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

other: merge 3.0

This commit is contained in:
Haojun Liao 2022-06-21 09:44:37 +08:00
parent e5d25f4634 4cacb51efa
commit bbe1deaf00
95 changed files with 3270 additions and 1427 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

15
examples/c/stream_demo.c
View File

@ -25,19 +25,21 @@ int32_t init_env() {
return -1;
}
TAOS_RES* pRes = taos_query(pConn, "create database if not exists abc1 vgroups 2");
TAOS_RES* pRes = taos_query(pConn, "create database if not exists abc1 vgroups 1");
if (taos_errno(pRes) != 0) {
printf("error in create db, reason:%s\n", taos_errstr(pRes));
return -1;
}
taos_free_result(pRes);
#if 0
pRes = taos_query(pConn, "create database if not exists abc2 vgroups 20");
if (taos_errno(pRes) != 0) {
printf("error in create db, reason:%s\n", taos_errstr(pRes));
return -1;
}
taos_free_result(pRes);
#endif
pRes = taos_query(pConn, "use abc1");
if (taos_errno(pRes) != 0) {
@ -89,8 +91,8 @@ int32_t create_stream() {
/*const char* sql = "select sum(k) from tu1 interval(10m)";*/
/*pRes = tmq_create_stream(pConn, "stream1", "out1", sql);*/
pRes = taos_query(pConn,
"create stream stream1 trigger at_once into abc2.outstb as select _wstartts, sum(k) from st1 "
"partition by tbname interval(10m) ");
"create stream stream1 trigger window_close into outstb as select _wstartts, sum(k) from st1 "
"interval(10s) ");
if (taos_errno(pRes) != 0) {
printf("failed to create stream stream1, reason:%s\n", taos_errstr(pRes));
return -1;
@ -107,11 +109,4 @@ int main(int argc, char* argv[]) {
code = init_env();
}
create_stream();
#if 0
tmq_t* tmq = build_consumer();
tmq_list_t* topic_list = build_topic_list();
/*perf_loop(tmq, topic_list);*/
/*basic_consume_loop(tmq, topic_list);*/
sync_consume_loop(tmq, topic_list);
#endif
}

10
include/common/tmsg.h
View File

@ -1776,12 +1776,10 @@ typedef struct {
} SDDropTopicReq;
typedef struct {
float xFilesFactor;
int32_t delay;
int32_t qmsg1Len;
int32_t qmsg2Len;
char* qmsg1; // pAst1:qmsg1:SRetention1 => trigger aggr task1
char* qmsg2; // pAst2:qmsg2:SRetention2 => trigger aggr task2
int64_t maxdelay[2];
int64_t watermark[2];
int32_t qmsgLen[2];
char* qmsg[2]; // pAst:qmsg:SRetention => trigger aggr task1/2
} SRSmaParam;
int32_t tEncodeSRSmaParam(SEncoder* pCoder, const SRSmaParam* pRSmaParam);

1
include/common/tmsgcb.h
View File

@ -34,7 +34,6 @@ typedef enum {
WRITE_QUEUE,
APPLY_QUEUE,
SYNC_QUEUE,
MERGE_QUEUE,
QUEUE_MAX,
} EQueueType;

1
include/common/tmsgdef.h
View File

@ -236,6 +236,7 @@ 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_CONFIG_CHANGE_FINISH, "sync-config-change-finish", 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)
TD_DEF_MSG_TYPE(TDMT_SYNC_LEADER_TRANSFER, "sync-leader-transfer", NULL, NULL)

6
include/dnode/snode/snode.h
View File

@ -16,8 +16,8 @@
#ifndef _TD_SNODE_H_
#define _TD_SNODE_H_
#include "tmsgcb.h"
#include "tmsg.h"
#include "tmsgcb.h"
#include "trpc.h"
#ifdef __cplusplus
@ -68,8 +68,8 @@ int32_t sndGetLoad(SSnode *pSnode, SSnodeLoad *pLoad);
* @param pMsg The request message
* @param pRsp The response message
*/
void sndProcessUMsg(SSnode *pSnode, SRpcMsg *pMsg);
void sndProcessSMsg(SSnode *pSnode, SRpcMsg *pMsg);
int32_t sndProcessUMsg(SSnode *pSnode, SRpcMsg *pMsg);
int32_t sndProcessSMsg(SSnode *pSnode, SRpcMsg *pMsg);
#ifdef __cplusplus
}

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

@ -36,7 +36,6 @@ typedef struct SPlanContext {
int64_t watermark;
char* pMsg;
int32_t msgLen;
// double filesFactor;
} SPlanContext;
// Create the physical plan for the query, according to the AST.

