logBuf, sizeof(logBuf), "==syncIOSendMsg== %s:%d msgType:%d", pEpSet->eps[0].fqdn, pEpSet->eps[0].port,
pMsg->msgType);
syncRpcMsgLog2(logBuf, pMsg);
syncUtilMsgHtoN(pMsg->pCont);
@ -89,8 +90,10 @@ int32_t syncIOSendMsg(const SEpSet *pEpSet, SRpcMsg *pMsg) {
int32_t syncIOEqMsg(const SMsgCb *msgcb, SRpcMsg *pMsg) {
int32_t ret = 0;
char logBuf[128] = {0};
syncRpcMsgLog2((char *)"==syncIOEqMsg==", pMsg);
char logBuf[256] = {0};
snprintf(logBuf, sizeof(logBuf), "==syncIOEqMsg== msgType:%d", pMsg->msgType);
syncRpcMsgLog2(logBuf, pMsg);
SRpcMsg *pTemp;
pTemp = taosAllocateQitem(sizeof(SRpcMsg), DEF_QITEM);
@ -253,7 +256,9 @@ static void *syncIOConsumerFunc(void *param) {
for (int i = 0; i < numOfMsgs; ++i) {
taosGetQitem(qall, (void **)&pRpcMsg);
syncRpcMsgLog2((char *)"==syncIOConsumerFunc==", pRpcMsg);
char logBuf[128];
snprintf(logBuf, sizeof(logBuf), "==syncIOConsumMsg== msgType:%d", pRpcMsg->msgType);
syncRpcMsgLog2(logBuf, pRpcMsg);
// use switch case instead of if else
if (pRpcMsg->msgType == TDMT_SYNC_PING) {
@ -319,6 +324,23 @@ static void *syncIOConsumerFunc(void *param) {
io->FpOnSyncTimeout(io->pSyncNode, pSyncMsg);
syncTimeoutDestroy(pSyncMsg);
}
} else if (pRpcMsg->msgType == TDMT_SYNC_SNAPSHOT_SEND) {
if (io->FpOnSyncSnapshotSend != NULL) {
SyncSnapshotSend *pSyncMsg = syncSnapshotSendFromRpcMsg2(pRpcMsg);
assert(pSyncMsg != NULL);
io->FpOnSyncSnapshotSend(io->pSyncNode, pSyncMsg);
syncSnapshotSendDestroy(pSyncMsg);
}
} else if (pRpcMsg->msgType == TDMT_SYNC_SNAPSHOT_RSP) {
if (io->FpOnSyncSnapshotRsp != NULL) {
SyncSnapshotRsp *pSyncMsg = syncSnapshotRspFromRpcMsg2(pRpcMsg);
assert(pSyncMsg != NULL);
io->FpOnSyncSnapshotRsp(io->pSyncNode, pSyncMsg);
syncSnapshotRspDestroy(pSyncMsg);
}
} else {
sTrace("unknown msgType:%d, no operator", pRpcMsg->msgType);
}

62
source/libs/sync/src/syncIndexMgr.c
View File

@ -46,6 +46,7 @@ void syncIndexMgrDestroy(SSyncIndexMgr *pSyncIndexMgr) {
void syncIndexMgrClear(SSyncIndexMgr *pSyncIndexMgr) {
memset(pSyncIndexMgr->index, 0, sizeof(pSyncIndexMgr->index));
memset(pSyncIndexMgr->privateTerm, 0, sizeof(pSyncIndexMgr->privateTerm));
/*
for (int i = 0; i < pSyncIndexMgr->replicaNum; ++i) {
pSyncIndexMgr->index[i] = 0;
@ -62,7 +63,7 @@ void syncIndexMgrSetIndex(SSyncIndexMgr *pSyncIndexMgr, const SRaftId *pRaftId,
}
// maybe config change
// assert(0);
assert(0);
}
SyncIndex syncIndexMgrGetIndex(SSyncIndexMgr *pSyncIndexMgr, const SRaftId *pRaftId) {
@ -86,14 +87,27 @@ cJSON *syncIndexMgr2Json(SSyncIndexMgr *pSyncIndexMgr) {
for (int i = 0; i < pSyncIndexMgr->replicaNum; ++i) {
cJSON_AddItemToArray(pReplicas, syncUtilRaftId2Json(&(*(pSyncIndexMgr->replicas))[i]));
}
int respondNum = 0;
int *arr = (int *)taosMemoryMalloc(sizeof(int) * pSyncIndexMgr->replicaNum);
for (int i = 0; i < pSyncIndexMgr->replicaNum; ++i) {
arr[i] = pSyncIndexMgr->index[i];
{
int *arr = (int *)taosMemoryMalloc(sizeof(int) * pSyncIndexMgr->replicaNum);
for (int i = 0; i < pSyncIndexMgr->replicaNum; ++i) {
arr[i] = pSyncIndexMgr->index[i];
}
cJSON *pIndex = cJSON_CreateIntArray(arr, pSyncIndexMgr->replicaNum);
taosMemoryFree(arr);
cJSON_AddItemToObject(pRoot, "index", pIndex);
}
cJSON *pIndex = cJSON_CreateIntArray(arr, pSyncIndexMgr->replicaNum);
taosMemoryFree(arr);
cJSON_AddItemToObject(pRoot, "index", pIndex);
{
int *arr = (int *)taosMemoryMalloc(sizeof(int) * pSyncIndexMgr->replicaNum);
for (int i = 0; i < pSyncIndexMgr->replicaNum; ++i) {
arr[i] = pSyncIndexMgr->privateTerm[i];
}
cJSON *pIndex = cJSON_CreateIntArray(arr, pSyncIndexMgr->replicaNum);
taosMemoryFree(arr);
cJSON_AddItemToObject(pRoot, "privateTerm", pIndex);
}
snprintf(u64buf, sizeof(u64buf), "%p", pSyncIndexMgr->pSyncNode);
cJSON_AddStringToObject(pRoot, "pSyncNode", u64buf);
}
@ -105,7 +119,7 @@ cJSON *syncIndexMgr2Json(SSyncIndexMgr *pSyncIndexMgr) {
char *syncIndexMgr2Str(SSyncIndexMgr *pSyncIndexMgr) {
cJSON *pJson = syncIndexMgr2Json(pSyncIndexMgr);
char * serialized = cJSON_Print(pJson);
char *serialized = cJSON_Print(pJson);
cJSON_Delete(pJson);
return serialized;
}
@ -132,7 +146,31 @@ void syncIndexMgrLog(SSyncIndexMgr *pObj) {
}
void syncIndexMgrLog2(char *s, SSyncIndexMgr *pObj) {
char *serialized = syncIndexMgr2Str(pObj);
sTrace("syncIndexMgrLog2 | len:%lu | %s | %s", strlen(serialized), s, serialized);
taosMemoryFree(serialized);
if (gRaftDetailLog) {
char *serialized = syncIndexMgr2Str(pObj);
sTrace("syncIndexMgrLog2 | len:%lu | %s | %s", strlen(serialized), s, serialized);
taosMemoryFree(serialized);
}
}
void syncIndexMgrSetTerm(SSyncIndexMgr *pSyncIndexMgr, const SRaftId *pRaftId, SyncTerm term) {
for (int i = 0; i < pSyncIndexMgr->replicaNum; ++i) {
if (syncUtilSameId(&((*(pSyncIndexMgr->replicas))[i]), pRaftId)) {
(pSyncIndexMgr->privateTerm)[i] = term;
return;
}
}
// maybe config change
assert(0);
}
SyncTerm syncIndexMgrGetTerm(SSyncIndexMgr *pSyncIndexMgr, const SRaftId *pRaftId) {
for (int i = 0; i < pSyncIndexMgr->replicaNum; ++i) {
if (syncUtilSameId(&((*(pSyncIndexMgr->replicas))[i]), pRaftId)) {
SyncTerm term = (pSyncIndexMgr->privateTerm)[i];
return term;
}
}
assert(0);
}