16
include/libs/stream/tstream.h
View File

@ -152,7 +152,6 @@ void* streamDataBlockDecode(const void* buf, SStreamDataBlock* pInput);
typedef struct {
char* qmsg;
// followings are not applicable to encoder and decoder
void* inputHandle;
void* executor;
} STaskExec;
@ -240,12 +239,13 @@ struct SStreamTask {
int8_t inputType;
int8_t status;
int8_t sourceType;
int8_t execType;
int8_t sinkType;
int8_t dispatchType;
int16_t dispatchMsgType;
int8_t dataScan;
// node info
int32_t childId;
int32_t nodeId;
@ -399,15 +399,13 @@ typedef struct {
int32_t tDecodeStreamDispatchReq(SDecoder* pDecoder, SStreamDispatchReq* pReq);
int32_t streamLaunchByWrite(SStreamTask* pTask, int32_t vgId, SMsgCb* pMsgCb);
int32_t streamLaunchByWrite(SStreamTask* pTask, int32_t vgId);
int32_t streamSetupTrigger(SStreamTask* pTask);
int32_t streamTaskRun(SStreamTask* pTask);
int32_t streamTaskProcessRunReq(SStreamTask* pTask, SMsgCb* pMsgCb);
int32_t streamProcessDispatchReq(SStreamTask* pTask, SMsgCb* pMsgCb, SStreamDispatchReq* pReq, SRpcMsg* pMsg);
int32_t streamProcessDispatchRsp(SStreamTask* pTask, SMsgCb* pMsgCb, SStreamDispatchRsp* pRsp);
int32_t streamProcessRecoverReq(SStreamTask* pTask, SMsgCb* pMsgCb, SStreamTaskRecoverReq* pReq, SRpcMsg* pMsg);
int32_t streamProcessRunReq(SStreamTask* pTask);
int32_t streamProcessDispatchReq(SStreamTask* pTask, SStreamDispatchReq* pReq, SRpcMsg* pMsg);
int32_t streamProcessDispatchRsp(SStreamTask* pTask, SStreamDispatchRsp* pRsp);
int32_t streamProcessRecoverReq(SStreamTask* pTask, SStreamTaskRecoverReq* pReq, SRpcMsg* pMsg);
int32_t streamProcessRecoverRsp(SStreamTask* pTask, SStreamTaskRecoverRsp* pRsp);
#ifdef __cplusplus

29
include/libs/sync/sync.h
View File

@ -24,6 +24,8 @@ extern "C" {
#include "tdef.h"
#include "tmsgcb.h"
extern bool gRaftDetailLog;
#define SYNC_INDEX_BEGIN 0
#define SYNC_INDEX_INVALID -1
@ -61,14 +63,14 @@ typedef struct SSyncCfg {
} SSyncCfg;
typedef struct SFsmCbMeta {
SyncIndex index;
SyncIndex lastConfigIndex;
bool isWeak;
int32_t code;
ESyncState state;
uint64_t seqNum;
SyncIndex index;
SyncTerm term;
uint64_t seqNum;
SyncIndex lastConfigIndex;
ESyncState state;
SyncTerm currentTerm;
bool isWeak;
uint64_t flag;
} SFsmCbMeta;
@ -76,13 +78,20 @@ typedef struct SReConfigCbMeta {
int32_t code;
SyncIndex index;
SyncTerm term;
uint64_t seqNum;
SyncIndex lastConfigIndex;
ESyncState state;
SyncTerm currentTerm;
bool isWeak;
uint64_t flag;
// config info
SSyncCfg oldCfg;
SSyncCfg newCfg;
bool isDrop;
uint64_t flag;
uint64_t seqNum;
SyncIndex newCfgIndex;
SyncTerm newCfgTerm;
uint64_t newCfgSeqNum;
} SReConfigCbMeta;
typedef struct SSnapshot {
@ -107,7 +116,6 @@ typedef struct SSyncFSM {
void (*FpReConfigCb)(struct SSyncFSM* pFsm, const SRpcMsg* pMsg, SReConfigCbMeta cbMeta);
void (*FpLeaderTransferCb)(struct SSyncFSM* pFsm, const SRpcMsg* pMsg, SFsmCbMeta cbMeta);
int32_t (*FpGetSnapshot)(struct SSyncFSM* pFsm, SSnapshot* pSnapshot, void* pReaderParam, void** ppReader);
int32_t (*FpGetSnapshotInfo)(struct SSyncFSM* pFsm, SSnapshot* pSnapshot);
@ -189,14 +197,13 @@ ESyncState syncGetMyRole(int64_t rid);
bool syncIsReady(int64_t rid);
const char* syncGetMyRoleStr(int64_t rid);
SyncTerm syncGetMyTerm(int64_t rid);
SyncGroupId syncGetVgId(int64_t rid);
void syncGetEpSet(int64_t rid, SEpSet* pEpSet);
int32_t syncGetVgId(int64_t rid);
int32_t syncPropose(int64_t rid, const SRpcMsg* pMsg, bool isWeak);
bool syncEnvIsStart();
const char* syncStr(ESyncState state);
bool syncIsRestoreFinish(int64_t rid);
int32_t syncReconfig(int64_t rid, const SSyncCfg* pNewCfg);
// build SRpcMsg, need to call syncPropose with SRpcMsg