431
source/libs/sync/src/syncMain.c
View File

@ -29,11 +29,14 @@
#include "syncRequestVote.h"
#include "syncRequestVoteReply.h"
#include "syncRespMgr.h"
#include "syncSnapshot.h"
#include "syncTimeout.h"
#include "syncUtil.h"
#include "syncVoteMgr.h"
#include "tref.h"
bool gRaftDetailLog = false;
static int32_t tsNodeRefId = -1;
// ------ local funciton ---------
@ -213,6 +216,18 @@ bool syncIsRestoreFinish(int64_t rid) {
return b;
}
int32_t syncGetSnapshotMeta(int64_t rid, struct SSnapshotMeta* sMeta) {
SSyncNode* pSyncNode = (SSyncNode*)taosAcquireRef(tsNodeRefId, rid);
if (pSyncNode == NULL) {
return -1;
}
assert(rid == pSyncNode->rid);
*sMeta = pSyncNode->sMeta;
taosReleaseRef(tsNodeRefId, pSyncNode->rid);
return 0;
}
const char* syncGetMyRoleStr(int64_t rid) {
const char* s = syncUtilState2String(syncGetMyRole(rid));
return s;
@ -411,8 +426,11 @@ SSyncNode* syncNodeOpen(const SSyncInfo* pOldSyncInfo) {
snprintf(pSyncNode->configPath, sizeof(pSyncNode->configPath), "%s/raft_config.json", pSyncInfo->path);
if (!taosCheckExistFile(pSyncNode->configPath)) {
// create raft config file
ret = raftCfgCreateFile((SSyncCfg*)&(pSyncInfo->syncCfg), pSyncInfo->isStandBy, pSyncNode->configPath);
// create a new raft config file
SRaftCfgMeta meta;
meta.isStandBy = pSyncInfo->isStandBy;
meta.snapshotEnable = pSyncInfo->snapshotEnable;
ret = raftCfgCreateFile((SSyncCfg*)&(pSyncInfo->syncCfg), meta, pSyncNode->configPath);
assert(ret == 0);
} else {
@ -552,35 +570,64 @@ SSyncNode* syncNodeOpen(const SSyncInfo* pOldSyncInfo) {
pSyncNode->FpOnPing = syncNodeOnPingCb;
pSyncNode->FpOnPingReply = syncNodeOnPingReplyCb;
pSyncNode->FpOnClientRequest = syncNodeOnClientRequestCb;
pSyncNode->FpOnRequestVote = syncNodeOnRequestVoteCb;
pSyncNode->FpOnRequestVoteReply = syncNodeOnRequestVoteReplyCb;
pSyncNode->FpOnAppendEntries = syncNodeOnAppendEntriesCb;
pSyncNode->FpOnAppendEntriesReply = syncNodeOnAppendEntriesReplyCb;
pSyncNode->FpOnTimeout = syncNodeOnTimeoutCb;
pSyncNode->FpOnSnapshotSend = syncNodeOnSnapshotSendCb;
pSyncNode->FpOnSnapshotRsp = syncNodeOnSnapshotRspCb;
if (pSyncNode->pRaftCfg->snapshotEnable) {
sInfo("sync node use snapshot");
pSyncNode->FpOnRequestVote = syncNodeOnRequestVoteSnapshotCb;
pSyncNode->FpOnRequestVoteReply = syncNodeOnRequestVoteReplySnapshotCb;
pSyncNode->FpOnAppendEntries = syncNodeOnAppendEntriesSnapshotCb;
pSyncNode->FpOnAppendEntriesReply = syncNodeOnAppendEntriesReplySnapshotCb;
} else {
sInfo("sync node do not use snapshot");
pSyncNode->FpOnRequestVote = syncNodeOnRequestVoteCb;
pSyncNode->FpOnRequestVoteReply = syncNodeOnRequestVoteReplyCb;
pSyncNode->FpOnAppendEntries = syncNodeOnAppendEntriesCb;
pSyncNode->FpOnAppendEntriesReply = syncNodeOnAppendEntriesReplyCb;
}
// tools
pSyncNode->pSyncRespMgr = syncRespMgrCreate(NULL, 0);
assert(pSyncNode->pSyncRespMgr != NULL);
// restore state
pSyncNode->restoreFinish = false;
pSyncNode->pSnapshot = NULL;
if (pSyncNode->pFsm->FpGetSnapshot != NULL) {
pSyncNode->pSnapshot = taosMemoryMalloc(sizeof(SSnapshot));
pSyncNode->pFsm->FpGetSnapshot(pSyncNode->pFsm, pSyncNode->pSnapshot);
}
// pSyncNode->pSnapshot = NULL;
// if (pSyncNode->pFsm->FpGetSnapshot != NULL) {
// pSyncNode->pSnapshot = taosMemoryMalloc(sizeof(SSnapshot));
// pSyncNode->pFsm->FpGetSnapshot(pSyncNode->pFsm, pSyncNode->pSnapshot);
// }
// tsem_init(&(pSyncNode->restoreSem), 0, 0);
// snapshot senders
for (int i = 0; i < TSDB_MAX_REPLICA; ++i) {
SSyncSnapshotSender* pSender = snapshotSenderCreate(pSyncNode, i);
// ASSERT(pSender != NULL);
(pSyncNode->senders)[i] = pSender;
}
// snapshot receivers
pSyncNode->pNewNodeReceiver = snapshotReceiverCreate(pSyncNode, 100);
// start in syncNodeStart
// start raft
// syncNodeBecomeFollower(pSyncNode);
// snapshot meta
pSyncNode->sMeta.lastConfigIndex = -1;
return pSyncNode;
}
void syncNodeStart(SSyncNode* pSyncNode) {
// start raft
if (pSyncNode->replicaNum == 1) {
raftStoreNextTerm(pSyncNode->pRaftStore);
syncNodeBecomeLeader(pSyncNode);
syncNodeLog2("==state change become leader immediately==", pSyncNode);
@ -662,9 +709,23 @@ void syncNodeClose(SSyncNode* pSyncNode) {
taosMemoryFree(pSyncNode->pFsm);
}
for (int i = 0; i < TSDB_MAX_REPLICA; ++i) {
if ((pSyncNode->senders)[i] != NULL) {
snapshotSenderDestroy((pSyncNode->senders)[i]);
(pSyncNode->senders)[i] = NULL;
}
}
if (pSyncNode->pNewNodeReceiver != NULL) {
snapshotReceiverDestroy(pSyncNode->pNewNodeReceiver);
pSyncNode->pNewNodeReceiver = NULL;
}
/*
if (pSyncNode->pSnapshot != NULL) {
taosMemoryFree(pSyncNode->pSnapshot);
}
*/
// tsem_destroy(&pSyncNode->restoreSem);
@ -672,6 +733,9 @@ void syncNodeClose(SSyncNode* pSyncNode) {
// taosMemoryFree(pSyncNode);
}
// option
bool syncNodeSnapshotEnable(SSyncNode* pSyncNode) { return pSyncNode->pRaftCfg->snapshotEnable; }
// ping --------------
int32_t syncNodePing(SSyncNode* pSyncNode, const SRaftId* destRaftId, SyncPing* pMsg) {
syncPingLog2((char*)"==syncNodePing==", pMsg);
@ -762,7 +826,13 @@ int32_t syncNodeRestartElectTimer(SSyncNode* pSyncNode, int32_t ms) {
int32_t syncNodeResetElectTimer(SSyncNode* pSyncNode) {
int32_t ret = 0;
int32_t electMS = syncUtilElectRandomMS(pSyncNode->electBaseLine, 2 * pSyncNode->electBaseLine);
int32_t electMS;
if (pSyncNode->pRaftCfg->isStandBy) {
electMS = TIMER_MAX_MS;
} else {
electMS = syncUtilElectRandomMS(pSyncNode->electBaseLine, 2 * pSyncNode->electBaseLine);
}
ret = syncNodeRestartElectTimer(pSyncNode, electMS);
return ret;
}
@ -788,6 +858,13 @@ int32_t syncNodeSendMsgById(const SRaftId* destRaftId, SSyncNode* pSyncNode, SRp
SEpSet epSet;
syncUtilraftId2EpSet(destRaftId, &epSet);
if (pSyncNode->FpSendMsg != NULL) {
if (gRaftDetailLog) {
char* JsonStr = syncRpcMsg2Str(pMsg);
syncUtilJson2Line(JsonStr);
sTrace("sync send msg, vgId:%d, type:%d, msg:%s", pSyncNode->vgId, pMsg->msgType, JsonStr);
taosMemoryFree(JsonStr);
}
// htonl
syncUtilMsgHtoN(pMsg->pCont);
@ -952,6 +1029,20 @@ cJSON* syncNode2Json(const SSyncNode* pSyncNode) {
cJSON_AddStringToObject(pRoot, "FpOnAppendEntriesReply", u64buf);
snprintf(u64buf, sizeof(u64buf), "%p", pSyncNode->FpOnTimeout);
cJSON_AddStringToObject(pRoot, "FpOnTimeout", u64buf);
// restoreFinish
cJSON_AddNumberToObject(pRoot, "restoreFinish", pSyncNode->restoreFinish);
// snapshot senders
cJSON* pSenders = cJSON_CreateArray();
cJSON_AddItemToObject(pRoot, "senders", pSenders);
for (int i = 0; i < TSDB_MAX_REPLICA; ++i) {
cJSON_AddItemToArray(pSenders, snapshotSender2Json((pSyncNode->senders)[i]));
}
// snapshot receivers
cJSON* pReceivers = cJSON_CreateArray();
cJSON_AddItemToObject(pRoot, "receiver", snapshotReceiver2Json(pSyncNode->pNewNodeReceiver));
}
cJSON* pJson = cJSON_CreateObject();
@ -973,10 +1064,10 @@ char* syncNode2SimpleStr(const SSyncNode* pSyncNode) {
"syncNode2SimpleStr vgId:%d currentTerm:%lu, commitIndex:%ld, state:%d %s, isStandBy:%d, "
"electTimerLogicClock:%lu, "
"electTimerLogicClockUser:%lu, "
"electTimerMS:%d",
"electTimerMS:%d, replicaNum:%d",
pSyncNode->vgId, pSyncNode->pRaftStore->currentTerm, pSyncNode->commitIndex, pSyncNode->state,
syncUtilState2String(pSyncNode->state), pSyncNode->pRaftCfg->isStandBy, pSyncNode->electTimerLogicClock,
pSyncNode->electTimerLogicClockUser, pSyncNode->electTimerMS);
pSyncNode->electTimerLogicClockUser, pSyncNode->electTimerMS, pSyncNode->replicaNum);
return s;
}
@ -1013,6 +1104,8 @@ void syncNodeUpdateConfig(SSyncNode* pSyncNode, SSyncCfg* newConfig, bool* isDro
voteGrantedUpdate(pSyncNode->pVotesGranted, pSyncNode);
votesRespondUpdate(pSyncNode->pVotesRespond, pSyncNode);
pSyncNode->quorum = syncUtilQuorum(pSyncNode->pRaftCfg->cfg.replicaNum);
// isDrop
*isDrop = true;
bool IamInOld, IamInNew;
@ -1103,7 +1196,15 @@ void syncNodeBecomeLeader(SSyncNode* pSyncNode) {
for (int i = 0; i < pSyncNode->pNextIndex->replicaNum; ++i) {
// maybe overwrite myself, no harm
// just do it!
pSyncNode->pNextIndex->index[i] = pSyncNode->pLogStore->getLastIndex(pSyncNode->pLogStore) + 1;
// pSyncNode->pNextIndex->index[i] = pSyncNode->pLogStore->getLastIndex(pSyncNode->pLogStore) + 1;
// maybe wal is deleted
SyncIndex lastIndex;
SyncTerm lastTerm;
int32_t code = syncNodeGetLastIndexTerm(pSyncNode, &lastIndex, &lastTerm);
ASSERT(code == 0);
pSyncNode->pNextIndex->index[i] = lastIndex + 1;
}
for (int i = 0; i < pSyncNode->pMatchIndex->replicaNum; ++i) {
@ -1112,6 +1213,17 @@ void syncNodeBecomeLeader(SSyncNode* pSyncNode) {
pSyncNode->pMatchIndex->index[i] = SYNC_INDEX_INVALID;
}
// update sender private term
SSyncSnapshotSender* pMySender = syncNodeGetSnapshotSender(pSyncNode, &(pSyncNode->myRaftId));
if (pMySender != NULL) {
for (int i = 0; i < pSyncNode->pMatchIndex->replicaNum; ++i) {
if ((pSyncNode->senders)[i]->privateTerm > pMySender->privateTerm) {
pMySender->privateTerm = (pSyncNode->senders)[i]->privateTerm;
}
}
(pMySender->privateTerm) += 100;
}
// stop elect timer
syncNodeStopElectTimer(pSyncNode);
@ -1186,6 +1298,153 @@ void syncNodeVoteForSelf(SSyncNode* pSyncNode) {
syncRequestVoteReplyDestroy(pMsg);
}
// snapshot --------------
bool syncNodeHasSnapshot(SSyncNode* pSyncNode) {
bool ret = false;
SSnapshot snapshot = {.data = NULL, .lastApplyIndex = -1, .lastApplyTerm = 0};
if (pSyncNode->pFsm->FpGetSnapshot != NULL) {
pSyncNode->pFsm->FpGetSnapshot(pSyncNode->pFsm, &snapshot);
if (snapshot.lastApplyIndex >= SYNC_INDEX_BEGIN) {
ret = true;
}
}
return ret;
}
bool syncNodeIsIndexInSnapshot(SSyncNode* pSyncNode, SyncIndex index) {
ASSERT(syncNodeHasSnapshot(pSyncNode));
ASSERT(pSyncNode->pFsm->FpGetSnapshot != NULL);
ASSERT(index >= SYNC_INDEX_BEGIN);
SSnapshot snapshot;
pSyncNode->pFsm->FpGetSnapshot(pSyncNode->pFsm, &snapshot);
bool b = (index <= snapshot.lastApplyIndex);
return b;
}
SyncIndex syncNodeGetLastIndex(SSyncNode* pSyncNode) {
SSnapshot snapshot = {.data = NULL, .lastApplyIndex = -1, .lastApplyTerm = 0};
if (pSyncNode->pFsm->FpGetSnapshot != NULL) {
pSyncNode->pFsm->FpGetSnapshot(pSyncNode->pFsm, &snapshot);
}
SyncIndex logLastIndex = pSyncNode->pLogStore->syncLogLastIndex(pSyncNode->pLogStore);
SyncIndex lastIndex = logLastIndex > snapshot.lastApplyIndex ? logLastIndex : snapshot.lastApplyIndex;
return lastIndex;
}
SyncTerm syncNodeGetLastTerm(SSyncNode* pSyncNode) {
SyncTerm lastTerm = 0;
if (syncNodeHasSnapshot(pSyncNode)) {
// has snapshot
SSnapshot snapshot = {.data = NULL, .lastApplyIndex = -1, .lastApplyTerm = 0};
if (pSyncNode->pFsm->FpGetSnapshot != NULL) {
pSyncNode->pFsm->FpGetSnapshot(pSyncNode->pFsm, &snapshot);
}
SyncIndex logLastIndex = pSyncNode->pLogStore->syncLogLastIndex(pSyncNode->pLogStore);
if (logLastIndex > snapshot.lastApplyIndex) {
lastTerm = pSyncNode->pLogStore->syncLogLastTerm(pSyncNode->pLogStore);
} else {
lastTerm = snapshot.lastApplyTerm;
}
} else {
// no snapshot
lastTerm = pSyncNode->pLogStore->syncLogLastTerm(pSyncNode->pLogStore);
}
return lastTerm;
}
// get last index and term along with snapshot
int32_t syncNodeGetLastIndexTerm(SSyncNode* pSyncNode, SyncIndex* pLastIndex, SyncTerm* pLastTerm) {
*pLastIndex = syncNodeGetLastIndex(pSyncNode);
*pLastTerm = syncNodeGetLastTerm(pSyncNode);
return 0;
}
SyncIndex syncNodeSyncStartIndex(SSyncNode* pSyncNode) {
SyncIndex syncStartIndex = syncNodeGetLastIndex(pSyncNode) + 1;
return syncStartIndex;
}
SyncIndex syncNodeGetPreIndex(SSyncNode* pSyncNode, SyncIndex index) {
ASSERT(index >= SYNC_INDEX_BEGIN);
SyncIndex syncStartIndex = syncNodeSyncStartIndex(pSyncNode);
ASSERT(index <= syncStartIndex);
SyncIndex preIndex = index - 1;
return preIndex;
}
SyncTerm syncNodeGetPreTerm(SSyncNode* pSyncNode, SyncIndex index) {
ASSERT(index >= SYNC_INDEX_BEGIN);
SyncIndex syncStartIndex = syncNodeSyncStartIndex(pSyncNode);
ASSERT(index <= syncStartIndex);
if (index == SYNC_INDEX_BEGIN) {
return 0;
}
SyncTerm preTerm = 0;
if (syncNodeHasSnapshot(pSyncNode)) {
// has snapshot
SSnapshot snapshot = {.data = NULL, .lastApplyIndex = -1, .lastApplyTerm = 0};
if (pSyncNode->pFsm->FpGetSnapshot != NULL) {
pSyncNode->pFsm->FpGetSnapshot(pSyncNode->pFsm, &snapshot);
}
if (index > snapshot.lastApplyIndex + 1) {
// should be log preTerm
SSyncRaftEntry* pPreEntry = NULL;
int32_t code = pSyncNode->pLogStore->syncLogGetEntry(pSyncNode->pLogStore, index - 1, &pPreEntry);
ASSERT(code == 0);
ASSERT(pPreEntry != NULL);
preTerm = pPreEntry->term;
taosMemoryFree(pPreEntry);
} else if (index == snapshot.lastApplyIndex + 1) {
preTerm = snapshot.lastApplyTerm;
} else {
// maybe snapshot change
sError("sync get pre term, bad scene. index:%ld", index);
logStoreLog2("sync get pre term, bad scene", pSyncNode->pLogStore);
SSyncRaftEntry* pPreEntry = NULL;
int32_t code = pSyncNode->pLogStore->syncLogGetEntry(pSyncNode->pLogStore, index - 1, &pPreEntry);
ASSERT(code == 0);
ASSERT(pPreEntry != NULL);
preTerm = pPreEntry->term;
taosMemoryFree(pPreEntry);
}
} else {
// no snapshot
ASSERT(index > SYNC_INDEX_BEGIN);
SSyncRaftEntry* pPreEntry = NULL;
int32_t code = pSyncNode->pLogStore->syncLogGetEntry(pSyncNode->pLogStore, index - 1, &pPreEntry);
ASSERT(code == 0);
ASSERT(pPreEntry != NULL);
preTerm = pPreEntry->term;
taosMemoryFree(pPreEntry);
}
return preTerm;
}
// get pre index and term of "index"
int32_t syncNodeGetPreIndexTerm(SSyncNode* pSyncNode, SyncIndex index, SyncIndex* pPreIndex, SyncTerm* pPreTerm) {
*pPreIndex = syncNodeGetPreIndex(pSyncNode, index);
*pPreTerm = syncNodeGetPreTerm(pSyncNode, index);
return 0;
}
// for debug --------------
void syncNodePrint(SSyncNode* pObj) {
char* serialized = syncNode2Str(pObj);
@ -1327,7 +1586,8 @@ static int32_t syncNodeAppendNoop(SSyncNode* ths) {
assert(pEntry != NULL);
if (ths->state == TAOS_SYNC_STATE_LEADER) {
ths->pLogStore->appendEntry(ths->pLogStore, pEntry);
// ths->pLogStore->appendEntry(ths->pLogStore, pEntry);
ths->pLogStore->syncLogAppendEntry(ths->pLogStore, pEntry);
syncNodeReplicate(ths);
}
@ -1383,13 +1643,14 @@ int32_t syncNodeOnClientRequestCb(SSyncNode* ths, SyncClientRequest* pMsg) {
int32_t ret = 0;
syncClientRequestLog2("==syncNodeOnClientRequestCb==", pMsg);
SyncIndex index = ths->pLogStore->getLastIndex(ths->pLogStore) + 1;
SyncIndex index = ths->pLogStore->syncLogWriteIndex(ths->pLogStore);
SyncTerm term = ths->pRaftStore->currentTerm;
SSyncRaftEntry* pEntry = syncEntryBuild2((SyncClientRequest*)pMsg, term, index);
assert(pEntry != NULL);
if (ths->state == TAOS_SYNC_STATE_LEADER) {
ths->pLogStore->appendEntry(ths->pLogStore, pEntry);
// ths->pLogStore->appendEntry(ths->pLogStore, pEntry);
ths->pLogStore->syncLogAppendEntry(ths->pLogStore, pEntry);
// start replicate right now!
syncNodeReplicate(ths);
@ -1459,3 +1720,137 @@ const char* syncStr(ESyncState state) {
return "error";
}
}
int32_t syncNodeCommit(SSyncNode* ths, SyncIndex beginIndex, SyncIndex endIndex, uint64_t flag) {
int32_t code = 0;
ESyncState state = flag;
sInfo("sync event commit from index:%" PRId64 " to index:%" PRId64 ", %s", beginIndex, endIndex,
syncUtilState2String(state));
// maybe execute by leader, skip snapshot
SSnapshot snapshot = {.data = NULL, .lastApplyIndex = -1, .lastApplyTerm = 0};
if (ths->pFsm->FpGetSnapshot != NULL) {
ths->pFsm->FpGetSnapshot(ths->pFsm, &snapshot);
}
if (beginIndex <= snapshot.lastApplyIndex) {
beginIndex = snapshot.lastApplyIndex + 1;
}
// execute fsm
if (ths->pFsm != NULL) {
for (SyncIndex i = beginIndex; i <= endIndex; ++i) {
if (i != SYNC_INDEX_INVALID) {
SSyncRaftEntry* pEntry;
code = ths->pLogStore->syncLogGetEntry(ths->pLogStore, i, &pEntry);
ASSERT(code == 0);
ASSERT(pEntry != NULL);
SRpcMsg rpcMsg;
syncEntry2OriginalRpc(pEntry, &rpcMsg);
if (ths->pFsm->FpCommitCb != NULL && syncUtilUserCommit(pEntry->originalRpcType)) {
SFsmCbMeta cbMeta;
cbMeta.index = pEntry->index;
cbMeta.isWeak = pEntry->isWeak;
cbMeta.code = 0;
cbMeta.state = ths->state;
cbMeta.seqNum = pEntry->seqNum;
cbMeta.term = pEntry->term;
cbMeta.currentTerm = ths->pRaftStore->currentTerm;
cbMeta.flag = flag;
ths->pFsm->FpCommitCb(ths->pFsm, &rpcMsg, cbMeta);
}
// config change
if (pEntry->originalRpcType == TDMT_SYNC_CONFIG_CHANGE) {
SSyncCfg oldSyncCfg = ths->pRaftCfg->cfg;
SSyncCfg newSyncCfg;
int32_t ret = syncCfgFromStr(rpcMsg.pCont, &newSyncCfg);
ASSERT(ret == 0);
// update new config myIndex
bool hit = false;
for (int i = 0; i < newSyncCfg.replicaNum; ++i) {
if (strcmp(ths->myNodeInfo.nodeFqdn, (newSyncCfg.nodeInfo)[i].nodeFqdn) == 0 &&
ths->myNodeInfo.nodePort == (newSyncCfg.nodeInfo)[i].nodePort) {
newSyncCfg.myIndex = i;
hit = true;
break;
}
}
SReConfigCbMeta cbMeta = {0};
bool isDrop;
// I am in newConfig
if (hit) {
syncNodeUpdateConfig(ths, &newSyncCfg, &isDrop);
// change isStandBy to normal
if (!isDrop) {
if (ths->state == TAOS_SYNC_STATE_LEADER) {
syncNodeBecomeLeader(ths);
} else {
syncNodeBecomeFollower(ths);
}
}
char* sOld = syncCfg2Str(&oldSyncCfg);
char* sNew = syncCfg2Str(&newSyncCfg);
sInfo("==config change== 0x11 old:%s new:%s isDrop:%d \n", sOld, sNew, isDrop);
taosMemoryFree(sOld);
taosMemoryFree(sNew);
}
// always call FpReConfigCb
if (ths->pFsm->FpReConfigCb != NULL) {
cbMeta.code = 0;
cbMeta.currentTerm = ths->pRaftStore->currentTerm;
cbMeta.index = pEntry->index;
cbMeta.term = pEntry->term;
cbMeta.oldCfg = oldSyncCfg;
cbMeta.flag = 0x11;
cbMeta.isDrop = isDrop;
ths->pFsm->FpReConfigCb(ths->pFsm, newSyncCfg, cbMeta);
}
}
// restore finish
if (pEntry->index == ths->pLogStore->syncLogLastIndex(ths->pLogStore)) {
if (ths->restoreFinish == false) {
if (ths->pFsm->FpRestoreFinishCb != NULL) {
ths->pFsm->FpRestoreFinishCb(ths->pFsm);
}
ths->restoreFinish = true;
sInfo("restore finish %p vgId:%d", ths, ths->vgId);
}
}
rpcFreeCont(rpcMsg.pCont);
syncEntryDestory(pEntry);
}
}
}
return 0;
}
bool syncNodeInRaftGroup(SSyncNode* ths, SRaftId* pRaftId) {
for (int i = 0; i < ths->replicaNum; ++i) {
if (syncUtilSameId(&((ths->replicasId)[i]), pRaftId)) {
return true;
}
}
return false;
}
SSyncSnapshotSender* syncNodeGetSnapshotSender(SSyncNode* ths, SRaftId* pDestId) {
SSyncSnapshotSender* pSender = NULL;
for (int i = 0; i < ths->replicaNum; ++i) {
if (syncUtilSameId(pDestId, &((ths->replicasId)[i]))) {
pSender = (ths->senders)[i];
}
}
return pSender;
}

424
source/libs/sync/src/syncMessage.c
View File

@ -65,6 +65,16 @@ cJSON* syncRpcMsg2Json(SRpcMsg* pRpcMsg) {
pRoot = syncAppendEntriesReply2Json(pSyncMsg);
syncAppendEntriesReplyDestroy(pSyncMsg);
} else if (pRpcMsg->msgType == TDMT_SYNC_SNAPSHOT_SEND) {
SyncSnapshotSend* pSyncMsg = syncSnapshotSendDeserialize2(pRpcMsg->pCont, pRpcMsg->contLen);
pRoot = syncSnapshotSend2Json(pSyncMsg);
syncSnapshotSendDestroy(pSyncMsg);
} else if (pRpcMsg->msgType == TDMT_SYNC_SNAPSHOT_RSP) {
SyncSnapshotRsp* pSyncMsg = syncSnapshotRspDeserialize2(pRpcMsg->pCont, pRpcMsg->contLen);
pRoot = syncSnapshotRsp2Json(pSyncMsg);
syncSnapshotRspDestroy(pSyncMsg);
} else if (pRpcMsg->msgType == TDMT_SYNC_COMMON_RESPONSE) {
pRoot = cJSON_CreateObject();
char* s;
@ -135,9 +145,11 @@ void syncRpcMsgLog(SRpcMsg* pMsg) {
}
void syncRpcMsgLog2(char* s, SRpcMsg* pMsg) {
char* serialized = syncRpcMsg2Str(pMsg);
sTrace("syncRpcMsgLog2 | len:%lu | %s | %s", strlen(serialized), s, serialized);
taosMemoryFree(serialized);
if (gRaftDetailLog) {
char* serialized = syncRpcMsg2Str(pMsg);
sTrace("syncRpcMsgLog2 | len:%lu | %s | %s", strlen(serialized), s, serialized);
taosMemoryFree(serialized);
}
}
// ---- message process SyncTimeout----
@ -264,9 +276,11 @@ void syncTimeoutLog(const SyncTimeout* pMsg) {
}
void syncTimeoutLog2(char* s, const SyncTimeout* pMsg) {
char* serialized = syncTimeout2Str(pMsg);
sTrace("syncTimeoutLog2 | len:%lu | %s | %s", strlen(serialized), s, serialized);
taosMemoryFree(serialized);
if (gRaftDetailLog) {
char* serialized = syncTimeout2Str(pMsg);
sTrace("syncTimeoutLog2 | len:%lu | %s | %s", strlen(serialized), s, serialized);
taosMemoryFree(serialized);
}
}
// ---- message process SyncPing----
@ -524,9 +538,11 @@ void syncPingLog(const SyncPing* pMsg) {
}
void syncPingLog2(char* s, const SyncPing* pMsg) {
char* serialized = syncPing2Str(pMsg);
sTrace("syncPingLog2 | len:%lu | %s | %s", strlen(serialized), s, serialized);
taosMemoryFree(serialized);
if (gRaftDetailLog) {
char* serialized = syncPing2Str(pMsg);
sTrace("syncPingLog2 | len:%lu | %s | %s", strlen(serialized), s, serialized);
taosMemoryFree(serialized);
}
}
// ---- message process SyncPingReply----
@ -784,9 +800,11 @@ void syncPingReplyLog(const SyncPingReply* pMsg) {
}
void syncPingReplyLog2(char* s, const SyncPingReply* pMsg) {
char* serialized = syncPingReply2Str(pMsg);
sTrace("syncPingReplyLog2 | len:%zu | %s | %s", strlen(serialized), s, serialized);
taosMemoryFree(serialized);
if (gRaftDetailLog) {
char* serialized = syncPingReply2Str(pMsg);
sTrace("syncPingReplyLog2 | len:%zu | %s | %s", strlen(serialized), s, serialized);
taosMemoryFree(serialized);
}
}
// ---- message process SyncClientRequest----
@ -925,9 +943,11 @@ void syncClientRequestLog(const SyncClientRequest* pMsg) {
}
void syncClientRequestLog2(char* s, const SyncClientRequest* pMsg) {
char* serialized = syncClientRequest2Str(pMsg);
sTrace("syncClientRequestLog2 | len:%lu | %s | %s", strlen(serialized), s, serialized);
taosMemoryFree(serialized);
if (gRaftDetailLog) {
char* serialized = syncClientRequest2Str(pMsg);
sTrace("syncClientRequestLog2 | len:%lu | %s | %s", strlen(serialized), s, serialized);
taosMemoryFree(serialized);
}
}
// ---- message process SyncRequestVote----
@ -1074,9 +1094,11 @@ void syncRequestVoteLog(const SyncRequestVote* pMsg) {
}
void syncRequestVoteLog2(char* s, const SyncRequestVote* pMsg) {
char* serialized = syncRequestVote2Str(pMsg);
sTrace("syncRequestVoteLog2 | len:%lu | %s | %s", strlen(serialized), s, serialized);
taosMemoryFree(serialized);
if (gRaftDetailLog) {
char* serialized = syncRequestVote2Str(pMsg);
sTrace("syncRequestVoteLog2 | len:%lu | %s | %s", strlen(serialized), s, serialized);
taosMemoryFree(serialized);
}
}
// ---- message process SyncRequestVoteReply----
@ -1220,9 +1242,11 @@ void syncRequestVoteReplyLog(const SyncRequestVoteReply* pMsg) {
}
void syncRequestVoteReplyLog2(char* s, const SyncRequestVoteReply* pMsg) {
char* serialized = syncRequestVoteReply2Str(pMsg);
sTrace("syncRequestVoteReplyLog2 | len:%lu | %s | %s", strlen(serialized), s, serialized);
taosMemoryFree(serialized);
if (gRaftDetailLog) {
char* serialized = syncRequestVoteReply2Str(pMsg);
sTrace("syncRequestVoteReplyLog2 | len:%lu | %s | %s", strlen(serialized), s, serialized);
taosMemoryFree(serialized);
}
}
// ---- message process SyncAppendEntries----
@ -1333,6 +1357,9 @@ cJSON* syncAppendEntries2Json(const SyncAppendEntries* pMsg) {
snprintf(u64buf, sizeof(u64buf), "%lu", pMsg->term);
cJSON_AddStringToObject(pRoot, "term", u64buf);
snprintf(u64buf, sizeof(u64buf), "%lu", pMsg->privateTerm);
cJSON_AddStringToObject(pRoot, "privateTerm", u64buf);
snprintf(u64buf, sizeof(u64buf), "%ld", pMsg->prevLogIndex);
cJSON_AddStringToObject(pRoot, "prevLogIndex", u64buf);
@ -1386,9 +1413,11 @@ void syncAppendEntriesLog(const SyncAppendEntries* pMsg) {
}
void syncAppendEntriesLog2(char* s, const SyncAppendEntries* pMsg) {
char* serialized = syncAppendEntries2Str(pMsg);
sTrace("syncAppendEntriesLog2 | len:%lu | %s | %s", strlen(serialized), s, serialized);
taosMemoryFree(serialized);
if (gRaftDetailLog) {
char* serialized = syncAppendEntries2Str(pMsg);
sTrace("syncAppendEntriesLog2 | len:%lu | %s | %s", strlen(serialized), s, serialized);
taosMemoryFree(serialized);
}
}
// ---- message process SyncAppendEntriesReply----
@ -1494,6 +1523,9 @@ cJSON* syncAppendEntriesReply2Json(const SyncAppendEntriesReply* pMsg) {
cJSON_AddNumberToObject(pDestId, "vgId", pMsg->destId.vgId);
cJSON_AddItemToObject(pRoot, "destId", pDestId);
snprintf(u64buf, sizeof(u64buf), "%lu", pMsg->privateTerm);
cJSON_AddStringToObject(pRoot, "privateTerm", u64buf);
snprintf(u64buf, sizeof(u64buf), "%lu", pMsg->term);
cJSON_AddStringToObject(pRoot, "term", u64buf);
cJSON_AddNumberToObject(pRoot, "success", pMsg->success);
@ -1535,9 +1567,11 @@ void syncAppendEntriesReplyLog(const SyncAppendEntriesReply* pMsg) {
}
void syncAppendEntriesReplyLog2(char* s, const SyncAppendEntriesReply* pMsg) {
char* serialized = syncAppendEntriesReply2Str(pMsg);
sTrace("syncAppendEntriesReplyLog2 | len:%lu | %s | %s", strlen(serialized), s, serialized);
taosMemoryFree(serialized);
if (gRaftDetailLog) {
char* serialized = syncAppendEntriesReply2Str(pMsg);
sTrace("syncAppendEntriesReplyLog2 | len:%lu | %s | %s", strlen(serialized), s, serialized);
taosMemoryFree(serialized);
}
}
// ---- message process SyncApplyMsg----
@ -1686,7 +1720,339 @@ void syncApplyMsgLog(const SyncApplyMsg* pMsg) {
}
void syncApplyMsgLog2(char* s, const SyncApplyMsg* pMsg) {
char* serialized = syncApplyMsg2Str(pMsg);
sTrace("syncApplyMsgLog2 | len:%lu | %s | %s", strlen(serialized), s, serialized);
if (gRaftDetailLog) {
char* serialized = syncApplyMsg2Str(pMsg);
sTrace("syncApplyMsgLog2 | len:%lu | %s | %s", strlen(serialized), s, serialized);
taosMemoryFree(serialized);
}
}
// ---------------------------------------------
SyncSnapshotSend* syncSnapshotSendBuild(uint32_t dataLen, int32_t vgId) {
uint32_t bytes = sizeof(SyncSnapshotSend) + dataLen;
SyncSnapshotSend* pMsg = taosMemoryMalloc(bytes);
memset(pMsg, 0, bytes);
pMsg->bytes = bytes;
pMsg->vgId = vgId;
pMsg->msgType = TDMT_SYNC_SNAPSHOT_SEND;
pMsg->dataLen = dataLen;
return pMsg;
}
void syncSnapshotSendDestroy(SyncSnapshotSend* pMsg) {
if (pMsg != NULL) {
taosMemoryFree(pMsg);
}
}
void syncSnapshotSendSerialize(const SyncSnapshotSend* pMsg, char* buf, uint32_t bufLen) {
assert(pMsg->bytes <= bufLen);
memcpy(buf, pMsg, pMsg->bytes);
}
void syncSnapshotSendDeserialize(const char* buf, uint32_t len, SyncSnapshotSend* pMsg) {
memcpy(pMsg, buf, len);
assert(len == pMsg->bytes);
assert(pMsg->bytes == sizeof(SyncSnapshotSend) + pMsg->dataLen);
}
char* syncSnapshotSendSerialize2(const SyncSnapshotSend* pMsg, uint32_t* len) {
char* buf = taosMemoryMalloc(pMsg->bytes);
assert(buf != NULL);
syncSnapshotSendSerialize(pMsg, buf, pMsg->bytes);
if (len != NULL) {
*len = pMsg->bytes;
}
return buf;
}
SyncSnapshotSend* syncSnapshotSendDeserialize2(const char* buf, uint32_t len) {
uint32_t bytes = *((uint32_t*)buf);
SyncSnapshotSend* pMsg = taosMemoryMalloc(bytes);
assert(pMsg != NULL);
syncSnapshotSendDeserialize(buf, len, pMsg);
assert(len == pMsg->bytes);
return pMsg;
}
void syncSnapshotSend2RpcMsg(const SyncSnapshotSend* pMsg, SRpcMsg* pRpcMsg) {
memset(pRpcMsg, 0, sizeof(*pRpcMsg));
pRpcMsg->msgType = pMsg->msgType;
pRpcMsg->contLen = pMsg->bytes;
pRpcMsg->pCont = rpcMallocCont(pRpcMsg->contLen);
syncSnapshotSendSerialize(pMsg, pRpcMsg->pCont, pRpcMsg->contLen);
}
void syncSnapshotSendFromRpcMsg(const SRpcMsg* pRpcMsg, SyncSnapshotSend* pMsg) {
syncSnapshotSendDeserialize(pRpcMsg->pCont, pRpcMsg->contLen, pMsg);
}
SyncSnapshotSend* syncSnapshotSendFromRpcMsg2(const SRpcMsg* pRpcMsg) {
SyncSnapshotSend* pMsg = syncSnapshotSendDeserialize2(pRpcMsg->pCont, pRpcMsg->contLen);
assert(pMsg != NULL);
return pMsg;
}
cJSON* syncSnapshotSend2Json(const SyncSnapshotSend* pMsg) {
char u64buf[128];
cJSON* pRoot = cJSON_CreateObject();
if (pMsg != NULL) {
cJSON_AddNumberToObject(pRoot, "bytes", pMsg->bytes);
cJSON_AddNumberToObject(pRoot, "vgId", pMsg->vgId);
cJSON_AddNumberToObject(pRoot, "msgType", pMsg->msgType);
cJSON* pSrcId = cJSON_CreateObject();
snprintf(u64buf, sizeof(u64buf), "%lu", pMsg->srcId.addr);
cJSON_AddStringToObject(pSrcId, "addr", u64buf);
{
uint64_t u64 = pMsg->srcId.addr;
cJSON* pTmp = pSrcId;
char host[128];
uint16_t port;
syncUtilU642Addr(u64, host, sizeof(host), &port);
cJSON_AddStringToObject(pTmp, "addr_host", host);
cJSON_AddNumberToObject(pTmp, "addr_port", port);
}
cJSON_AddNumberToObject(pSrcId, "vgId", pMsg->srcId.vgId);
cJSON_AddItemToObject(pRoot, "srcId", pSrcId);
cJSON* pDestId = cJSON_CreateObject();
snprintf(u64buf, sizeof(u64buf), "%lu", pMsg->destId.addr);
cJSON_AddStringToObject(pDestId, "addr", u64buf);
{
uint64_t u64 = pMsg->destId.addr;
cJSON* pTmp = pDestId;
char host[128];
uint16_t port;
syncUtilU642Addr(u64, host, sizeof(host), &port);
cJSON_AddStringToObject(pTmp, "addr_host", host);
cJSON_AddNumberToObject(pTmp, "addr_port", port);
}
cJSON_AddNumberToObject(pDestId, "vgId", pMsg->destId.vgId);
cJSON_AddItemToObject(pRoot, "destId", pDestId);
snprintf(u64buf, sizeof(u64buf), "%lu", pMsg->term);
cJSON_AddStringToObject(pRoot, "term", u64buf);
snprintf(u64buf, sizeof(u64buf), "%lu", pMsg->privateTerm);
cJSON_AddStringToObject(pRoot, "privateTerm", u64buf);
snprintf(u64buf, sizeof(u64buf), "%ld", pMsg->lastIndex);
cJSON_AddStringToObject(pRoot, "lastIndex", u64buf);
snprintf(u64buf, sizeof(u64buf), "%lu", pMsg->lastTerm);
cJSON_AddStringToObject(pRoot, "lastTerm", u64buf);
cJSON_AddNumberToObject(pRoot, "seq", pMsg->seq);
cJSON_AddNumberToObject(pRoot, "dataLen", pMsg->dataLen);
char* s;
s = syncUtilprintBin((char*)(pMsg->data), pMsg->dataLen);
cJSON_AddStringToObject(pRoot, "data", s);
taosMemoryFree(s);
s = syncUtilprintBin2((char*)(pMsg->data), pMsg->dataLen);
cJSON_AddStringToObject(pRoot, "data2", s);
taosMemoryFree(s);
}
cJSON* pJson = cJSON_CreateObject();
cJSON_AddItemToObject(pJson, "SyncSnapshotSend", pRoot);
return pJson;
}
char* syncSnapshotSend2Str(const SyncSnapshotSend* pMsg) {
cJSON* pJson = syncSnapshotSend2Json(pMsg);
char* serialized = cJSON_Print(pJson);
cJSON_Delete(pJson);
return serialized;
}
// for debug ----------------------
void syncSnapshotSendPrint(const SyncSnapshotSend* pMsg) {
char* serialized = syncSnapshotSend2Str(pMsg);
printf("syncSnapshotSendPrint | len:%lu | %s \n", strlen(serialized), serialized);
fflush(NULL);
taosMemoryFree(serialized);
}
void syncSnapshotSendPrint2(char* s, const SyncSnapshotSend* pMsg) {
char* serialized = syncSnapshotSend2Str(pMsg);
printf("syncSnapshotSendPrint2 | len:%lu | %s | %s \n", strlen(serialized), s, serialized);
fflush(NULL);
taosMemoryFree(serialized);
}
void syncSnapshotSendLog(const SyncSnapshotSend* pMsg) {
char* serialized = syncSnapshotSend2Str(pMsg);
sTrace("syncSnapshotSendLog | len:%lu | %s", strlen(serialized), serialized);
taosMemoryFree(serialized);
}
void syncSnapshotSendLog2(char* s, const SyncSnapshotSend* pMsg) {
if (gRaftDetailLog) {
char* serialized = syncSnapshotSend2Str(pMsg);
sTrace("syncSnapshotSendLog2 | len:%lu | %s | %s", strlen(serialized), s, serialized);
taosMemoryFree(serialized);
}
}
// ---------------------------------------------
SyncSnapshotRsp* syncSnapshotRspBuild(int32_t vgId) {
uint32_t bytes = sizeof(SyncSnapshotRsp);
SyncSnapshotRsp* pMsg = taosMemoryMalloc(bytes);
memset(pMsg, 0, bytes);
pMsg->bytes = bytes;
pMsg->vgId = vgId;
pMsg->msgType = TDMT_SYNC_SNAPSHOT_RSP;
return pMsg;
}
void syncSnapshotRspDestroy(SyncSnapshotRsp* pMsg) {
if (pMsg != NULL) {
taosMemoryFree(pMsg);
}
}
void syncSnapshotRspSerialize(const SyncSnapshotRsp* pMsg, char* buf, uint32_t bufLen) {
assert(pMsg->bytes <= bufLen);
memcpy(buf, pMsg, pMsg->bytes);
}
void syncSnapshotRspDeserialize(const char* buf, uint32_t len, SyncSnapshotRsp* pMsg) {
memcpy(pMsg, buf, len);
assert(len == pMsg->bytes);
}
char* syncSnapshotRspSerialize2(const SyncSnapshotRsp* pMsg, uint32_t* len) {
char* buf = taosMemoryMalloc(pMsg->bytes);
assert(buf != NULL);
syncSnapshotRspSerialize(pMsg, buf, pMsg->bytes);
if (len != NULL) {
*len = pMsg->bytes;
}
return buf;
}
SyncSnapshotRsp* syncSnapshotRspDeserialize2(const char* buf, uint32_t len) {
uint32_t bytes = *((uint32_t*)buf);
SyncSnapshotRsp* pMsg = taosMemoryMalloc(bytes);
assert(pMsg != NULL);
syncSnapshotRspDeserialize(buf, len, pMsg);
assert(len == pMsg->bytes);
return pMsg;
}
void syncSnapshotRsp2RpcMsg(const SyncSnapshotRsp* pMsg, SRpcMsg* pRpcMsg) {
memset(pRpcMsg, 0, sizeof(*pRpcMsg));
pRpcMsg->msgType = pMsg->msgType;
pRpcMsg->contLen = pMsg->bytes;
pRpcMsg->pCont = rpcMallocCont(pRpcMsg->contLen);
syncSnapshotRspSerialize(pMsg, pRpcMsg->pCont, pRpcMsg->contLen);
}
void syncSnapshotRspFromRpcMsg(const SRpcMsg* pRpcMsg, SyncSnapshotRsp* pMsg) {
syncSnapshotRspDeserialize(pRpcMsg->pCont, pRpcMsg->contLen, pMsg);
}
SyncSnapshotRsp* syncSnapshotRspFromRpcMsg2(const SRpcMsg* pRpcMsg) {
SyncSnapshotRsp* pMsg = syncSnapshotRspDeserialize2(pRpcMsg->pCont, pRpcMsg->contLen);
assert(pMsg != NULL);
return pMsg;
}
cJSON* syncSnapshotRsp2Json(const SyncSnapshotRsp* pMsg) {
char u64buf[128];
cJSON* pRoot = cJSON_CreateObject();
if (pMsg != NULL) {
cJSON_AddNumberToObject(pRoot, "bytes", pMsg->bytes);
cJSON_AddNumberToObject(pRoot, "vgId", pMsg->vgId);
cJSON_AddNumberToObject(pRoot, "msgType", pMsg->msgType);
cJSON* pSrcId = cJSON_CreateObject();
snprintf(u64buf, sizeof(u64buf), "%lu", pMsg->srcId.addr);
cJSON_AddStringToObject(pSrcId, "addr", u64buf);
{
uint64_t u64 = pMsg->srcId.addr;
cJSON* pTmp = pSrcId;
char host[128];
uint16_t port;
syncUtilU642Addr(u64, host, sizeof(host), &port);
cJSON_AddStringToObject(pTmp, "addr_host", host);
cJSON_AddNumberToObject(pTmp, "addr_port", port);
}
cJSON_AddNumberToObject(pSrcId, "vgId", pMsg->srcId.vgId);
cJSON_AddItemToObject(pRoot, "srcId", pSrcId);
cJSON* pDestId = cJSON_CreateObject();
snprintf(u64buf, sizeof(u64buf), "%lu", pMsg->destId.addr);
cJSON_AddStringToObject(pDestId, "addr", u64buf);
{
uint64_t u64 = pMsg->destId.addr;
cJSON* pTmp = pDestId;
char host[128];
uint16_t port;
syncUtilU642Addr(u64, host, sizeof(host), &port);
cJSON_AddStringToObject(pTmp, "addr_host", host);
cJSON_AddNumberToObject(pTmp, "addr_port", port);
}
cJSON_AddNumberToObject(pDestId, "vgId", pMsg->destId.vgId);
cJSON_AddItemToObject(pRoot, "destId", pDestId);
snprintf(u64buf, sizeof(u64buf), "%lu", pMsg->term);
cJSON_AddStringToObject(pRoot, "term", u64buf);
snprintf(u64buf, sizeof(u64buf), "%lu", pMsg->privateTerm);
cJSON_AddStringToObject(pRoot, "privateTerm", u64buf);
snprintf(u64buf, sizeof(u64buf), "%ld", pMsg->lastIndex);
cJSON_AddStringToObject(pRoot, "lastIndex", u64buf);
snprintf(u64buf, sizeof(u64buf), "%lu", pMsg->lastTerm);
cJSON_AddStringToObject(pRoot, "lastTerm", u64buf);
cJSON_AddNumberToObject(pRoot, "ack", pMsg->ack);
cJSON_AddNumberToObject(pRoot, "code", pMsg->code);
}
cJSON* pJson = cJSON_CreateObject();
cJSON_AddItemToObject(pJson, "SyncSnapshotRsp", pRoot);
return pJson;
}
char* syncSnapshotRsp2Str(const SyncSnapshotRsp* pMsg) {
cJSON* pJson = syncSnapshotRsp2Json(pMsg);
char* serialized = cJSON_Print(pJson);
cJSON_Delete(pJson);
return serialized;
}
// for debug ----------------------
void syncSnapshotRspPrint(const SyncSnapshotRsp* pMsg) {
char* serialized = syncSnapshotRsp2Str(pMsg);
printf("syncSnapshotRspPrint | len:%lu | %s \n", strlen(serialized), serialized);
fflush(NULL);
taosMemoryFree(serialized);
}
void syncSnapshotRspPrint2(char* s, const SyncSnapshotRsp* pMsg) {
char* serialized = syncSnapshotRsp2Str(pMsg);
printf("syncSnapshotRspPrint2 | len:%lu | %s | %s \n", strlen(serialized), s, serialized);
fflush(NULL);
taosMemoryFree(serialized);
}
void syncSnapshotRspLog(const SyncSnapshotRsp* pMsg) {
char* serialized = syncSnapshotRsp2Str(pMsg);
sTrace("syncSnapshotRspLog | len:%lu | %s", strlen(serialized), serialized);
taosMemoryFree(serialized);
}
void syncSnapshotRspLog2(char* s, const SyncSnapshotRsp* pMsg) {
if (gRaftDetailLog) {
char* serialized = syncSnapshotRsp2Str(pMsg);
sTrace("syncSnapshotRspLog2 | len:%lu | %s | %s", strlen(serialized), s, serialized);
taosMemoryFree(serialized);
}
}