34
include/libs/sync/syncTools.h
View File

@ -489,6 +489,40 @@ void syncLeaderTransferPrint2(char* s, const SyncLeaderTransfer* pMsg);
void syncLeaderTransferLog(const SyncLeaderTransfer* pMsg);
void syncLeaderTransferLog2(char* s, const SyncLeaderTransfer* pMsg);
// ---------------------------------------------
typedef struct SyncReconfigFinish {
uint32_t bytes;
int32_t vgId;
uint32_t msgType;
SSyncCfg oldCfg;
SSyncCfg newCfg;
SyncIndex newCfgIndex;
SyncTerm newCfgTerm;
uint64_t newCfgSeqNum;
} SyncReconfigFinish;
SyncReconfigFinish* syncReconfigFinishBuild(int32_t vgId);
void syncReconfigFinishDestroy(SyncReconfigFinish* pMsg);
void syncReconfigFinishSerialize(const SyncReconfigFinish* pMsg, char* buf, uint32_t bufLen);
void syncReconfigFinishDeserialize(const char* buf, uint32_t len, SyncReconfigFinish* pMsg);
char* syncReconfigFinishSerialize2(const SyncReconfigFinish* pMsg, uint32_t* len);
SyncReconfigFinish* syncReconfigFinishDeserialize2(const char* buf, uint32_t len);
void syncReconfigFinish2RpcMsg(const SyncReconfigFinish* pMsg, SRpcMsg* pRpcMsg);
void syncReconfigFinishFromRpcMsg(const SRpcMsg* pRpcMsg, SyncReconfigFinish* pMsg);
SyncReconfigFinish* syncReconfigFinishFromRpcMsg2(const SRpcMsg* pRpcMsg);
cJSON* syncReconfigFinish2Json(const SyncReconfigFinish* pMsg);
char* syncReconfigFinish2Str(const SyncReconfigFinish* pMsg);
// for debug ----------------------
void syncReconfigFinishPrint(const SyncReconfigFinish* pMsg);
void syncReconfigFinishPrint2(char* s, const SyncReconfigFinish* pMsg);
void syncReconfigFinishLog(const SyncReconfigFinish* pMsg);
void syncReconfigFinishLog2(char* s, const SyncReconfigFinish* pMsg);
// on message ----------------------
int32_t syncNodeOnPingCb(SSyncNode* ths, SyncPing* pMsg);
int32_t syncNodeOnPingReplyCb(SSyncNode* ths, SyncPingReply* pMsg);

3
include/util/taoserror.h
View File

@ -416,6 +416,9 @@ int32_t* taosGetErrno();
#define TSDB_CODE_SYN_NOT_LEADER TAOS_DEF_ERROR_CODE(0, 0x090C)
#define TSDB_CODE_SYN_ONE_REPLICA TAOS_DEF_ERROR_CODE(0, 0x090D)
#define TSDB_CODE_SYN_NOT_IN_NEW_CONFIG TAOS_DEF_ERROR_CODE(0, 0x090E)
#define TSDB_CODE_SYN_NEW_CONFIG_ERROR TAOS_DEF_ERROR_CODE(0, 0x090F)
#define TSDB_CODE_SYN_RECONFIG_NOT_READY TAOS_DEF_ERROR_CODE(0, 0x0910)
#define TSDB_CODE_SYN_PROPOSE_NOT_READY TAOS_DEF_ERROR_CODE(0, 0x0911)
#define TSDB_CODE_SYN_INTERNAL_ERROR TAOS_DEF_ERROR_CODE(0, 0x09FF)
// tq

4
source/common/src/tglobal.c
View File

@ -161,7 +161,7 @@ int32_t tsDiskCfgNum = 0;
SDiskCfg tsDiskCfg[TFS_MAX_DISKS] = {0};
// stream scheduler
bool tsStreamSchedV = true;
bool tsSchedStreamToSnode = true;
/*
* minimum scale for whole system, millisecond by default
@ -185,7 +185,7 @@ char tsCompressor[32] = "ZSTD_COMPRESSOR"; // ZSTD_COMPRESSOR or GZIP_COMPR
bool tsStartUdfd = true;
// internal
int32_t tsTransPullupInterval = 6;
int32_t tsTransPullupInterval = 2;
int32_t tsMqRebalanceInterval = 2;
void taosAddDataDir(int32_t index, char *v1, int32_t level, int32_t primary) {

37
source/common/src/tmsg.c
View File

@ -4273,39 +4273,34 @@ void tFreeSCMCreateStreamReq(SCMCreateStreamReq *pReq) {
}
int32_t tEncodeSRSmaParam(SEncoder *pCoder, const SRSmaParam *pRSmaParam) {
if (tEncodeFloat(pCoder, pRSmaParam->xFilesFactor) < 0) return -1;
if (tEncodeI32v(pCoder, pRSmaParam->delay) < 0) return -1;
if (tEncodeI32v(pCoder, pRSmaParam->qmsg1Len) < 0) return -1;
if (tEncodeI32v(pCoder, pRSmaParam->qmsg2Len) < 0) return -1;
if (pRSmaParam->qmsg1Len > 0) {
if (tEncodeBinary(pCoder, pRSmaParam->qmsg1, (uint64_t)pRSmaParam->qmsg1Len) < 0) // qmsg1Len contains len of '\0'
for (int32_t i = 0; i < 2; ++i) {
if (tEncodeI64v(pCoder, pRSmaParam->maxdelay[i]) < 0) return -1;
if (tEncodeI64v(pCoder, pRSmaParam->watermark[i]) < 0) return -1;
if (tEncodeI32v(pCoder, pRSmaParam->qmsgLen[i]) < 0) return -1;
if (pRSmaParam->qmsgLen[i] > 0) {
if (tEncodeBinary(pCoder, pRSmaParam->qmsg[i], (uint64_t)pRSmaParam->qmsgLen[i]) <
0) // qmsgLen contains len of '\0'
return -1;
}
if (pRSmaParam->qmsg2Len > 0) {
if (tEncodeBinary(pCoder, pRSmaParam->qmsg2, (uint64_t)pRSmaParam->qmsg2Len) < 0) // qmsg2Len contains len of '\0'
return -1;
}
return 0;
}
int32_t tDecodeSRSmaParam(SDecoder *pCoder, SRSmaParam *pRSmaParam) {
if (tDecodeFloat(pCoder, &pRSmaParam->xFilesFactor) < 0) return -1;
if (tDecodeI32v(pCoder, &pRSmaParam->delay) < 0) return -1;
if (tDecodeI32v(pCoder, &pRSmaParam->qmsg1Len) < 0) return -1;
if (tDecodeI32v(pCoder, &pRSmaParam->qmsg2Len) < 0) return -1;
if (pRSmaParam->qmsg1Len > 0) {
for (int32_t i = 0; i < 2; ++i) {
if (tDecodeI64v(pCoder, &pRSmaParam->maxdelay[i]) < 0) return -1;
if (tDecodeI64v(pCoder, &pRSmaParam->watermark[i]) < 0) return -1;
if (tDecodeI32v(pCoder, &pRSmaParam->qmsgLen[i]) < 0) return -1;
if (pRSmaParam->qmsgLen[i] > 0) {
uint64_t len;
if (tDecodeBinaryAlloc(pCoder, (void **)&pRSmaParam->qmsg1, &len) < 0) return -1; // qmsg1Len contains len of '\0'
if (tDecodeBinaryAlloc(pCoder, (void **)&pRSmaParam->qmsg[i], &len) < 0)
return -1; // qmsgLen contains len of '\0'
} else {
pRSmaParam->qmsg1 = NULL;
pRSmaParam->qmsg[i] = NULL;
}
if (pRSmaParam->qmsg2Len > 0) {
uint64_t len;
if (tDecodeBinaryAlloc(pCoder, (void **)&pRSmaParam->qmsg2, &len) < 0) return -1; // qmsg2Len contains len of '\0'
} else {
pRSmaParam->qmsg2 = NULL;
}
return 0;
}