9
source/libs/sync/src/syncRaftCfg.c
View File

@ -148,6 +148,7 @@ cJSON *raftCfg2Json(SRaftCfg *pRaftCfg) {
cJSON *pRoot = cJSON_CreateObject();
cJSON_AddItemToObject(pRoot, "SSyncCfg", syncCfg2Json(&(pRaftCfg->cfg)));
cJSON_AddNumberToObject(pRoot, "isStandBy", pRaftCfg->isStandBy);
cJSON_AddNumberToObject(pRoot, "snapshotEnable", pRaftCfg->snapshotEnable);
cJSON *pJson = cJSON_CreateObject();
cJSON_AddItemToObject(pJson, "RaftCfg", pRoot);
@ -161,7 +162,7 @@ char *raftCfg2Str(SRaftCfg *pRaftCfg) {
return serialized;
}
int32_t raftCfgCreateFile(SSyncCfg *pCfg, int8_t isStandBy, const char *path) {
int32_t raftCfgCreateFile(SSyncCfg *pCfg, SRaftCfgMeta meta, const char *path) {
assert(pCfg != NULL);
TdFilePtr pFile = taosOpenFile(path, TD_FILE_CREATE | TD_FILE_WRITE);
@ -169,7 +170,8 @@ int32_t raftCfgCreateFile(SSyncCfg *pCfg, int8_t isStandBy, const char *path) {
SRaftCfg raftCfg;
raftCfg.cfg = *pCfg;
raftCfg.isStandBy = isStandBy;
raftCfg.isStandBy = meta.isStandBy;
raftCfg.snapshotEnable = meta.snapshotEnable;
char *s = raftCfg2Str(&raftCfg);
char buf[CONFIG_FILE_LEN] = {0};
@ -194,6 +196,9 @@ int32_t raftCfgFromJson(const cJSON *pRoot, SRaftCfg *pRaftCfg) {
cJSON *pJsonIsStandBy = cJSON_GetObjectItem(pJson, "isStandBy");
pRaftCfg->isStandBy = cJSON_GetNumberValue(pJsonIsStandBy);
cJSON *pJsonSnapshotEnable = cJSON_GetObjectItem(pJson, "snapshotEnable");
pRaftCfg->snapshotEnable = cJSON_GetNumberValue(pJsonSnapshotEnable);
cJSON * pJsonSyncCfg = cJSON_GetObjectItem(pJson, "SSyncCfg");
int32_t code = syncCfgFromJson(pJsonSyncCfg, &(pRaftCfg->cfg));
ASSERT(code == 0);

304
source/libs/sync/src/syncRaftLog.c
View File

@ -16,6 +16,23 @@
#include "syncRaftLog.h"
#include "wal.h"
// refactor, log[0 .. n] ==> log[m .. n]
static int32_t raftLogSetBeginIndex(struct SSyncLogStore* pLogStore, SyncIndex beginIndex);
static SyncIndex raftLogBeginIndex(struct SSyncLogStore* pLogStore);
static SyncIndex raftLogEndIndex(struct SSyncLogStore* pLogStore);
static SyncIndex raftLogWriteIndex(struct SSyncLogStore* pLogStore);
static bool raftLogIsEmpty(struct SSyncLogStore* pLogStore);
static int32_t raftLogEntryCount(struct SSyncLogStore* pLogStore);
static bool raftLogInRange(struct SSyncLogStore* pLogStore, SyncIndex index);
static SyncIndex raftLogLastIndex(struct SSyncLogStore* pLogStore);
static SyncTerm raftLogLastTerm(struct SSyncLogStore* pLogStore);
static int32_t raftLogAppendEntry(struct SSyncLogStore* pLogStore, SSyncRaftEntry* pEntry);
static int32_t raftLogGetEntry(struct SSyncLogStore* pLogStore, SyncIndex index, SSyncRaftEntry** ppEntry);
static int32_t raftLogTruncate(struct SSyncLogStore* pLogStore, SyncIndex fromIndex);
static int32_t raftLogGetLastEntry(SSyncLogStore* pLogStore, SSyncRaftEntry** ppLastEntry);
//-------------------------------
static SSyncRaftEntry* logStoreGetLastEntry(SSyncLogStore* pLogStore);
static SyncIndex logStoreLastIndex(SSyncLogStore* pLogStore);
static SyncTerm logStoreLastTerm(SSyncLogStore* pLogStore);
@ -25,6 +42,202 @@ static int32_t logStoreTruncate(SSyncLogStore* pLogStore, SyncIndex from
static int32_t logStoreUpdateCommitIndex(SSyncLogStore* pLogStore, SyncIndex index);
static SyncIndex logStoreGetCommitIndex(SSyncLogStore* pLogStore);
// refactor, log[0 .. n] ==> log[m .. n]
static int32_t raftLogSetBeginIndex(struct SSyncLogStore* pLogStore, SyncIndex beginIndex) {
sTrace("raftLogSetBeginIndex beginIndex:%ld", beginIndex);
// if beginIndex == 0, donot need call this funciton
ASSERT(beginIndex > 0);
SSyncLogStoreData* pData = pLogStore->data;
SWal* pWal = pData->pWal;
pData->beginIndex = beginIndex;
walRestoreFromSnapshot(pWal, beginIndex - 1);
return 0;
}
int32_t raftLogResetBeginIndex(struct SSyncLogStore* pLogStore) { return 0; }
static SyncIndex raftLogBeginIndex(struct SSyncLogStore* pLogStore) {
SSyncLogStoreData* pData = pLogStore->data;
SWal* pWal = pData->pWal;
return pData->beginIndex;
}
static SyncIndex raftLogEndIndex(struct SSyncLogStore* pLogStore) { return raftLogLastIndex(pLogStore); }
static bool raftLogIsEmpty(struct SSyncLogStore* pLogStore) {
SyncIndex beginIndex = raftLogBeginIndex(pLogStore);
SyncIndex endIndex = raftLogEndIndex(pLogStore);
return (endIndex < beginIndex);
}
static int32_t raftLogEntryCount(struct SSyncLogStore* pLogStore) {
SyncIndex beginIndex = raftLogBeginIndex(pLogStore);
SyncIndex endIndex = raftLogEndIndex(pLogStore);
int32_t count = endIndex - beginIndex + 1;
return count > 0 ? count : 0;
}
static bool raftLogInRange(struct SSyncLogStore* pLogStore, SyncIndex index) {
SyncIndex beginIndex = raftLogBeginIndex(pLogStore);
SyncIndex endIndex = raftLogEndIndex(pLogStore);
if (index >= beginIndex && index <= endIndex) {
return true;
} else {
return false;
}
}
static SyncIndex raftLogLastIndex(struct SSyncLogStore* pLogStore) {
SyncIndex lastIndex;
SSyncLogStoreData* pData = pLogStore->data;
SWal* pWal = pData->pWal;
SyncIndex lastVer = walGetLastVer(pWal);
SyncIndex firstVer = walGetFirstVer(pWal);
if (lastVer < firstVer) {
// no record
lastIndex = -1;
} else {
if (firstVer >= 0) {
lastIndex = lastVer;
} else if (firstVer == -1) {
lastIndex = -1;
} else {
ASSERT(0);
}
}
return lastIndex;
}
static SyncIndex raftLogWriteIndex(struct SSyncLogStore* pLogStore) {
SSyncLogStoreData* pData = pLogStore->data;
SWal* pWal = pData->pWal;
SyncIndex lastVer = walGetLastVer(pWal);
return lastVer + 1;
}
static SyncTerm raftLogLastTerm(struct SSyncLogStore* pLogStore) {
SyncTerm lastTerm = 0;
if (raftLogEntryCount(pLogStore) == 0) {
lastTerm = 0;
} else {
SSyncRaftEntry* pLastEntry;
int32_t code = raftLogGetLastEntry(pLogStore, &pLastEntry);
ASSERT(code == 0);
if (pLastEntry != NULL) {
lastTerm = pLastEntry->term;
taosMemoryFree(pLastEntry);
}
}
return lastTerm;
}
static int32_t raftLogAppendEntry(struct SSyncLogStore* pLogStore, SSyncRaftEntry* pEntry) {
SSyncLogStoreData* pData = pLogStore->data;
SWal* pWal = pData->pWal;
SyncIndex writeIndex = raftLogWriteIndex(pLogStore);
ASSERT(pEntry->index == writeIndex);
int code = 0;
SSyncLogMeta syncMeta;
syncMeta.isWeek = pEntry->isWeak;
syncMeta.seqNum = pEntry->seqNum;
syncMeta.term = pEntry->term;
code = walWriteWithSyncInfo(pWal, pEntry->index, pEntry->originalRpcType, syncMeta, pEntry->data, pEntry->dataLen);
if (code != 0) {
int32_t err = terrno;
const char* errStr = tstrerror(err);
int32_t linuxErr = errno;
const char* linuxErrMsg = strerror(errno);
sError("raftLogAppendEntry error, err:%d %X, msg:%s, linuxErr:%d, linuxErrMsg:%s", err, err, errStr, linuxErr,
linuxErrMsg);
ASSERT(0);
}
walFsync(pWal, true);
sTrace("sync event write index:%" PRId64, pEntry->index);
return code;
}
static int32_t raftLogGetEntry(struct SSyncLogStore* pLogStore, SyncIndex index, SSyncRaftEntry** ppEntry) {
SSyncLogStoreData* pData = pLogStore->data;
SWal* pWal = pData->pWal;
int32_t code;
*ppEntry = NULL;
if (raftLogInRange(pLogStore, index)) {
SWalReadHandle* pWalHandle = walOpenReadHandle(pWal);
ASSERT(pWalHandle != NULL);
code = walReadWithHandle(pWalHandle, index);
if (code != 0) {
int32_t err = terrno;
const char* errStr = tstrerror(err);
int32_t linuxErr = errno;
const char* linuxErrMsg = strerror(errno);
sError("raftLogGetEntry error, err:%d %X, msg:%s, linuxErr:%d, linuxErrMsg:%s", err, err, errStr, linuxErr,
linuxErrMsg);
ASSERT(0);
walCloseReadHandle(pWalHandle);
return code;
}
*ppEntry = syncEntryBuild(pWalHandle->pHead->head.bodyLen);
ASSERT(*ppEntry != NULL);
(*ppEntry)->msgType = TDMT_SYNC_CLIENT_REQUEST;
(*ppEntry)->originalRpcType = pWalHandle->pHead->head.msgType;
(*ppEntry)->seqNum = pWalHandle->pHead->head.syncMeta.seqNum;
(*ppEntry)->isWeak = pWalHandle->pHead->head.syncMeta.isWeek;
(*ppEntry)->term = pWalHandle->pHead->head.syncMeta.term;
(*ppEntry)->index = index;
ASSERT((*ppEntry)->dataLen == pWalHandle->pHead->head.bodyLen);
memcpy((*ppEntry)->data, pWalHandle->pHead->head.body, pWalHandle->pHead->head.bodyLen);
// need to hold, do not new every time!!
walCloseReadHandle(pWalHandle);
} else {
// index not in range
code = 0;
}
return code;
}
static int32_t raftLogTruncate(struct SSyncLogStore* pLogStore, SyncIndex fromIndex) {
SSyncLogStoreData* pData = pLogStore->data;
SWal* pWal = pData->pWal;
int32_t code = walRollback(pWal, fromIndex);
if (code != 0) {
int32_t err = terrno;
const char* errStr = tstrerror(err);
int32_t linuxErr = errno;
const char* linuxErrMsg = strerror(errno);
sError("raftLogTruncate error, err:%d %X, msg:%s, linuxErr:%d, linuxErrMsg:%s", err, err, errStr, linuxErr,
linuxErrMsg);
ASSERT(0);
}
return code;
}
static int32_t raftLogGetLastEntry(SSyncLogStore* pLogStore, SSyncRaftEntry** ppLastEntry) {
*ppLastEntry = NULL;
if (raftLogEntryCount(pLogStore) == 0) {
return 0;
}
SyncIndex lastIndex = raftLogLastIndex(pLogStore);
int32_t code = raftLogGetEntry(pLogStore, lastIndex, ppLastEntry);
return code;
}
//-------------------------------
SSyncLogStore* logStoreCreate(SSyncNode* pSyncNode) {
SSyncLogStore* pLogStore = taosMemoryMalloc(sizeof(SSyncLogStore));
assert(pLogStore != NULL);
@ -36,6 +249,16 @@ SSyncLogStore* logStoreCreate(SSyncNode* pSyncNode) {
pData->pSyncNode = pSyncNode;
pData->pWal = pSyncNode->pWal;
SyncIndex firstVer = walGetFirstVer(pData->pWal);
SyncIndex lastVer = walGetLastVer(pData->pWal);
if (firstVer >= 0) {
pData->beginIndex = firstVer;
} else if (firstVer == -1) {
pData->beginIndex = lastVer + 1;
} else {
ASSERT(0);
}
pLogStore->appendEntry = logStoreAppendEntry;
pLogStore->getEntry = logStoreGetEntry;
pLogStore->truncate = logStoreTruncate;
@ -43,6 +266,20 @@ SSyncLogStore* logStoreCreate(SSyncNode* pSyncNode) {
pLogStore->getLastTerm = logStoreLastTerm;
pLogStore->updateCommitIndex = logStoreUpdateCommitIndex;
pLogStore->getCommitIndex = logStoreGetCommitIndex;
pLogStore->syncLogSetBeginIndex = raftLogSetBeginIndex;
pLogStore->syncLogBeginIndex = raftLogBeginIndex;
pLogStore->syncLogEndIndex = raftLogEndIndex;
pLogStore->syncLogIsEmpty = raftLogIsEmpty;
pLogStore->syncLogEntryCount = raftLogEntryCount;
pLogStore->syncLogInRange = raftLogInRange;
pLogStore->syncLogLastIndex = raftLogLastIndex;
pLogStore->syncLogLastTerm = raftLogLastTerm;
pLogStore->syncLogAppendEntry = raftLogAppendEntry;
pLogStore->syncLogGetEntry = raftLogGetEntry;
pLogStore->syncLogTruncate = raftLogTruncate;
pLogStore->syncLogWriteIndex = raftLogWriteIndex;
return pLogStore;
}
@ -53,6 +290,7 @@ void logStoreDestory(SSyncLogStore* pLogStore) {
}
}
//-------------------------------
int32_t logStoreAppendEntry(SSyncLogStore* pLogStore, SSyncRaftEntry* pEntry) {
SSyncLogStoreData* pData = pLogStore->data;
SWal* pWal = pData->pWal;
@ -78,6 +316,8 @@ int32_t logStoreAppendEntry(SSyncLogStore* pLogStore, SSyncRaftEntry* pEntry) {
// assert(code == 0);
walFsync(pWal, true);
sTrace("sync event old write wal: %ld", pEntry->index);
return code;
}
@ -136,7 +376,7 @@ int32_t logStoreTruncate(SSyncLogStore* pLogStore, SyncIndex fromIndex) {
linuxErrMsg);
ASSERT(0);
}
return 0; // to avoid compiler error
return 0;
}
SyncIndex logStoreLastIndex(SSyncLogStore* pLogStore) {
@ -169,7 +409,7 @@ int32_t logStoreUpdateCommitIndex(SSyncLogStore* pLogStore, SyncIndex index) {
sError("walCommit error, err:%d %X, msg:%s, linuxErr:%d, linuxErrMsg:%s", err, err, errStr, linuxErr, linuxErrMsg);
ASSERT(0);
}
return 0; // to avoid compiler error
return 0;
}
SyncIndex logStoreGetCommitIndex(SSyncLogStore* pLogStore) {
@ -199,15 +439,32 @@ cJSON* logStore2Json(SSyncLogStore* pLogStore) {
cJSON_AddStringToObject(pRoot, "pSyncNode", u64buf);
snprintf(u64buf, sizeof(u64buf), "%p", pData->pWal);
cJSON_AddStringToObject(pRoot, "pWal", u64buf);
snprintf(u64buf, sizeof(u64buf), "%ld", logStoreLastIndex(pLogStore));
snprintf(u64buf, sizeof(u64buf), "%ld", pData->beginIndex);
cJSON_AddStringToObject(pRoot, "beginIndex", u64buf);
SyncIndex endIndex = raftLogEndIndex(pLogStore);
snprintf(u64buf, sizeof(u64buf), "%ld", endIndex);
cJSON_AddStringToObject(pRoot, "endIndex", u64buf);
int32_t count = raftLogEntryCount(pLogStore);
cJSON_AddNumberToObject(pRoot, "entryCount", count);
snprintf(u64buf, sizeof(u64buf), "%ld", raftLogWriteIndex(pLogStore));
cJSON_AddStringToObject(pRoot, "WriteIndex", u64buf);
snprintf(u64buf, sizeof(u64buf), "%d", raftLogIsEmpty(pLogStore));
cJSON_AddStringToObject(pRoot, "IsEmpty", u64buf);
snprintf(u64buf, sizeof(u64buf), "%ld", raftLogLastIndex(pLogStore));
cJSON_AddStringToObject(pRoot, "LastIndex", u64buf);
snprintf(u64buf, sizeof(u64buf), "%lu", logStoreLastTerm(pLogStore));
snprintf(u64buf, sizeof(u64buf), "%lu", raftLogLastTerm(pLogStore));
cJSON_AddStringToObject(pRoot, "LastTerm", u64buf);
cJSON* pEntries = cJSON_CreateArray();
cJSON_AddItemToObject(pRoot, "pEntries", pEntries);
SyncIndex lastIndex = logStoreLastIndex(pLogStore);
for (SyncIndex i = 0; i <= lastIndex; ++i) {
for (SyncIndex i = pData->beginIndex; i <= endIndex; ++i) {
SSyncRaftEntry* pEntry = logStoreGetEntry(pLogStore, i);
cJSON_AddItemToArray(pEntries, syncEntry2Json(pEntry));
syncEntryDestory(pEntry);
@ -236,9 +493,26 @@ cJSON* logStoreSimple2Json(SSyncLogStore* pLogStore) {
cJSON_AddStringToObject(pRoot, "pSyncNode", u64buf);
snprintf(u64buf, sizeof(u64buf), "%p", pData->pWal);
cJSON_AddStringToObject(pRoot, "pWal", u64buf);
snprintf(u64buf, sizeof(u64buf), "%ld", logStoreLastIndex(pLogStore));
snprintf(u64buf, sizeof(u64buf), "%ld", pData->beginIndex);
cJSON_AddStringToObject(pRoot, "beginIndex", u64buf);
SyncIndex endIndex = raftLogEndIndex(pLogStore);
snprintf(u64buf, sizeof(u64buf), "%ld", endIndex);
cJSON_AddStringToObject(pRoot, "endIndex", u64buf);
int32_t count = raftLogEntryCount(pLogStore);
cJSON_AddNumberToObject(pRoot, "entryCount", count);
snprintf(u64buf, sizeof(u64buf), "%ld", raftLogWriteIndex(pLogStore));
cJSON_AddStringToObject(pRoot, "WriteIndex", u64buf);
snprintf(u64buf, sizeof(u64buf), "%d", raftLogIsEmpty(pLogStore));
cJSON_AddStringToObject(pRoot, "IsEmpty", u64buf);
snprintf(u64buf, sizeof(u64buf), "%ld", raftLogLastIndex(pLogStore));
cJSON_AddStringToObject(pRoot, "LastIndex", u64buf);
snprintf(u64buf, sizeof(u64buf), "%lu", logStoreLastTerm(pLogStore));
snprintf(u64buf, sizeof(u64buf), "%lu", raftLogLastTerm(pLogStore));
cJSON_AddStringToObject(pRoot, "LastTerm", u64buf);
}
@ -254,6 +528,12 @@ char* logStoreSimple2Str(SSyncLogStore* pLogStore) {
return serialized;
}
SyncIndex logStoreFirstIndex(SSyncLogStore* pLogStore) {
SSyncLogStoreData* pData = pLogStore->data;
SWal* pWal = pData->pWal;
return walGetFirstVer(pWal);
}
// for debug -----------------
void logStorePrint(SSyncLogStore* pLogStore) {
char* serialized = logStore2Str(pLogStore);
@ -303,7 +583,9 @@ void logStoreSimpleLog(SSyncLogStore* pLogStore) {
}
void logStoreSimpleLog2(char* s, SSyncLogStore* pLogStore) {
char* serialized = logStoreSimple2Str(pLogStore);
sTrace("logStoreSimpleLog2 | len:%lu | %s | %s", strlen(serialized), s, serialized);
taosMemoryFree(serialized);
if (gRaftDetailLog) {
char* serialized = logStoreSimple2Str(pLogStore);
sTrace("logStoreSimpleLog2 | len:%lu | %s | %s", strlen(serialized), s, serialized);
taosMemoryFree(serialized);
}
}