1
source/dnode/mgmt/mgmt_mnode/src/mmHandle.c
View File

@ -183,6 +183,7 @@ SArray *mmGetMsgHandles() {
if (dmSetMgmtHandle(pArray, TDMT_MND_CREATE_SMA, mmPutMsgToWriteQueue, 0) == NULL) goto _OVER;
if (dmSetMgmtHandle(pArray, TDMT_MND_DROP_SMA, mmPutMsgToWriteQueue, 0) == NULL) goto _OVER;
if (dmSetMgmtHandle(pArray, TDMT_MND_CREATE_STREAM, mmPutMsgToWriteQueue, 0) == NULL) goto _OVER;
if (dmSetMgmtHandle(pArray, TDMT_MND_DROP_STREAM, mmPutMsgToWriteQueue, 0) == NULL) goto _OVER;
if (dmSetMgmtHandle(pArray, TDMT_MND_GET_INDEX, mmPutMsgToReadQueue, 0) == NULL) goto _OVER;
if (dmSetMgmtHandle(pArray, TDMT_MND_GET_TABLE_INDEX, mmPutMsgToReadQueue, 0) == NULL) goto _OVER;
if (dmSetMgmtHandle(pArray, TDMT_MND_CREATE_TOPIC, mmPutMsgToWriteQueue, 0) == NULL) goto _OVER;

9
source/dnode/mgmt/mgmt_snode/src/smHandle.c
View File

@ -95,9 +95,12 @@ SArray *smGetMsgHandles() {
if (dmSetMgmtHandle(pArray, TDMT_MON_SM_INFO, smPutNodeMsgToMonitorQueue, 0) == NULL) goto _OVER;
if (dmSetMgmtHandle(pArray, TDMT_STREAM_TASK_DEPLOY, smPutNodeMsgToMgmtQueue, 1) == NULL) goto _OVER;
if (dmSetMgmtHandle(pArray, TDMT_STREAM_TASK_RUN, smPutNodeMsgToMgmtQueue, 1) == NULL) goto _OVER;
if (dmSetMgmtHandle(pArray, TDMT_STREAM_TASK_DISPATCH, smPutNodeMsgToMgmtQueue, 1) == NULL) goto _OVER;
if (dmSetMgmtHandle(pArray, TDMT_STREAM_TASK_RECOVER, smPutNodeMsgToMgmtQueue, 1) == NULL) goto _OVER;
if (dmSetMgmtHandle(pArray, TDMT_VND_STREAM_TASK_DROP, smPutNodeMsgToMgmtQueue, 1) == NULL) goto _OVER;
if (dmSetMgmtHandle(pArray, TDMT_STREAM_TASK_RUN, smPutNodeMsgToSharedQueue, 1) == NULL) goto _OVER;
if (dmSetMgmtHandle(pArray, TDMT_STREAM_TASK_DISPATCH, smPutNodeMsgToSharedQueue, 1) == NULL) goto _OVER;
if (dmSetMgmtHandle(pArray, TDMT_STREAM_TASK_DISPATCH_RSP, smPutNodeMsgToSharedQueue, 1) == NULL) goto _OVER;
if (dmSetMgmtHandle(pArray, TDMT_STREAM_TASK_RECOVER, smPutNodeMsgToSharedQueue, 1) == NULL) goto _OVER;
if (dmSetMgmtHandle(pArray, TDMT_STREAM_TASK_RECOVER_RSP, smPutNodeMsgToSharedQueue, 1) == NULL) goto _OVER;
code = 0;
_OVER:

10
source/dnode/mgmt/mgmt_snode/src/smWorker.c
View File

@ -55,7 +55,10 @@ static void smProcessUniqueQueue(SQueueInfo *pInfo, STaosQall *qall, int32_t num
taosGetQitem(qall, (void **)&pMsg);
dTrace("msg:%p, get from snode-unique queue", pMsg);
sndProcessUMsg(pMgmt->pSnode, pMsg);
if (sndProcessUMsg(pMgmt->pSnode, pMsg) < 0) {
ASSERT(0);
}
smSendRsp(pMsg, 0);
dTrace("msg:%p, is freed", pMsg);
rpcFreeCont(pMsg->pCont);
@ -67,7 +70,10 @@ static void smProcessSharedQueue(SQueueInfo *pInfo, SRpcMsg *pMsg) {
SSnodeMgmt *pMgmt = pInfo->ahandle;
dTrace("msg:%p, get from snode-shared queue", pMsg);
sndProcessSMsg(pMgmt->pSnode, pMsg);
if (sndProcessSMsg(pMgmt->pSnode, pMsg) < 0) {
smSendRsp(pMsg, terrno);
ASSERT(0);
}
dTrace("msg:%p, is freed", pMsg);
rpcFreeCont(pMsg->pCont);

2
source/dnode/mgmt/mgmt_vnode/inc/vmInt.h
View File

@ -35,7 +35,6 @@ typedef struct SVnodeMgmt {
SWWorkerPool syncPool;
SWWorkerPool writePool;
SWWorkerPool applyPool;
SWWorkerPool mergePool;
SSingleWorker mgmtWorker;
SSingleWorker monitorWorker;
SHashObj *hash;
@ -63,7 +62,6 @@ typedef struct {
STaosQueue *pApplyQ;
STaosQueue *pQueryQ;
STaosQueue *pFetchQ;
STaosQueue *pMergeQ;
} SVnodeObj;
typedef struct {

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

@ -86,7 +86,6 @@ void vmCloseVnode(SVnodeMgmt *pMgmt, SVnodeObj *pVnode) {
while (!taosQueueEmpty(pVnode->pApplyQ)) taosMsleep(10);
while (!taosQueueEmpty(pVnode->pQueryQ)) taosMsleep(10);
while (!taosQueueEmpty(pVnode->pFetchQ)) taosMsleep(10);
while (!taosQueueEmpty(pVnode->pMergeQ)) taosMsleep(10);
vmFreeQueue(pMgmt, pVnode);
vnodeClose(pVnode->pImpl);

20
source/dnode/mgmt/mgmt_vnode/src/vmWorker.c
View File

@ -170,10 +170,6 @@ static int32_t vmPutMsgToQueue(SVnodeMgmt *pMgmt, SRpcMsg *pMsg, EQueueType qtyp
dTrace("vgId:%d, msg:%p put into vnode-sync queue", pVnode->vgId, pMsg);
taosWriteQitem(pVnode->pSyncQ, pMsg);
break;
case MERGE_QUEUE:
dTrace("vgId:%d, msg:%p put into vnode-merge queue", pVnode->vgId, pMsg);
taosWriteQitem(pVnode->pMergeQ, pMsg);
break;
case APPLY_QUEUE:
dTrace("vgId:%d, msg:%p put into vnode-apply queue", pVnode->vgId, pMsg);
taosWriteQitem(pVnode->pApplyQ, pMsg);
@ -196,8 +192,6 @@ int32_t vmPutMsgToQueryQueue(SVnodeMgmt *pMgmt, SRpcMsg *pMsg) { return vmPutMsg
int32_t vmPutMsgToFetchQueue(SVnodeMgmt *pMgmt, SRpcMsg *pMsg) { return vmPutMsgToQueue(pMgmt, pMsg, FETCH_QUEUE); }
int32_t vmPutMsgToMergeQueue(SVnodeMgmt *pMgmt, SRpcMsg *pMsg) { return vmPutMsgToQueue(pMgmt, pMsg, MERGE_QUEUE); }
int32_t vmPutMsgToMgmtQueue(SVnodeMgmt *pMgmt, SRpcMsg *pMsg) {
dTrace("msg:%p, put into vnode-mgmt queue", pMsg);
taosWriteQitem(pMgmt->mgmtWorker.queue, pMsg);
@ -243,9 +237,6 @@ int32_t vmGetQueueSize(SVnodeMgmt *pMgmt, int32_t vgId, EQueueType qtype) {
case FETCH_QUEUE:
size = taosQueueItemSize(pVnode->pFetchQ);
break;
case MERGE_QUEUE:
size = taosQueueItemSize(pVnode->pMergeQ);
break;
default:
break;
}
@ -260,10 +251,9 @@ int32_t vmAllocQueue(SVnodeMgmt *pMgmt, SVnodeObj *pVnode) {
pVnode->pApplyQ = tWWorkerAllocQueue(&pMgmt->applyPool, pVnode->pImpl, (FItems)vnodeApplyMsg);
pVnode->pQueryQ = tQWorkerAllocQueue(&pMgmt->queryPool, pVnode, (FItem)vmProcessQueryQueue);
pVnode->pFetchQ = tQWorkerAllocQueue(&pMgmt->fetchPool, pVnode, (FItem)vmProcessFetchQueue);
pVnode->pMergeQ = tWWorkerAllocQueue(&pMgmt->mergePool, pVnode, (FItems)vmProcessMergeQueue);
if (pVnode->pWriteQ == NULL || pVnode->pSyncQ == NULL || pVnode->pApplyQ == NULL || pVnode->pQueryQ == NULL ||
pVnode->pFetchQ == NULL || pVnode->pMergeQ == NULL) {
pVnode->pFetchQ == NULL) {
terrno = TSDB_CODE_OUT_OF_MEMORY;
return -1;
}
@ -278,13 +268,11 @@ void vmFreeQueue(SVnodeMgmt *pMgmt, SVnodeObj *pVnode) {
tWWorkerFreeQueue(&pMgmt->syncPool, pVnode->pSyncQ);
tQWorkerFreeQueue(&pMgmt->queryPool, pVnode->pQueryQ);
tQWorkerFreeQueue(&pMgmt->fetchPool, pVnode->pFetchQ);
tWWorkerFreeQueue(&pMgmt->mergePool, pVnode->pMergeQ);
pVnode->pWriteQ = NULL;
pVnode->pSyncQ = NULL;
pVnode->pApplyQ = NULL;
pVnode->pQueryQ = NULL;
pVnode->pFetchQ = NULL;
pVnode->pMergeQ = NULL;
dDebug("vgId:%d, queue is freed", pVnode->vgId);
}
@ -316,11 +304,6 @@ int32_t vmStartWorker(SVnodeMgmt *pMgmt) {
pSPool->max = tsNumOfVnodeSyncThreads;
if (tWWorkerInit(pSPool) != 0) return -1;
SWWorkerPool *pMPool = &pMgmt->mergePool;
pMPool->name = "vnode-merge";
pMPool->max = tsNumOfVnodeMergeThreads;
if (tWWorkerInit(pMPool) != 0) return -1;
SSingleWorkerCfg mgmtCfg = {
.min = 1,
.max = 1,
@ -351,6 +334,5 @@ void vmStopWorker(SVnodeMgmt *pMgmt) {
tWWorkerCleanup(&pMgmt->syncPool);
tQWorkerCleanup(&pMgmt->queryPool);
tQWorkerCleanup(&pMgmt->fetchPool);
tWWorkerCleanup(&pMgmt->mergePool);
dDebug("vnode workers are closed");
}

4
source/dnode/mnode/impl/inc/mndDef.h
View File

@ -341,8 +341,8 @@ typedef struct {
int32_t colVer;
int32_t smaVer;
int32_t nextColId;
float xFilesFactor;
int32_t delay;
int64_t watermark[2];
int64_t maxdelay[2];
int32_t ttl;
int32_t numOfColumns;
int32_t numOfTags;

2
source/dnode/mnode/impl/inc/mndScheduler.h
View File

@ -30,7 +30,7 @@ int32_t mndSchedInitSubEp(SMnode* pMnode, const SMqTopicObj* pTopic, SMqSubscrib
int32_t mndScheduleStream(SMnode* pMnode, STrans* pTrans, SStreamObj* pStream);
int32_t mndConvertRsmaTask(char** pDst, int32_t* pDstLen, const char* ast, int64_t uid, int8_t triggerType,
int64_t watermark, double filesFactor);
int64_t watermark);
int32_t mndScheduleStream(SMnode* pMnode, STrans* pTrans, SStreamObj* pStream);

10
source/dnode/mnode/impl/src/mndMain.c
View File

@ -433,15 +433,19 @@ int32_t mndProcessSyncMsg(SRpcMsg *pMsg) {
return -1;
}
char logBuf[512] = {0};
do {
char *syncNodeStr = sync2SimpleStr(pMgmt->sync);
snprintf(logBuf, sizeof(logBuf), "==mndProcessSyncMsg== msgType:%d, syncNode: %s", pMsg->msgType, syncNodeStr);
static int64_t mndTick = 0;
if (++mndTick % 10 == 1) {
mTrace("sync trace msg:%s, %s", TMSG_INFO(pMsg->msgType), syncNodeStr);
mTrace("vgId:%d, sync heartbeat msg:%s, %s", syncGetVgId(pMgmt->sync), TMSG_INFO(pMsg->msgType), syncNodeStr);
}
if (gRaftDetailLog) {
char logBuf[512] = {0};
snprintf(logBuf, sizeof(logBuf), "==mndProcessSyncMsg== msgType:%d, syncNode: %s", pMsg->msgType, syncNodeStr);
syncRpcMsgLog2(logBuf, pMsg);
}
taosMemoryFree(syncNodeStr);
} while (0);
// ToDo: ugly! use function pointer
if (syncNodeSnapshotEnable(pSyncNode)) {