83
source/libs/sync/src/syncReplication.c
View File

@ -16,9 +16,11 @@
#include "syncReplication.h"
#include "syncIndexMgr.h"
#include "syncMessage.h"
#include "syncRaftCfg.h"
#include "syncRaftEntry.h"
#include "syncRaftLog.h"
#include "syncRaftStore.h"
#include "syncSnapshot.h"
#include "syncUtil.h"
// TLA+ Spec
@ -59,6 +61,7 @@ int32_t syncNodeAppendEntriesPeers(SSyncNode* pSyncNode) {
// set prevLogIndex
SyncIndex nextIndex = syncIndexMgrGetIndex(pSyncNode->pNextIndex, pDestId);
SyncIndex preLogIndex = nextIndex - 1;
// set preLogTerm
@ -113,9 +116,87 @@ int32_t syncNodeAppendEntriesPeers(SSyncNode* pSyncNode) {
return ret;
}
int32_t syncNodeAppendEntriesPeersSnapshot(SSyncNode* pSyncNode) {
ASSERT(pSyncNode->state == TAOS_SYNC_STATE_LEADER);
syncIndexMgrLog2("begin append entries peers pNextIndex:", pSyncNode->pNextIndex);
syncIndexMgrLog2("begin append entries peers pMatchIndex:", pSyncNode->pMatchIndex);
logStoreSimpleLog2("begin append entries peers LogStore:", pSyncNode->pLogStore);
{
SSnapshot snapshot;
pSyncNode->pFsm->FpGetSnapshot(pSyncNode->pFsm, &snapshot);
sTrace("begin append entries peers, snapshot.lastApplyIndex:%ld, snapshot.lastApplyTerm:%lu",
snapshot.lastApplyIndex, snapshot.lastApplyTerm);
}
int32_t ret = 0;
for (int i = 0; i < pSyncNode->peersNum; ++i) {
SRaftId* pDestId = &(pSyncNode->peersId[i]);
// next index
SyncIndex nextIndex = syncIndexMgrGetIndex(pSyncNode->pNextIndex, pDestId);
// pre index, pre term
SyncIndex preLogIndex = syncNodeGetPreIndex(pSyncNode, nextIndex);
SyncTerm preLogTerm = syncNodeGetPreTerm(pSyncNode, nextIndex);
// batch optimized
// SyncIndex lastIndex = syncUtilMinIndex(pSyncNode->pLogStore->getLastIndex(pSyncNode->pLogStore), nextIndex);
// prepare entry
SyncAppendEntries* pMsg = NULL;
SSyncRaftEntry* pEntry;
int32_t code = pSyncNode->pLogStore->syncLogGetEntry(pSyncNode->pLogStore, nextIndex, &pEntry);
ASSERT(code == 0);
if (pEntry != NULL) {
pMsg = syncAppendEntriesBuild(pEntry->bytes, pSyncNode->vgId);
ASSERT(pMsg != NULL);
// add pEntry into msg
uint32_t len;
char* serialized = syncEntrySerialize(pEntry, &len);
assert(len == pEntry->bytes);
memcpy(pMsg->data, serialized, len);
taosMemoryFree(serialized);
syncEntryDestory(pEntry);
} else {
// no entry in log
pMsg = syncAppendEntriesBuild(0, pSyncNode->vgId);
ASSERT(pMsg != NULL);
}
// prepare msg
ASSERT(pMsg != NULL);
pMsg->srcId = pSyncNode->myRaftId;
pMsg->destId = *pDestId;
pMsg->term = pSyncNode->pRaftStore->currentTerm;
pMsg->prevLogIndex = preLogIndex;
pMsg->prevLogTerm = preLogTerm;
pMsg->commitIndex = pSyncNode->commitIndex;
pMsg->privateTerm = 0;
// pMsg->privateTerm = syncIndexMgrGetTerm(pSyncNode->pNextIndex, pDestId);
// send msg
syncNodeAppendEntries(pSyncNode, pDestId, pMsg);
syncAppendEntriesDestroy(pMsg);
}
return ret;
}
int32_t syncNodeReplicate(SSyncNode* pSyncNode) {
// start replicate
int32_t ret = syncNodeAppendEntriesPeers(pSyncNode);
int32_t ret = 0;
if (pSyncNode->pRaftCfg->snapshotEnable) {
ret = syncNodeAppendEntriesPeersSnapshot(pSyncNode);
} else {
ret = syncNodeAppendEntriesPeers(pSyncNode);
}
return ret;
}

65
source/libs/sync/src/syncRequestVote.c
View File

@ -15,6 +15,7 @@
#include "syncRequestVote.h"
#include "syncInt.h"
#include "syncRaftCfg.h"
#include "syncRaftStore.h"
#include "syncUtil.h"
#include "syncVoteMgr.h"
@ -62,6 +63,9 @@ int32_t syncNodeOnRequestVoteCb(SSyncNode* ths, SyncRequestVote* pMsg) {
// maybe has already voted for pMsg->srcId
// vote again, no harm
raftStoreVote(ths->pRaftStore, &(pMsg->srcId));
// forbid elect for this round
syncNodeResetElectTimer(ths);
}
SyncRequestVoteReply* pReply = syncRequestVoteReplyBuild(ths->vgId);
@ -77,3 +81,64 @@ int32_t syncNodeOnRequestVoteCb(SSyncNode* ths, SyncRequestVote* pMsg) {
return ret;
}
static bool syncNodeOnRequestVoteLogOK(SSyncNode* pSyncNode, SyncRequestVote* pMsg) {
SyncTerm myLastTerm = syncNodeGetLastTerm(pSyncNode);
SyncIndex myLastIndex = syncNodeGetLastIndex(pSyncNode);
if (pMsg->lastLogTerm > myLastTerm) {
return true;
}
if (pMsg->lastLogTerm == myLastTerm && pMsg->lastLogIndex >= myLastIndex) {
return true;
}
return false;
}
int32_t syncNodeOnRequestVoteSnapshotCb(SSyncNode* ths, SyncRequestVote* pMsg) {
int32_t ret = 0;
// print log
char logBuf[128] = {0};
snprintf(logBuf, sizeof(logBuf), "recv SyncRequestVote, currentTerm:%lu", ths->pRaftStore->currentTerm);
syncRequestVoteLog2(logBuf, pMsg);
// if already drop replica, do not process
if (!syncNodeInRaftGroup(ths, &(pMsg->srcId)) && !ths->pRaftCfg->isStandBy) {
sInfo("recv SyncRequestVote maybe replica already dropped");
return ret;
}
// maybe update term
if (pMsg->term > ths->pRaftStore->currentTerm) {
syncNodeUpdateTerm(ths, pMsg->term);
}
ASSERT(pMsg->term <= ths->pRaftStore->currentTerm);
bool logOK = syncNodeOnRequestVoteLogOK(ths, pMsg);
bool grant = (pMsg->term == ths->pRaftStore->currentTerm) && logOK &&
((!raftStoreHasVoted(ths->pRaftStore)) || (syncUtilSameId(&(ths->pRaftStore->voteFor), &(pMsg->srcId))));
if (grant) {
// maybe has already voted for pMsg->srcId
// vote again, no harm
raftStoreVote(ths->pRaftStore, &(pMsg->srcId));
// forbid elect for this round
syncNodeResetElectTimer(ths);
}
// send msg
SyncRequestVoteReply* pReply = syncRequestVoteReplyBuild(ths->vgId);
pReply->srcId = ths->myRaftId;
pReply->destId = pMsg->srcId;
pReply->term = ths->pRaftStore->currentTerm;
pReply->voteGranted = grant;
SRpcMsg rpcMsg;
syncRequestVoteReply2RpcMsg(pReply, &rpcMsg);
syncNodeSendMsgById(&pReply->destId, ths, &rpcMsg);
syncRequestVoteReplyDestroy(pReply);
return ret;
}

66
source/libs/sync/src/syncRequestVoteReply.c
View File

@ -15,6 +15,7 @@
#include "syncRequestVoteReply.h"
#include "syncInt.h"
#include "syncRaftCfg.h"
#include "syncRaftStore.h"
#include "syncUtil.h"
#include "syncVoteMgr.h"
@ -92,3 +93,68 @@ int32_t syncNodeOnRequestVoteReplyCb(SSyncNode* ths, SyncRequestVoteReply* pMsg)
return ret;
}
int32_t syncNodeOnRequestVoteReplySnapshotCb(SSyncNode* ths, SyncRequestVoteReply* pMsg) {
int32_t ret = 0;
// print log
char logBuf[128] = {0};
snprintf(logBuf, sizeof(logBuf), "recv SyncRequestVoteReply, term:%lu", ths->pRaftStore->currentTerm);
syncRequestVoteReplyLog2(logBuf, pMsg);
// if already drop replica, do not process
if (!syncNodeInRaftGroup(ths, &(pMsg->srcId)) && !ths->pRaftCfg->isStandBy) {
sInfo("recv SyncRequestVoteReply, maybe replica already dropped");
return ret;
}
// drop stale response
if (pMsg->term < ths->pRaftStore->currentTerm) {
sTrace("recv SyncRequestVoteReply, drop stale response, receive_term:%lu current_term:%lu", pMsg->term,
ths->pRaftStore->currentTerm);
return ret;
}
// assert(!(pMsg->term > ths->pRaftStore->currentTerm));
// no need this code, because if I receive reply.term, then I must have sent for that term.
// if (pMsg->term > ths->pRaftStore->currentTerm) {
// syncNodeUpdateTerm(ths, pMsg->term);
// }
if (pMsg->term > ths->pRaftStore->currentTerm) {
char logBuf[128] = {0};
snprintf(logBuf, sizeof(logBuf), "recv SyncRequestVoteReply, error term, receive_term:%lu current_term:%lu",
pMsg->term, ths->pRaftStore->currentTerm);
syncNodePrint2(logBuf, ths);
sError("%s", logBuf);
return ret;
}
ASSERT(pMsg->term == ths->pRaftStore->currentTerm);
// This tallies votes even when the current state is not Candidate,
// but they won't be looked at, so it doesn't matter.
if (ths->state == TAOS_SYNC_STATE_CANDIDATE) {
votesRespondAdd(ths->pVotesRespond, pMsg);
if (pMsg->voteGranted) {
// add vote
voteGrantedVote(ths->pVotesGranted, pMsg);
// maybe to leader
if (voteGrantedMajority(ths->pVotesGranted)) {
if (!ths->pVotesGranted->toLeader) {
syncNodeCandidate2Leader(ths);
// prevent to leader again!
ths->pVotesGranted->toLeader = true;
}
}
} else {
;
// do nothing
// UNCHANGED <<votesGranted, voterLog>>
}
}
return ret;
}

595
source/libs/sync/src/syncSnapshot.c
View File

@ -14,23 +14,598 @@
*/
#include "syncSnapshot.h"
#include "syncIndexMgr.h"
#include "syncRaftLog.h"
#include "syncRaftStore.h"
#include "syncUtil.h"
#include "wal.h"
SSyncSnapshotSender *snapshotSenderCreate(SSyncNode *pSyncNode) { return NULL; }
static void snapshotReceiverDoStart(SSyncSnapshotReceiver *pReceiver, SyncTerm privateTerm);
void snapshotSenderDestroy(SSyncSnapshotSender *pSender) {}
//----------------------------------
SSyncSnapshotSender *snapshotSenderCreate(SSyncNode *pSyncNode, int32_t replicaIndex) {
bool condition = (pSyncNode->pFsm->FpSnapshotStartRead != NULL) && (pSyncNode->pFsm->FpSnapshotStopRead != NULL) &&
(pSyncNode->pFsm->FpSnapshotDoRead != NULL);
int32_t snapshotSend(SSyncSnapshotSender *pSender) { return 0; }
SSyncSnapshotSender *pSender = NULL;
if (condition) {
pSender = taosMemoryMalloc(sizeof(SSyncSnapshotSender));
ASSERT(pSender != NULL);
memset(pSender, 0, sizeof(*pSender));
cJSON *snapshotSender2Json(SSyncSnapshotSender *pSender) { return NULL; }
pSender->start = false;
pSender->seq = SYNC_SNAPSHOT_SEQ_INVALID;
pSender->ack = SYNC_SNAPSHOT_SEQ_INVALID;
pSender->pReader = NULL;
pSender->pCurrentBlock = NULL;
pSender->blockLen = 0;
pSender->sendingMS = SYNC_SNAPSHOT_RETRY_MS;
pSender->pSyncNode = pSyncNode;
pSender->replicaIndex = replicaIndex;
pSender->term = pSyncNode->pRaftStore->currentTerm;
pSender->privateTerm = taosGetTimestampMs() + 100;
pSender->pSyncNode->pFsm->FpGetSnapshot(pSender->pSyncNode->pFsm, &(pSender->snapshot));
pSender->finish = false;
} else {
sError("snapshotSenderCreate cannot create sender");
}
return pSender;
}
char *snapshotSender2Str(SSyncSnapshotSender *pSender) { return NULL; }
void snapshotSenderDestroy(SSyncSnapshotSender *pSender) {
if (pSender != NULL) {
if (pSender->pCurrentBlock != NULL) {
taosMemoryFree(pSender->pCurrentBlock);
}
taosMemoryFree(pSender);
}
}
SSyncSnapshotReceiver *snapshotReceiverCreate(SSyncNode *pSyncNode) { return NULL; }
bool snapshotSenderIsStart(SSyncSnapshotSender *pSender) { return pSender->start; }
void snapshotReceiverDestroy(SSyncSnapshotReceiver *pReceiver) {}
// begin send snapshot (current term, seq begin)
void snapshotSenderStart(SSyncSnapshotSender *pSender) {
ASSERT(!snapshotSenderIsStart(pSender));
int32_t snapshotReceive(SSyncSnapshotReceiver *pReceiver) { return 0; }
pSender->seq = SYNC_SNAPSHOT_SEQ_BEGIN;
pSender->ack = SYNC_SNAPSHOT_SEQ_INVALID;
cJSON *snapshotReceiver2Json(SSyncSnapshotReceiver *pReceiver) { return NULL; }
// open snapshot reader
ASSERT(pSender->pReader == NULL);
int32_t ret = pSender->pSyncNode->pFsm->FpSnapshotStartRead(pSender->pSyncNode->pFsm, &(pSender->pReader));
ASSERT(ret == 0);
char *snapshotReceiver2Str(SSyncSnapshotReceiver *pReceiver) { return NULL; }
if (pSender->pCurrentBlock != NULL) {
taosMemoryFree(pSender->pCurrentBlock);
}
pSender->blockLen = 0;
// get current snapshot info
pSender->pSyncNode->pFsm->FpGetSnapshot(pSender->pSyncNode->pFsm, &(pSender->snapshot));
pSender->sendingMS = SYNC_SNAPSHOT_RETRY_MS;
pSender->term = pSender->pSyncNode->pRaftStore->currentTerm;
++(pSender->privateTerm);
pSender->finish = false;
pSender->start = true;
// build begin msg
SyncSnapshotSend *pMsg = syncSnapshotSendBuild(0, pSender->pSyncNode->vgId);
pMsg->srcId = pSender->pSyncNode->myRaftId;
pMsg->destId = (pSender->pSyncNode->replicasId)[pSender->replicaIndex];
pMsg->term = pSender->pSyncNode->pRaftStore->currentTerm;
pMsg->lastIndex = pSender->snapshot.lastApplyIndex;
pMsg->lastTerm = pSender->snapshot.lastApplyTerm;
pMsg->seq = pSender->seq; // SYNC_SNAPSHOT_SEQ_BEGIN
pMsg->privateTerm = pSender->privateTerm;
// send msg
SRpcMsg rpcMsg;
syncSnapshotSend2RpcMsg(pMsg, &rpcMsg);
syncNodeSendMsgById(&(pMsg->destId), pSender->pSyncNode, &rpcMsg);
char *msgStr = syncSnapshotSend2Str(pMsg);
char host[128];
uint16_t port;
syncUtilU642Addr(pSender->pSyncNode->replicasId[pSender->replicaIndex].addr, host, sizeof(host), &port);
sTrace("sync event snapshot send to %s:%d begin seq:%d ack:%d lastApplyIndex:%ld lastApplyTerm:%lu send msg:%s", host,
port, pSender->seq, pSender->ack, pSender->snapshot.lastApplyIndex, pSender->snapshot.lastApplyTerm, msgStr);
taosMemoryFree(msgStr);
syncSnapshotSendDestroy(pMsg);
}
#if 0
// when entry in snapshot, start sender
void snapshotSenderStart(SSyncSnapshotSender *pSender) {
if (!(pSender->start)) {
// start
snapshotSenderDoStart(pSender);
pSender->start = true;
} else {
// already start
ASSERT(pSender->pSyncNode->pRaftStore->currentTerm >= pSender->term);
// if current term is higher, need start again
if (pSender->pSyncNode->pRaftStore->currentTerm > pSender->term) {
// force peer rollback
SyncSnapshotSend *pMsg = syncSnapshotSendBuild(0, pSender->pSyncNode->vgId);
pMsg->srcId = pSender->pSyncNode->myRaftId;
pMsg->destId = (pSender->pSyncNode->replicasId)[pSender->replicaIndex];
pMsg->term = pSender->pSyncNode->pRaftStore->currentTerm;
pMsg->lastIndex = pSender->snapshot.lastApplyIndex;
pMsg->lastTerm = pSender->snapshot.lastApplyTerm;
pMsg->seq = SYNC_SNAPSHOT_SEQ_FORCE_CLOSE;
SRpcMsg rpcMsg;
syncSnapshotSend2RpcMsg(pMsg, &rpcMsg);
syncNodeSendMsgById(&(pMsg->destId), pSender->pSyncNode, &rpcMsg);
char *msgStr = syncSnapshotSend2Str(pMsg);
sTrace("snapshot send force close seq:%d ack:%d send msg:%s", pSender->seq, pSender->ack, msgStr);
taosMemoryFree(msgStr);
syncSnapshotSendDestroy(pMsg);
// close reader
int32_t ret = pSender->pSyncNode->pFsm->FpSnapshotStopRead(pSender->pSyncNode->pFsm, pSender->pReader);
ASSERT(ret == 0);
pSender->pReader = NULL;
// start again
snapshotSenderDoStart(pSender);
pSender->start = true;
} else {
// current term, do nothing
ASSERT(pSender->pSyncNode->pRaftStore->currentTerm == pSender->term);
}
}
char *s = snapshotSender2Str(pSender);
sInfo("snapshotSenderStart %s", s);
taosMemoryFree(s);
}
#endif
void snapshotSenderStop(SSyncSnapshotSender *pSender) {
if (pSender->pReader != NULL) {
int32_t ret = pSender->pSyncNode->pFsm->FpSnapshotStopRead(pSender->pSyncNode->pFsm, pSender->pReader);
ASSERT(ret == 0);
pSender->pReader = NULL;
}
if (pSender->pCurrentBlock != NULL) {
taosMemoryFree(pSender->pCurrentBlock);
pSender->pCurrentBlock = NULL;
pSender->blockLen = 0;
}
pSender->start = false;
char *s = snapshotSender2Str(pSender);
sInfo("snapshotSenderStop %s", s);
taosMemoryFree(s);
}
// when sender receiver ack, call this function to send msg from seq
// seq = ack + 1, already updated
int32_t snapshotSend(SSyncSnapshotSender *pSender) {
// free memory last time (seq - 1)
if (pSender->pCurrentBlock != NULL) {
taosMemoryFree(pSender->pCurrentBlock);
pSender->pCurrentBlock = NULL;
pSender->blockLen = 0;
}
// read data
int32_t ret = pSender->pSyncNode->pFsm->FpSnapshotDoRead(pSender->pSyncNode->pFsm, pSender->pReader,
&(pSender->pCurrentBlock), &(pSender->blockLen));
ASSERT(ret == 0);
if (pSender->blockLen > 0) {
// has read data
} else {
// read finish
pSender->seq = SYNC_SNAPSHOT_SEQ_END;
}
// build msg
SyncSnapshotSend *pMsg = syncSnapshotSendBuild(pSender->blockLen, pSender->pSyncNode->vgId);
pMsg->srcId = pSender->pSyncNode->myRaftId;
pMsg->destId = (pSender->pSyncNode->replicasId)[pSender->replicaIndex];
pMsg->term = pSender->pSyncNode->pRaftStore->currentTerm;
pMsg->lastIndex = pSender->snapshot.lastApplyIndex;
pMsg->lastTerm = pSender->snapshot.lastApplyTerm;
pMsg->seq = pSender->seq;
pMsg->privateTerm = pSender->privateTerm;
memcpy(pMsg->data, pSender->pCurrentBlock, pSender->blockLen);
// send msg
SRpcMsg rpcMsg;
syncSnapshotSend2RpcMsg(pMsg, &rpcMsg);
syncNodeSendMsgById(&(pMsg->destId), pSender->pSyncNode, &rpcMsg);
char *msgStr = syncSnapshotSend2Str(pMsg);
char host[128];
uint16_t port;
syncUtilU642Addr(pSender->pSyncNode->replicasId[pSender->replicaIndex].addr, host, sizeof(host), &port);
if (pSender->seq == SYNC_SNAPSHOT_SEQ_END) {
sTrace("sync event snapshot send to %s:%d finish seq:%d ack:%d lastApplyIndex:%ld lastApplyTerm:%lu send msg:%s",
host, port, pSender->seq, pSender->ack, pSender->snapshot.lastApplyIndex, pSender->snapshot.lastApplyTerm,
msgStr);
} else {
sTrace("sync event snapshot send to %s:%d sending seq:%d ack:%d lastApplyIndex:%ld lastApplyTerm:%lu send msg:%s",
host, port, pSender->seq, pSender->ack, pSender->snapshot.lastApplyIndex, pSender->snapshot.lastApplyTerm,
msgStr);
}
taosMemoryFree(msgStr);
syncSnapshotSendDestroy(pMsg);
return 0;
}
// send snapshot data from cache
int32_t snapshotReSend(SSyncSnapshotSender *pSender) {
if (pSender->pCurrentBlock != NULL) {
SyncSnapshotSend *pMsg = syncSnapshotSendBuild(pSender->blockLen, pSender->pSyncNode->vgId);
pMsg->srcId = pSender->pSyncNode->myRaftId;
pMsg->destId = (pSender->pSyncNode->replicasId)[pSender->replicaIndex];
pMsg->term = pSender->pSyncNode->pRaftStore->currentTerm;
pMsg->lastIndex = pSender->snapshot.lastApplyIndex;
pMsg->lastTerm = pSender->snapshot.lastApplyTerm;
pMsg->seq = pSender->seq;
memcpy(pMsg->data, pSender->pCurrentBlock, pSender->blockLen);
SRpcMsg rpcMsg;
syncSnapshotSend2RpcMsg(pMsg, &rpcMsg);
syncNodeSendMsgById(&(pMsg->destId), pSender->pSyncNode, &rpcMsg);
char *msgStr = syncSnapshotSend2Str(pMsg);
char host[128];
uint16_t port;
syncUtilU642Addr(pSender->pSyncNode->replicasId[pSender->replicaIndex].addr, host, sizeof(host), &port);
sTrace("sync event snapshot send to %s:%d resend seq:%d ack:%d send msg:%s", host, port, pSender->seq, pSender->ack,
msgStr);
taosMemoryFree(msgStr);
syncSnapshotSendDestroy(pMsg);
}
return 0;
}
cJSON *snapshotSender2Json(SSyncSnapshotSender *pSender) {
char u64buf[128];
cJSON *pRoot = cJSON_CreateObject();
if (pSender != NULL) {
cJSON_AddNumberToObject(pRoot, "start", pSender->start);
cJSON_AddNumberToObject(pRoot, "seq", pSender->seq);
cJSON_AddNumberToObject(pRoot, "ack", pSender->ack);
snprintf(u64buf, sizeof(u64buf), "%p", pSender->pReader);
cJSON_AddStringToObject(pRoot, "pReader", u64buf);
snprintf(u64buf, sizeof(u64buf), "%p", pSender->pCurrentBlock);
cJSON_AddStringToObject(pRoot, "pCurrentBlock", u64buf);
cJSON_AddNumberToObject(pRoot, "blockLen", pSender->blockLen);
if (pSender->pCurrentBlock != NULL) {
char *s;
s = syncUtilprintBin((char *)(pSender->pCurrentBlock), pSender->blockLen);
cJSON_AddStringToObject(pRoot, "pCurrentBlock", s);
taosMemoryFree(s);
s = syncUtilprintBin2((char *)(pSender->pCurrentBlock), pSender->blockLen);
cJSON_AddStringToObject(pRoot, "pCurrentBlock2", s);
taosMemoryFree(s);
}
cJSON *pSnapshot = cJSON_CreateObject();
snprintf(u64buf, sizeof(u64buf), "%lu", pSender->snapshot.lastApplyIndex);
cJSON_AddStringToObject(pSnapshot, "lastApplyIndex", u64buf);
snprintf(u64buf, sizeof(u64buf), "%lu", pSender->snapshot.lastApplyTerm);
cJSON_AddStringToObject(pSnapshot, "lastApplyTerm", u64buf);
cJSON_AddItemToObject(pRoot, "snapshot", pSnapshot);
snprintf(u64buf, sizeof(u64buf), "%lu", pSender->sendingMS);
cJSON_AddStringToObject(pRoot, "sendingMS", u64buf);
snprintf(u64buf, sizeof(u64buf), "%p", pSender->pSyncNode);
cJSON_AddStringToObject(pRoot, "pSyncNode", u64buf);
cJSON_AddNumberToObject(pRoot, "replicaIndex", pSender->replicaIndex);
snprintf(u64buf, sizeof(u64buf), "%lu", pSender->term);
cJSON_AddStringToObject(pRoot, "term", u64buf);
snprintf(u64buf, sizeof(u64buf), "%lu", pSender->privateTerm);
cJSON_AddStringToObject(pRoot, "privateTerm", u64buf);
cJSON_AddNumberToObject(pRoot, "finish", pSender->finish);
}
cJSON *pJson = cJSON_CreateObject();
cJSON_AddItemToObject(pJson, "SSyncSnapshotSender", pRoot);
return pJson;
}
char *snapshotSender2Str(SSyncSnapshotSender *pSender) {
cJSON *pJson = snapshotSender2Json(pSender);
char *serialized = cJSON_Print(pJson);
cJSON_Delete(pJson);
return serialized;
}
// -------------------------------------
SSyncSnapshotReceiver *snapshotReceiverCreate(SSyncNode *pSyncNode, int32_t replicaIndex) {
bool condition = (pSyncNode->pFsm->FpSnapshotStartWrite != NULL) && (pSyncNode->pFsm->FpSnapshotStopWrite != NULL) &&
(pSyncNode->pFsm->FpSnapshotDoWrite != NULL);
SSyncSnapshotReceiver *pReceiver = NULL;
if (condition) {
pReceiver = taosMemoryMalloc(sizeof(SSyncSnapshotReceiver));
ASSERT(pReceiver != NULL);
memset(pReceiver, 0, sizeof(*pReceiver));
pReceiver->start = false;
pReceiver->ack = SYNC_SNAPSHOT_SEQ_BEGIN;
pReceiver->pWriter = NULL;
pReceiver->pSyncNode = pSyncNode;
pReceiver->replicaIndex = replicaIndex;
pReceiver->term = pSyncNode->pRaftStore->currentTerm;
pReceiver->privateTerm = 0;
} else {
sInfo("snapshotReceiverCreate cannot create receiver");
}
return pReceiver;
}
void snapshotReceiverDestroy(SSyncSnapshotReceiver *pReceiver) {
if (pReceiver != NULL) {
taosMemoryFree(pReceiver);
}
}
bool snapshotReceiverIsStart(SSyncSnapshotReceiver *pReceiver) { return pReceiver->start; }
// begin receive snapshot msg (current term, seq begin)
static void snapshotReceiverDoStart(SSyncSnapshotReceiver *pReceiver, SyncTerm privateTerm) {
pReceiver->term = pReceiver->pSyncNode->pRaftStore->currentTerm;
pReceiver->privateTerm = privateTerm;
pReceiver->ack = SYNC_SNAPSHOT_SEQ_BEGIN;
ASSERT(pReceiver->pWriter == NULL);
int32_t ret = pReceiver->pSyncNode->pFsm->FpSnapshotStartWrite(pReceiver->pSyncNode->pFsm, &(pReceiver->pWriter));
ASSERT(ret == 0);
}
// if receiver receive msg from seq = SYNC_SNAPSHOT_SEQ_BEGIN, start receiver
// if already start, force close, start again
void snapshotReceiverStart(SSyncSnapshotReceiver *pReceiver, SyncTerm privateTerm) {
if (!snapshotReceiverIsStart(pReceiver)) {
// start
snapshotReceiverDoStart(pReceiver, privateTerm);
pReceiver->start = true;
} else {
// already start
// force close, abandon incomplete data
int32_t ret =
pReceiver->pSyncNode->pFsm->FpSnapshotStopWrite(pReceiver->pSyncNode->pFsm, pReceiver->pWriter, false);
ASSERT(ret == 0);
pReceiver->pWriter = NULL;
// start again
snapshotReceiverDoStart(pReceiver, privateTerm);
pReceiver->start = true;
ASSERT(0);
}
char *s = snapshotReceiver2Str(pReceiver);
sInfo("snapshotReceiverStart %s", s);
taosMemoryFree(s);
}
void snapshotReceiverStop(SSyncSnapshotReceiver *pReceiver, bool apply) {
if (pReceiver->pWriter != NULL) {
int32_t ret =
pReceiver->pSyncNode->pFsm->FpSnapshotStopWrite(pReceiver->pSyncNode->pFsm, pReceiver->pWriter, false);
ASSERT(ret == 0);
pReceiver->pWriter = NULL;
}
pReceiver->start = false;
if (apply) {
++(pReceiver->privateTerm);
}
char *s = snapshotReceiver2Str(pReceiver);
sInfo("snapshotReceiverStop %s", s);
taosMemoryFree(s);
}
cJSON *snapshotReceiver2Json(SSyncSnapshotReceiver *pReceiver) {
char u64buf[128];
cJSON *pRoot = cJSON_CreateObject();
if (pReceiver != NULL) {
cJSON_AddNumberToObject(pRoot, "start", pReceiver->start);
cJSON_AddNumberToObject(pRoot, "ack", pReceiver->ack);
snprintf(u64buf, sizeof(u64buf), "%p", pReceiver->pWriter);
cJSON_AddStringToObject(pRoot, "pWriter", u64buf);
snprintf(u64buf, sizeof(u64buf), "%p", pReceiver->pSyncNode);
cJSON_AddStringToObject(pRoot, "pSyncNode", u64buf);
cJSON_AddNumberToObject(pRoot, "replicaIndex", pReceiver->replicaIndex);
snprintf(u64buf, sizeof(u64buf), "%lu", pReceiver->term);
cJSON_AddStringToObject(pRoot, "term", u64buf);
snprintf(u64buf, sizeof(u64buf), "%lu", pReceiver->privateTerm);
cJSON_AddStringToObject(pRoot, "privateTerm", u64buf);
}
cJSON *pJson = cJSON_CreateObject();
cJSON_AddItemToObject(pJson, "SSyncSnapshotReceiver", pRoot);
return pJson;
}
char *snapshotReceiver2Str(SSyncSnapshotReceiver *pReceiver) {
cJSON *pJson = snapshotReceiver2Json(pReceiver);
char *serialized = cJSON_Print(pJson);
cJSON_Delete(pJson);
return serialized;
}
// receiver do something
int32_t syncNodeOnSnapshotSendCb(SSyncNode *pSyncNode, SyncSnapshotSend *pMsg) {
// get receiver
SSyncSnapshotReceiver *pReceiver = pSyncNode->pNewNodeReceiver;
bool needRsp = false;
int32_t writeCode = 0;
// state, term, seq/ack
if (pSyncNode->state == TAOS_SYNC_STATE_FOLLOWER) {
if (pMsg->term == pSyncNode->pRaftStore->currentTerm) {
if (pMsg->seq == SYNC_SNAPSHOT_SEQ_BEGIN) {
// begin
snapshotReceiverStart(pReceiver, pMsg->privateTerm);
pReceiver->ack = pMsg->seq;
needRsp = true;
char *msgStr = syncSnapshotSend2Str(pMsg);
char host[128];
uint16_t port;
syncUtilU642Addr(pMsg->srcId.addr, host, sizeof(host), &port);
sTrace("sync event snapshot recv from %s:%d begin ack:%d, lastIndex:%ld, lastTerm:%lu, recv msg:%s", host, port,
pReceiver->ack, pMsg->lastIndex, pMsg->lastTerm, msgStr);
taosMemoryFree(msgStr);
} else if (pMsg->seq == SYNC_SNAPSHOT_SEQ_END) {
// end, finish FSM
writeCode = pSyncNode->pFsm->FpSnapshotDoWrite(pSyncNode->pFsm, pReceiver->pWriter, pMsg->data, pMsg->dataLen);
ASSERT(writeCode == 0);
pSyncNode->pFsm->FpSnapshotStopWrite(pSyncNode->pFsm, pReceiver->pWriter, true);
pSyncNode->pLogStore->syncLogSetBeginIndex(pSyncNode->pLogStore, pMsg->lastIndex + 1);
char *logSimpleStr = logStoreSimple2Str(pSyncNode->pLogStore);
SSnapshot snapshot;
pSyncNode->pFsm->FpGetSnapshot(pSyncNode->pFsm, &snapshot);
char host[128];
uint16_t port;
syncUtilU642Addr(pMsg->srcId.addr, host, sizeof(host), &port);
sInfo(
"sync event snapshot recv from %s:%d finish, update log begin index:%ld, snapshot.lastApplyIndex:%ld, "
"snapshot.lastApplyTerm:%lu, raft log:%s",
host, port, pMsg->lastIndex + 1, snapshot.lastApplyIndex, snapshot.lastApplyTerm, logSimpleStr);
taosMemoryFree(logSimpleStr);
pReceiver->pWriter = NULL;
snapshotReceiverStop(pReceiver, true);
pReceiver->ack = pMsg->seq;
needRsp = true;
char *msgStr = syncSnapshotSend2Str(pMsg);
sTrace("sync event snapshot recv from %s:%d end ack:%d, lastIndex:%ld, lastTerm:%lu, recv msg:%s", host, port,
pReceiver->ack, pMsg->lastIndex, pMsg->lastTerm, msgStr);
taosMemoryFree(msgStr);
} else if (pMsg->seq == SYNC_SNAPSHOT_SEQ_FORCE_CLOSE) {
pSyncNode->pFsm->FpSnapshotStopWrite(pSyncNode->pFsm, pReceiver->pWriter, false);
snapshotReceiverStop(pReceiver, false);
needRsp = false;
char host[128];
uint16_t port;
syncUtilU642Addr(pMsg->srcId.addr, host, sizeof(host), &port);
char *msgStr = syncSnapshotSend2Str(pMsg);
sTrace("sync event snapshot recv from %s:%d force close ack:%d, lastIndex:%ld, lastTerm:%lu, recv msg:%s", host,
port, pReceiver->ack, pMsg->lastIndex, pMsg->lastTerm, msgStr);
taosMemoryFree(msgStr);
} else if (pMsg->seq > SYNC_SNAPSHOT_SEQ_BEGIN && pMsg->seq < SYNC_SNAPSHOT_SEQ_END) {
// transfering
if (pMsg->seq == pReceiver->ack + 1) {
writeCode =
pSyncNode->pFsm->FpSnapshotDoWrite(pSyncNode->pFsm, pReceiver->pWriter, pMsg->data, pMsg->dataLen);
ASSERT(writeCode == 0);
pReceiver->ack = pMsg->seq;
}
needRsp = true;
char *msgStr = syncSnapshotSend2Str(pMsg);
char host[128];
uint16_t port;
syncUtilU642Addr(pMsg->srcId.addr, host, sizeof(host), &port);
sTrace("sync event snapshot recv from %s:%d receiving ack:%d, lastIndex:%ld, lastTerm:%lu, recv msg:%s", host,
port, pReceiver->ack, pMsg->lastIndex, pMsg->lastTerm, msgStr);
taosMemoryFree(msgStr);
} else {
ASSERT(0);
}
if (needRsp) {
SyncSnapshotRsp *pRspMsg = syncSnapshotRspBuild(pSyncNode->vgId);
pRspMsg->srcId = pSyncNode->myRaftId;
pRspMsg->destId = pMsg->srcId;
pRspMsg->term = pSyncNode->pRaftStore->currentTerm;
pRspMsg->lastIndex = pMsg->lastIndex;
pRspMsg->lastTerm = pMsg->lastTerm;
pRspMsg->ack = pReceiver->ack;
pRspMsg->code = writeCode;
pRspMsg->privateTerm = pReceiver->privateTerm;
SRpcMsg rpcMsg;
syncSnapshotRsp2RpcMsg(pRspMsg, &rpcMsg);
syncNodeSendMsgById(&(pRspMsg->destId), pSyncNode, &rpcMsg);
syncSnapshotRspDestroy(pRspMsg);
}
}
} else {
syncNodeLog2("syncNodeOnSnapshotSendCb not follower", pSyncNode);
}
return 0;
}
// sender receives ack, set seq = ack + 1, send msg from seq
// if ack == SYNC_SNAPSHOT_SEQ_END, stop sender
int32_t syncNodeOnSnapshotRspCb(SSyncNode *pSyncNode, SyncSnapshotRsp *pMsg) {
// get sender
SSyncSnapshotSender *pSender = syncNodeGetSnapshotSender(pSyncNode, &(pMsg->srcId));
ASSERT(pSender != NULL);
// state, term, seq/ack
if (pSyncNode->state == TAOS_SYNC_STATE_LEADER) {
if (pMsg->term == pSyncNode->pRaftStore->currentTerm) {
// receiver ack is finish, close sender
if (pMsg->ack == SYNC_SNAPSHOT_SEQ_END) {
pSender->finish = true;
snapshotSenderStop(pSender);
return 0;
}
// send next msg
if (pMsg->ack == pSender->seq) {
// update sender ack
pSender->ack = pMsg->ack;
(pSender->seq)++;
snapshotSend(pSender);
} else if (pMsg->ack == pSender->seq - 1) {
snapshotReSend(pSender);
} else {
ASSERT(0);
}
}
} else {
syncNodeLog2("syncNodeOnSnapshotRspCb not leader", pSyncNode);
}
return 0;
}