72
source/dnode/mnode/impl/src/mndScheduler.c
View File

@ -33,7 +33,7 @@
#include "tname.h"
#include "tuuid.h"
extern bool tsStreamSchedV;
extern bool tsSchedStreamToSnode;
static int32_t mndAddTaskToTaskSet(SArray* pArray, SStreamTask* pTask) {
int32_t childId = taosArrayGetSize(pArray);
@ -43,7 +43,7 @@ static int32_t mndAddTaskToTaskSet(SArray* pArray, SStreamTask* pTask) {
}
int32_t mndConvertRsmaTask(char** pDst, int32_t* pDstLen, const char* ast, int64_t uid, int8_t triggerType,
int64_t watermark, double filesFactor) {
int64_t watermark) {
SNode* pAst = NULL;
SQueryPlan* pPlan = NULL;
terrno = TSDB_CODE_SUCCESS;
@ -63,9 +63,8 @@ int32_t mndConvertRsmaTask(char** pDst, int32_t* pDstLen, const char* ast, int64
.topicQuery = false,
.streamQuery = true,
.rSmaQuery = true,
.triggerType = STREAM_TRIGGER_AT_ONCE,
.triggerType = triggerType,
.watermark = watermark,
/*.filesFactor = filesFactor,*/
};
if (qCreateQueryPlan(&cxt, &pPlan, NULL) < 0) {
@ -205,9 +204,11 @@ int32_t mndAssignTaskToVg(SMnode* pMnode, STrans* pTrans, SStreamTask* pTask, SS
return 0;
}
SSnodeObj* mndSchedFetchSnode(SMnode* pMnode) {
SSnodeObj* mndSchedFetchOneSnode(SMnode* pMnode) {
SSnodeObj* pObj = NULL;
pObj = sdbFetch(pMnode->pSdb, SDB_SNODE, NULL, (void**)&pObj);
void* pIter = NULL;
// TODO random fetch
pIter = sdbFetch(pMnode->pSdb, SDB_SNODE, pIter, (void**)&pObj);
return pObj;
}
@ -215,7 +216,7 @@ int32_t mndAssignTaskToSnode(SMnode* pMnode, STrans* pTrans, SStreamTask* pTask,
const SSnodeObj* pSnode) {
int32_t msgLen;
pTask->nodeId = 0;
pTask->nodeId = SNODE_HANDLE;
pTask->epSet = mndAcquireEpFromSnode(pMnode, pSnode);
plan->execNode.nodeId = 0;
@ -225,7 +226,7 @@ int32_t mndAssignTaskToSnode(SMnode* pMnode, STrans* pTrans, SStreamTask* pTask,
terrno = TSDB_CODE_QRY_INVALID_INPUT;
return -1;
}
mndPersistTaskDeployReq(pTrans, pTask, &plan->execNode.epSet, TDMT_STREAM_TASK_DEPLOY, 0);
mndPersistTaskDeployReq(pTrans, pTask, &plan->execNode.epSet, TDMT_STREAM_TASK_DEPLOY, SNODE_HANDLE);
return 0;
}
@ -270,7 +271,6 @@ int32_t mndAddShuffleSinkTasksToStream(SMnode* pMnode, STrans* pTrans, SStreamOb
pTask->epSet = mndGetVgroupEpset(pMnode, pVgroup);
// source
pTask->sourceType = TASK_SOURCE__MERGE;
pTask->inputType = TASK_INPUT_TYPE__DATA_BLOCK;
// exec
@ -316,7 +316,6 @@ int32_t mndAddFixedSinkTaskToStream(SMnode* pMnode, STrans* pTrans, SStreamObj*
#endif
pTask->epSet = mndGetVgroupEpset(pMnode, &pStream->fixedSinkVg);
// source
pTask->sourceType = TASK_SOURCE__MERGE;
pTask->inputType = TASK_INPUT_TYPE__DATA_BLOCK;
// exec
@ -373,8 +372,8 @@ int32_t mndScheduleStream(SMnode* pMnode, STrans* pTrans, SStreamObj* pStream) {
}
if (totLevel > 1) {
SStreamTask* pFinalTask;
// inner plan
SStreamTask* pInnerTask;
// inner level
{
SArray* taskInnerLevel = taosArrayInit(0, sizeof(void*));
taosArrayPush(pStream->tasks, &taskInnerLevel);
@ -383,31 +382,51 @@ int32_t mndScheduleStream(SMnode* pMnode, STrans* pTrans, SStreamObj* pStream) {
SSubplan* plan = (SSubplan*)nodesListGetNode(inner->pNodeList, 0);
ASSERT(plan->subplanType == SUBPLAN_TYPE_MERGE);
pFinalTask = tNewSStreamTask(pStream->uid);
mndAddTaskToTaskSet(taskInnerLevel, pFinalTask);
pInnerTask = tNewSStreamTask(pStream->uid);
mndAddTaskToTaskSet(taskInnerLevel, pInnerTask);
// input
pFinalTask->inputType = TASK_INPUT_TYPE__DATA_BLOCK;
pInnerTask->inputType = TASK_INPUT_TYPE__DATA_BLOCK;
// trigger
pFinalTask->triggerParam = pStream->triggerParam;
pInnerTask->triggerParam = pStream->triggerParam;
// dispatch
if (mndAddDispatcherToInnerTask(pMnode, pTrans, pStream, pFinalTask) < 0) {
if (mndAddDispatcherToInnerTask(pMnode, pTrans, pStream, pInnerTask) < 0) {
qDestroyQueryPlan(pPlan);
return -1;
}
// exec
pFinalTask->execType = TASK_EXEC__PIPE;
pInnerTask->execType = TASK_EXEC__PIPE;
if (tsSchedStreamToSnode) {
SSnodeObj* pSnode = mndSchedFetchOneSnode(pMnode);
if (pSnode == NULL) {
SVgObj* pVgroup = mndSchedFetchOneVg(pMnode, pStream->sourceDbUid);
if (mndAssignTaskToVg(pMnode, pTrans, pFinalTask, plan, pVgroup) < 0) {
if (mndAssignTaskToVg(pMnode, pTrans, pInnerTask, plan, pVgroup) < 0) {
sdbRelease(pSdb, pVgroup);
qDestroyQueryPlan(pPlan);
return -1;
}
} else {
if (mndAssignTaskToSnode(pMnode, pTrans, pInnerTask, plan, pSnode) < 0) {
ASSERT(0);
sdbRelease(pSdb, pSnode);
qDestroyQueryPlan(pPlan);
return -1;
}
}
} else {
SVgObj* pVgroup = mndSchedFetchOneVg(pMnode, pStream->sourceDbUid);
if (mndAssignTaskToVg(pMnode, pTrans, pInnerTask, plan, pVgroup) < 0) {
sdbRelease(pSdb, pVgroup);
qDestroyQueryPlan(pPlan);
return -1;
}
}
}
// source plan
// source level
SArray* taskSourceLevel = taosArrayInit(0, sizeof(void*));
taosArrayPush(pStream->tasks, &taskSourceLevel);
@ -427,6 +446,8 @@ int32_t mndScheduleStream(SMnode* pMnode, STrans* pTrans, SStreamObj* pStream) {
SStreamTask* pTask = tNewSStreamTask(pStream->uid);
mndAddTaskToTaskSet(taskSourceLevel, pTask);
pTask->dataScan = 1;
// input
pTask->inputType = TASK_INPUT_TYPE__SUMBIT_BLOCK;
@ -435,9 +456,9 @@ int32_t mndScheduleStream(SMnode* pMnode, STrans* pTrans, SStreamObj* pStream) {
pTask->dispatchMsgType = TDMT_STREAM_TASK_DISPATCH;
pTask->dispatchType = TASK_DISPATCH__FIXED;
pTask->fixedEpDispatcher.taskId = pFinalTask->taskId;
pTask->fixedEpDispatcher.nodeId = pFinalTask->nodeId;
pTask->fixedEpDispatcher.epSet = pFinalTask->epSet;
pTask->fixedEpDispatcher.taskId = pInnerTask->taskId;
pTask->fixedEpDispatcher.nodeId = pInnerTask->nodeId;
pTask->fixedEpDispatcher.epSet = pInnerTask->epSet;
// exec
pTask->execType = TASK_EXEC__PIPE;
@ -470,9 +491,14 @@ int32_t mndScheduleStream(SMnode* pMnode, STrans* pTrans, SStreamObj* pStream) {
SStreamTask* pTask = tNewSStreamTask(pStream->uid);
mndAddTaskToTaskSet(taskOneLevel, pTask);
pTask->dataScan = 1;
// input
pTask->inputType = TASK_INPUT_TYPE__SUMBIT_BLOCK;
// trigger
pTask->triggerParam = pStream->triggerParam;
// sink or dispatch
if (hasExtraSink) {
mndAddDispatcherToInnerTask(pMnode, pTrans, pStream, pTask);