22
source/libs/sync/src/syncUtil.c
View File

@ -240,4 +240,26 @@ bool syncUtilUserRollback(tmsg_t msgType) {
return true;
}
return false;
}
void syncUtilJson2Line(char* jsonStr) {
int p, q, len;
p = 0;
q = 1;
len = strlen(jsonStr);
while (1) {
if (jsonStr[q] == '\0') {
jsonStr[p + 1] = '\0';
break;
}
if (jsonStr[q] == '\n' || jsonStr[q] == ' ' || jsonStr[q] == '\t') {
q++;
continue;
} else {
jsonStr[p + 1] = jsonStr[q];
p++;
q++;
}
}
}

126
source/libs/sync/test/CMakeLists.txt
View File

@ -38,6 +38,15 @@ add_executable(syncRespMgrTest "")
add_executable(syncSnapshotTest "")
add_executable(syncApplyMsgTest "")
add_executable(syncConfigChangeTest "")
add_executable(syncConfigChangeSnapshotTest "")
add_executable(syncSnapshotSendTest "")
add_executable(syncSnapshotRspTest "")
add_executable(syncSnapshotSenderTest "")
add_executable(syncSnapshotReceiverTest "")
add_executable(syncTestTool "")
add_executable(syncRaftLogTest "")
add_executable(syncRaftLogTest2 "")
add_executable(syncRaftLogTest3 "")
target_sources(syncTest
@ -200,6 +209,42 @@ target_sources(syncConfigChangeTest
PRIVATE
"syncConfigChangeTest.cpp"
)
target_sources(syncConfigChangeSnapshotTest
PRIVATE
"syncConfigChangeSnapshotTest.cpp"
)
target_sources(syncSnapshotSendTest
PRIVATE
"syncSnapshotSendTest.cpp"
)
target_sources(syncSnapshotRspTest
PRIVATE
"syncSnapshotRspTest.cpp"
)
target_sources(syncSnapshotSenderTest
PRIVATE
"syncSnapshotSenderTest.cpp"
)
target_sources(syncSnapshotReceiverTest
PRIVATE
"syncSnapshotReceiverTest.cpp"
)
target_sources(syncTestTool
PRIVATE
"syncTestTool.cpp"
)
target_sources(syncRaftLogTest
PRIVATE
"syncRaftLogTest.cpp"
)
target_sources(syncRaftLogTest2
PRIVATE
"syncRaftLogTest2.cpp"
)
target_sources(syncRaftLogTest3
PRIVATE
"syncRaftLogTest3.cpp"
)
target_include_directories(syncTest
@ -402,6 +447,51 @@ target_include_directories(syncConfigChangeTest
"${TD_SOURCE_DIR}/include/libs/sync"
"${CMAKE_CURRENT_SOURCE_DIR}/../inc"
)
target_include_directories(syncConfigChangeSnapshotTest
PUBLIC
"${TD_SOURCE_DIR}/include/libs/sync"
"${CMAKE_CURRENT_SOURCE_DIR}/../inc"
)
target_include_directories(syncSnapshotSendTest
PUBLIC
"${TD_SOURCE_DIR}/include/libs/sync"
"${CMAKE_CURRENT_SOURCE_DIR}/../inc"
)
target_include_directories(syncSnapshotRspTest
PUBLIC
"${TD_SOURCE_DIR}/include/libs/sync"
"${CMAKE_CURRENT_SOURCE_DIR}/../inc"
)
target_include_directories(syncSnapshotSenderTest
PUBLIC
"${TD_SOURCE_DIR}/include/libs/sync"
"${CMAKE_CURRENT_SOURCE_DIR}/../inc"
)
target_include_directories(syncSnapshotReceiverTest
PUBLIC
"${TD_SOURCE_DIR}/include/libs/sync"
"${CMAKE_CURRENT_SOURCE_DIR}/../inc"
)
target_include_directories(syncTestTool
PUBLIC
"${TD_SOURCE_DIR}/include/libs/sync"
"${CMAKE_CURRENT_SOURCE_DIR}/../inc"
)
target_include_directories(syncRaftLogTest
PUBLIC
"${TD_SOURCE_DIR}/include/libs/sync"
"${CMAKE_CURRENT_SOURCE_DIR}/../inc"
)
target_include_directories(syncRaftLogTest2
PUBLIC
"${TD_SOURCE_DIR}/include/libs/sync"
"${CMAKE_CURRENT_SOURCE_DIR}/../inc"
)
target_include_directories(syncRaftLogTest3
PUBLIC
"${TD_SOURCE_DIR}/include/libs/sync"
"${CMAKE_CURRENT_SOURCE_DIR}/../inc"
)
target_link_libraries(syncTest
@ -564,6 +654,42 @@ target_link_libraries(syncConfigChangeTest
sync
gtest_main
)
target_link_libraries(syncConfigChangeSnapshotTest
sync
gtest_main
)
target_link_libraries(syncSnapshotSendTest
sync
gtest_main
)
target_link_libraries(syncSnapshotRspTest
sync
gtest_main
)
target_link_libraries(syncSnapshotSenderTest
sync
gtest_main
)
target_link_libraries(syncSnapshotReceiverTest
sync
gtest_main
)
target_link_libraries(syncTestTool
sync
gtest_main
)
target_link_libraries(syncRaftLogTest
sync
gtest_main
)
target_link_libraries(syncRaftLogTest2
sync
gtest_main
)
target_link_libraries(syncRaftLogTest3
sync
gtest_main
)
enable_testing()

2
source/libs/sync/test/syncAppendEntriesReplyTest.cpp
View File

@ -22,6 +22,8 @@ SyncAppendEntriesReply *createMsg() {
pMsg->destId.vgId = 100;
pMsg->success = true;
pMsg->matchIndex = 77;
pMsg->term = 33;
pMsg->privateTerm = 44;
return pMsg;
}

1
source/libs/sync/test/syncAppendEntriesTest.cpp
View File

@ -23,6 +23,7 @@ SyncAppendEntries *createMsg() {
pMsg->prevLogIndex = 11;
pMsg->prevLogTerm = 22;
pMsg->commitIndex = 33;
pMsg->privateTerm = 44;
strcpy(pMsg->data, "hello world");
return pMsg;
}

366
source/libs/sync/test/syncConfigChangeSnapshotTest.cpp Normal file
View File

@ -0,0 +1,366 @@
#include <gtest/gtest.h>
#include <stdio.h>
#include "os.h"
#include "syncEnv.h"
#include "syncIO.h"
#include "syncInt.h"
#include "syncUtil.h"
#include "wal.h"
void logTest() {
sTrace("--- sync log test: trace");
sDebug("--- sync log test: debug");
sInfo("--- sync log test: info");
sWarn("--- sync log test: warn");
sError("--- sync log test: error");
sFatal("--- sync log test: fatal");
}
uint16_t gPorts[] = {7010, 7110, 7210, 7310, 7410};
const char* gDir = "./syncReplicateTest";
int32_t gVgId = 1234;
SyncIndex gSnapshotLastApplyIndex;
SyncIndex gSnapshotLastApplyTerm;
void init() {
int code = walInit();
assert(code == 0);
code = syncInit();
assert(code == 0);
sprintf(tsTempDir, "%s", ".");
}
void cleanup() { walCleanUp(); }
void CommitCb(struct SSyncFSM* pFsm, const SRpcMsg* pMsg, SFsmCbMeta cbMeta) {
SyncIndex beginIndex = SYNC_INDEX_INVALID;
if (pFsm->FpGetSnapshot != NULL) {
SSnapshot snapshot;
pFsm->FpGetSnapshot(pFsm, &snapshot);
beginIndex = snapshot.lastApplyIndex;
}
if (cbMeta.index > beginIndex) {
char logBuf[256] = {0};
snprintf(logBuf, sizeof(logBuf),
"==callback== ==CommitCb== pFsm:%p, index:%ld, isWeak:%d, code:%d, state:%d %s, flag:%lu, term:%lu \n",
pFsm, cbMeta.index, cbMeta.isWeak, cbMeta.code, cbMeta.state, syncUtilState2String(cbMeta.state),
cbMeta.flag, cbMeta.term);
syncRpcMsgLog2(logBuf, (SRpcMsg*)pMsg);
} else {
sTrace("==callback== ==CommitCb== do not apply again %ld", cbMeta.index);
}
}
void PreCommitCb(struct SSyncFSM* pFsm, const SRpcMsg* pMsg, SFsmCbMeta cbMeta) {
char logBuf[256] = {0};
snprintf(logBuf, sizeof(logBuf),
"==callback== ==PreCommitCb== pFsm:%p, index:%ld, isWeak:%d, code:%d, state:%d %s flag:%lu\n", pFsm,
cbMeta.index, cbMeta.isWeak, cbMeta.code, cbMeta.state, syncUtilState2String(cbMeta.state), cbMeta.flag);
syncRpcMsgLog2(logBuf, (SRpcMsg*)pMsg);
}
void RollBackCb(struct SSyncFSM* pFsm, const SRpcMsg* pMsg, SFsmCbMeta cbMeta) {
char logBuf[256];
snprintf(logBuf, sizeof(logBuf),
"==callback== ==RollBackCb== pFsm:%p, index:%ld, isWeak:%d, code:%d, state:%d %s flag:%lu\n", pFsm,
cbMeta.index, cbMeta.isWeak, cbMeta.code, cbMeta.state, syncUtilState2String(cbMeta.state), cbMeta.flag);
syncRpcMsgLog2(logBuf, (SRpcMsg*)pMsg);
}
int32_t GetSnapshotCb(struct SSyncFSM* pFsm, SSnapshot* pSnapshot) {
pSnapshot->data = NULL;
pSnapshot->lastApplyIndex = gSnapshotLastApplyIndex;
pSnapshot->lastApplyTerm = gSnapshotLastApplyTerm;
return 0;
}
int32_t SnapshotStartRead(struct SSyncFSM* pFsm, void** ppReader) {
*ppReader = (void*)0xABCD;
char logBuf[256] = {0};
snprintf(logBuf, sizeof(logBuf), "==callback== ==SnapshotStartRead== pFsm:%p, *ppReader:%p", pFsm, *ppReader);
sTrace("%s", logBuf);
return 0;
}
int32_t SnapshotStopRead(struct SSyncFSM* pFsm, void* pReader) {
char logBuf[256] = {0};
snprintf(logBuf, sizeof(logBuf), "==callback== ==SnapshotStopRead== pFsm:%p, pReader:%p", pFsm, pReader);
sTrace("%s", logBuf);
return 0;
}
int32_t SnapshotDoRead(struct SSyncFSM* pFsm, void* pReader, void** ppBuf, int32_t* len) {
static int readIter = 0;
if (readIter == 5) {
*len = 0;
*ppBuf = NULL;
} else if (readIter < 5) {
*len = 20;
*ppBuf = taosMemoryMalloc(*len);
snprintf((char*)*ppBuf, *len, "data iter:%d", readIter);
}
char logBuf[256] = {0};
snprintf(logBuf, sizeof(logBuf),
"==callback== ==SnapshotDoRead== pFsm:%p, pReader:%p, *len:%d, *ppBuf:%s, readIter:%d", pFsm, pReader, *len,
(char*)(*ppBuf), readIter);
sTrace("%s", logBuf);
readIter++;
return 0;
}
int32_t SnapshotStartWrite(struct SSyncFSM* pFsm, void** ppWriter) {
*ppWriter = (void*)0xCDEF;
char logBuf[256] = {0};
snprintf(logBuf, sizeof(logBuf), "==callback== ==SnapshotStartWrite== pFsm:%p, *ppWriter:%p", pFsm, *ppWriter);
sTrace("%s", logBuf);
return 0;
}
int32_t SnapshotStopWrite(struct SSyncFSM* pFsm, void* pWriter, bool isApply) {
char logBuf[256] = {0};
snprintf(logBuf, sizeof(logBuf), "==callback== ==SnapshotStopWrite== pFsm:%p, pWriter:%p, isApply:%d", pFsm, pWriter,
isApply);
sTrace("%s", logBuf);
if (isApply) {
gSnapshotLastApplyIndex = 10;
gSnapshotLastApplyTerm = 1;
}
return 0;
}
int32_t SnapshotDoWrite(struct SSyncFSM* pFsm, void* pWriter, void* pBuf, int32_t len) {
char logBuf[256] = {0};
snprintf(logBuf, sizeof(logBuf), "==callback== ==SnapshotDoWrite== pFsm:%p, pWriter:%p, len:%d pBuf:%s", pFsm,
pWriter, len, (char*)pBuf);
sTrace("%s", logBuf);
return 0;
}
void RestoreFinishCb(struct SSyncFSM* pFsm) { sTrace("==callback== ==RestoreFinishCb=="); }
void ReConfigCb(struct SSyncFSM* pFsm, SSyncCfg newCfg, SReConfigCbMeta cbMeta) {
sTrace("==callback== ==ReConfigCb== flag:0x%lX, isDrop:%d, index:%ld, code:%d, currentTerm:%lu, term:%lu",
cbMeta.flag, cbMeta.isDrop, cbMeta.index, cbMeta.code, cbMeta.currentTerm, cbMeta.term);
}
SSyncFSM* createFsm() {
SSyncFSM* pFsm = (SSyncFSM*)taosMemoryMalloc(sizeof(SSyncFSM));
memset(pFsm, 0, sizeof(*pFsm));
pFsm->FpCommitCb = CommitCb;
pFsm->FpPreCommitCb = PreCommitCb;
pFsm->FpRollBackCb = RollBackCb;
pFsm->FpGetSnapshot = GetSnapshotCb;
pFsm->FpRestoreFinishCb = RestoreFinishCb;
pFsm->FpSnapshotStartRead = SnapshotStartRead;
pFsm->FpSnapshotStopRead = SnapshotStopRead;
pFsm->FpSnapshotDoRead = SnapshotDoRead;
pFsm->FpSnapshotStartWrite = SnapshotStartWrite;
pFsm->FpSnapshotStopWrite = SnapshotStopWrite;
pFsm->FpSnapshotDoWrite = SnapshotDoWrite;
pFsm->FpReConfigCb = ReConfigCb;
return pFsm;
}
SWal* createWal(char* path, int32_t vgId) {
SWalCfg walCfg;
memset(&walCfg, 0, sizeof(SWalCfg));
walCfg.vgId = vgId;
walCfg.fsyncPeriod = 1000;
walCfg.retentionPeriod = 1000;
walCfg.rollPeriod = 1000;
walCfg.retentionSize = 1000;
walCfg.segSize = 1000;
walCfg.level = TAOS_WAL_FSYNC;
SWal* pWal = walOpen(path, &walCfg);
assert(pWal != NULL);
return pWal;
}
int64_t createSyncNode(int32_t replicaNum, int32_t myIndex, int32_t vgId, SWal* pWal, char* path, bool isStandBy) {
SSyncInfo syncInfo;
syncInfo.vgId = vgId;
syncInfo.msgcb = &gSyncIO->msgcb;
syncInfo.FpSendMsg = syncIOSendMsg;
syncInfo.FpEqMsg = syncIOEqMsg;
syncInfo.pFsm = createFsm();
snprintf(syncInfo.path, sizeof(syncInfo.path), "%s_sync_replica%d_index%d", path, replicaNum, myIndex);
syncInfo.pWal = pWal;
syncInfo.isStandBy = isStandBy;
syncInfo.snapshotEnable = true;
SSyncCfg* pCfg = &syncInfo.syncCfg;
if (isStandBy) {
pCfg->myIndex = 0;
pCfg->replicaNum = 1;
pCfg->nodeInfo[0].nodePort = gPorts[myIndex];
taosGetFqdn(pCfg->nodeInfo[0].nodeFqdn);
} else {
pCfg->myIndex = myIndex;
pCfg->replicaNum = replicaNum;
for (int i = 0; i < replicaNum; ++i) {
pCfg->nodeInfo[i].nodePort = gPorts[i];
taosGetFqdn(pCfg->nodeInfo[i].nodeFqdn);
// snprintf(pCfg->nodeInfo[i].nodeFqdn, sizeof(pCfg->nodeInfo[i].nodeFqdn), "%s", "127.0.0.1");
}
}
int64_t rid = syncOpen(&syncInfo);
assert(rid > 0);
SSyncNode* pSyncNode = (SSyncNode*)syncNodeAcquire(rid);
assert(pSyncNode != NULL);
gSyncIO->FpOnSyncPing = pSyncNode->FpOnPing;
gSyncIO->FpOnSyncPingReply = pSyncNode->FpOnPingReply;
gSyncIO->FpOnSyncTimeout = pSyncNode->FpOnTimeout;
gSyncIO->FpOnSyncClientRequest = pSyncNode->FpOnClientRequest;
gSyncIO->FpOnSyncRequestVote = pSyncNode->FpOnRequestVote;
gSyncIO->FpOnSyncRequestVoteReply = pSyncNode->FpOnRequestVoteReply;
gSyncIO->FpOnSyncAppendEntries = pSyncNode->FpOnAppendEntries;
gSyncIO->FpOnSyncAppendEntriesReply = pSyncNode->FpOnAppendEntriesReply;
gSyncIO->FpOnSyncSnapshotSend = pSyncNode->FpOnSnapshotSend;
gSyncIO->FpOnSyncSnapshotRsp = pSyncNode->FpOnSnapshotRsp;
gSyncIO->pSyncNode = pSyncNode;
syncNodeRelease(pSyncNode);
return rid;
}
void configChange(int64_t rid, int32_t replicaNum, int32_t myIndex) {
SSyncCfg syncCfg;
syncCfg.myIndex = myIndex;
syncCfg.replicaNum = replicaNum;
for (int i = 0; i < replicaNum; ++i) {
syncCfg.nodeInfo[i].nodePort = gPorts[i];
taosGetFqdn(syncCfg.nodeInfo[i].nodeFqdn);
}
syncReconfig(rid, &syncCfg);
}
void usage(char* exe) {
printf("usage: %s replicaNum myIndex lastApplyIndex writeRecordNum isStandBy isConfigChange lastApplyTerm \n", exe);
}
SRpcMsg* createRpcMsg(int i, int count, int myIndex) {
SRpcMsg* pMsg = (SRpcMsg*)taosMemoryMalloc(sizeof(SRpcMsg));
memset(pMsg, 0, sizeof(SRpcMsg));
pMsg->msgType = 9999;
pMsg->contLen = 256;
pMsg->pCont = rpcMallocCont(pMsg->contLen);
snprintf((char*)(pMsg->pCont), pMsg->contLen, "value-myIndex:%u-%d-%d-%ld", myIndex, i, count, taosGetTimestampMs());
return pMsg;
}
int main(int argc, char** argv) {
tsAsyncLog = 0;
sDebugFlag = DEBUG_TRACE + DEBUG_SCREEN + DEBUG_FILE + DEBUG_INFO;
if (argc != 8) {
usage(argv[0]);
exit(-1);
}
int32_t replicaNum = atoi(argv[1]);
int32_t myIndex = atoi(argv[2]);
int32_t lastApplyIndex = atoi(argv[3]);
int32_t writeRecordNum = atoi(argv[4]);
bool isStandBy = atoi(argv[5]);
bool isConfigChange = atoi(argv[6]);
int32_t lastApplyTerm = atoi(argv[7]);
sTrace(
"args: replicaNum:%d, myIndex:%d, lastApplyIndex:%d, writeRecordNum:%d, isStandBy:%d, isConfigChange:%d, "
"lastApplyTerm:%d",
replicaNum, myIndex, lastApplyIndex, writeRecordNum, isStandBy, isConfigChange, lastApplyTerm);
gSnapshotLastApplyIndex = lastApplyIndex;
gSnapshotLastApplyTerm = lastApplyTerm;
if (!isStandBy) {
assert(replicaNum >= 1 && replicaNum <= 5);
assert(myIndex >= 0 && myIndex < replicaNum);
assert(lastApplyIndex >= -1);
assert(writeRecordNum >= 0);
}
init();
int32_t ret = syncIOStart((char*)"127.0.0.1", gPorts[myIndex]);
assert(ret == 0);
char walPath[128];
snprintf(walPath, sizeof(walPath), "%s_wal_replica%d_index%d", gDir, replicaNum, myIndex);
SWal* pWal = createWal(walPath, gVgId);
int64_t rid = createSyncNode(replicaNum, myIndex, gVgId, pWal, (char*)gDir, isStandBy);
assert(rid > 0);
syncStart(rid);
/*
if (isStandBy) {
syncStartStandBy(rid);
} else {
syncStart(rid);
}
*/
SSyncNode* pSyncNode = (SSyncNode*)syncNodeAcquire(rid);
assert(pSyncNode != NULL);
if (isConfigChange) {
configChange(rid, 2, myIndex);
}
//---------------------------
int32_t alreadySend = 0;
while (1) {
char* s = syncNode2SimpleStr(pSyncNode);
if (alreadySend < writeRecordNum) {
SRpcMsg* pRpcMsg = createRpcMsg(alreadySend, writeRecordNum, myIndex);
int32_t ret = syncPropose(rid, pRpcMsg, false);
if (ret == TAOS_SYNC_PROPOSE_NOT_LEADER) {
sTrace("%s value%d write not leader", s, alreadySend);
} else {
assert(ret == 0);
sTrace("%s value%d write ok", s, alreadySend);
}
alreadySend++;
rpcFreeCont(pRpcMsg->pCont);
taosMemoryFree(pRpcMsg);
} else {
sTrace("%s", s);
}
taosMsleep(1000);
taosMemoryFree(s);
taosMsleep(1000);
}
syncNodeRelease(pSyncNode);
syncStop(rid);
walClose(pWal);
syncIOStop();
cleanup();
return 0;
}

1
source/libs/sync/test/syncConfigChangeTest.cpp
View File

@ -93,7 +93,6 @@ SSyncFSM* createFsm() {
pFsm->FpGetSnapshot = GetSnapshotCb;
pFsm->FpRestoreFinishCb = RestoreFinishCb;
pFsm->FpReConfigCb = ReConfigCb;
return pFsm;

82
source/libs/sync/test/syncIndexMgrTest.cpp
View File

@ -22,55 +22,23 @@ int32_t replicaNum = 3;
int32_t myIndex = 0;
SRaftId ids[TSDB_MAX_REPLICA];
SSyncInfo syncInfo;
SSyncFSM* pFsm;
SSyncNode* pSyncNode;
SSyncNode* syncNodeInit() {
syncInfo.vgId = 1234;
syncInfo.msgcb = &gSyncIO->msgcb;
syncInfo.FpSendMsg = syncIOSendMsg;
syncInfo.FpEqMsg = syncIOEqMsg;
syncInfo.pFsm = pFsm;
snprintf(syncInfo.path, sizeof(syncInfo.path), "%s", "./");
SSyncCfg* pCfg = &syncInfo.syncCfg;
pCfg->myIndex = myIndex;
pCfg->replicaNum = replicaNum;
pSyncNode = (SSyncNode*)taosMemoryMalloc(sizeof(SSyncNode));
memset(pSyncNode, 0, sizeof(SSyncNode));
pSyncNode->replicaNum = replicaNum;
for (int i = 0; i < replicaNum; ++i) {
pCfg->nodeInfo[i].nodePort = ports[i];
snprintf(pCfg->nodeInfo[i].nodeFqdn, sizeof(pCfg->nodeInfo[i].nodeFqdn), "%s", "127.0.0.1");
// taosGetFqdn(pCfg->nodeInfo[0].nodeFqdn);
pSyncNode->replicasId[i].addr = syncUtilAddr2U64("127.0.0.1", ports[i]);
pSyncNode->replicasId[i].vgId = 1234;
ids[i].addr = pSyncNode->replicasId[i].addr;
ids[i].vgId = pSyncNode->replicasId[i].vgId;
}
pSyncNode = syncNodeOpen(&syncInfo);
assert(pSyncNode != NULL);
gSyncIO->FpOnSyncPing = pSyncNode->FpOnPing;
gSyncIO->FpOnSyncPingReply = pSyncNode->FpOnPingReply;
gSyncIO->FpOnSyncRequestVote = pSyncNode->FpOnRequestVote;
gSyncIO->FpOnSyncRequestVoteReply = pSyncNode->FpOnRequestVoteReply;
gSyncIO->FpOnSyncAppendEntries = pSyncNode->FpOnAppendEntries;
gSyncIO->FpOnSyncAppendEntriesReply = pSyncNode->FpOnAppendEntriesReply;
gSyncIO->FpOnSyncPing = pSyncNode->FpOnPing;
gSyncIO->FpOnSyncPingReply = pSyncNode->FpOnPingReply;
gSyncIO->pSyncNode = pSyncNode;
return pSyncNode;
}
SSyncNode* syncInitTest() { return syncNodeInit(); }
void initRaftId(SSyncNode* pSyncNode) {
for (int i = 0; i < replicaNum; ++i) {
ids[i] = pSyncNode->replicasId[i];
char* s = syncUtilRaftId2Str(&ids[i]);
printf("raftId[%d] : %s\n", i, s);
taosMemoryFree(s);
}
}
int main(int argc, char** argv) {
tsAsyncLog = 0;
sDebugFlag = DEBUG_TRACE + DEBUG_SCREEN + DEBUG_FILE;
@ -80,58 +48,52 @@ int main(int argc, char** argv) {
myIndex = atoi(argv[1]);
}
int32_t ret = syncIOStart((char*)"127.0.0.1", ports[myIndex]);
assert(ret == 0);
ret = syncEnvStart();
assert(ret == 0);
SSyncNode* pSyncNode = syncInitTest();
SSyncNode* pSyncNode = syncNodeInit();
assert(pSyncNode != NULL);
char* serialized = syncNode2Str(pSyncNode);
printf("%s\n", serialized);
taosMemoryFree(serialized);
initRaftId(pSyncNode);
printf("---------------------------------------\n");
SSyncIndexMgr* pSyncIndexMgr = syncIndexMgrCreate(pSyncNode);
assert(pSyncIndexMgr != NULL);
printf("---------------------------------------\n");
{
char* serialized = syncIndexMgr2Str(pSyncIndexMgr);
assert(serialized != NULL);
printf("%s\n", serialized);
taosMemoryFree(serialized);
}
printf("---------------------------------------\n");
printf("---------------------------------------\n");
syncIndexMgrSetIndex(pSyncIndexMgr, &ids[0], 100);
syncIndexMgrSetIndex(pSyncIndexMgr, &ids[1], 200);
syncIndexMgrSetIndex(pSyncIndexMgr, &ids[2], 300);
printf("---------------------------------------\n");
// syncIndexMgrSetTerm(pSyncIndexMgr, &ids[0], 700);
// syncIndexMgrSetTerm(pSyncIndexMgr, &ids[1], 800);
// syncIndexMgrSetTerm(pSyncIndexMgr, &ids[2], 900);
{
char* serialized = syncIndexMgr2Str(pSyncIndexMgr);
assert(serialized != NULL);
printf("%s\n", serialized);
taosMemoryFree(serialized);
}
printf("---------------------------------------\n");
printf("---------------------------------------\n");
for (int i = 0; i < pSyncIndexMgr->replicaNum; ++i) {
SyncIndex idx = syncIndexMgrGetIndex(pSyncIndexMgr, &ids[i]);
printf("index %d : %lu \n", i, idx);
// SyncTerm term = syncIndexMgrGetTerm(pSyncIndexMgr, &ids[i]);
// printf("%d: index:%ld term:%lu \n", i, idx, term);
}
syncIndexMgrClear(pSyncIndexMgr);
printf("---------------------------------------\n");
printf("---------------------------------------\n");
syncIndexMgrClear(pSyncIndexMgr);
{
char* serialized = syncIndexMgr2Str(pSyncIndexMgr);
assert(serialized != NULL);
printf("%s\n", serialized);
taosMemoryFree(serialized);
}
printf("---------------------------------------\n");
syncIndexMgrDestroy(pSyncIndexMgr);
return 0;

6
source/libs/sync/test/syncRaftCfgTest.cpp
View File

@ -71,7 +71,10 @@ void test3() {
if (taosCheckExistFile(s)) {
printf("%s file: %s already exist! \n", (char*)__FUNCTION__, s);
} else {
raftCfgCreateFile(pCfg, 7, s);
SRaftCfgMeta meta;
meta.isStandBy = 7;
meta.snapshotEnable = 9;
raftCfgCreateFile(pCfg, meta, s);
printf("%s create json file: %s \n", (char*)__FUNCTION__, s);
}
@ -94,6 +97,7 @@ void test5() {
pCfg->cfg.myIndex = taosGetTimestampSec();
pCfg->isStandBy += 2;
pCfg->snapshotEnable += 3;
raftCfgPersist(pCfg);
printf("%s update json file: %s myIndex->%d \n", (char*)__FUNCTION__, "./test3_raft_cfg.json", pCfg->cfg.myIndex);

Some files were not shown because too many files have changed in this diff Show More