1
source/dnode/mnode/impl/src/mndSnode.c
View File

@ -408,6 +408,7 @@ static int32_t mndProcessDropSnodeReq(SRpcMsg *pReq) {
goto _OVER;
}
// check deletable
code = mndDropSnode(pMnode, pReq, pObj);
if (code == 0) code = TSDB_CODE_ACTION_IN_PROGRESS;

54
source/dnode/mnode/impl/src/mndStb.c
View File

@ -89,8 +89,10 @@ SSdbRaw *mndStbActionEncode(SStbObj *pStb) {
SDB_SET_INT32(pRaw, dataPos, pStb->tagVer, _OVER)
SDB_SET_INT32(pRaw, dataPos, pStb->colVer, _OVER)
SDB_SET_INT32(pRaw, dataPos, pStb->nextColId, _OVER)
SDB_SET_INT32(pRaw, dataPos, (int32_t)(pStb->xFilesFactor * 10000), _OVER)
SDB_SET_INT32(pRaw, dataPos, pStb->delay, _OVER)
SDB_SET_INT64(pRaw, dataPos, pStb->maxdelay[0], _OVER)
SDB_SET_INT64(pRaw, dataPos, pStb->maxdelay[1], _OVER)
SDB_SET_INT64(pRaw, dataPos, pStb->watermark[0], _OVER)
SDB_SET_INT64(pRaw, dataPos, pStb->watermark[1], _OVER)
SDB_SET_INT32(pRaw, dataPos, pStb->ttl, _OVER)
SDB_SET_INT32(pRaw, dataPos, pStb->numOfColumns, _OVER)
SDB_SET_INT32(pRaw, dataPos, pStb->numOfTags, _OVER)
@ -168,10 +170,10 @@ static SSdbRow *mndStbActionDecode(SSdbRaw *pRaw) {
SDB_GET_INT32(pRaw, dataPos, &pStb->tagVer, _OVER)
SDB_GET_INT32(pRaw, dataPos, &pStb->colVer, _OVER)
SDB_GET_INT32(pRaw, dataPos, &pStb->nextColId, _OVER)
int32_t xFilesFactor = 0;
SDB_GET_INT32(pRaw, dataPos, &xFilesFactor, _OVER)
pStb->xFilesFactor = xFilesFactor / 10000.0f;
SDB_GET_INT32(pRaw, dataPos, &pStb->delay, _OVER)
SDB_GET_INT64(pRaw, dataPos, &pStb->maxdelay[0], _OVER)
SDB_GET_INT64(pRaw, dataPos, &pStb->maxdelay[1], _OVER)
SDB_GET_INT64(pRaw, dataPos, &pStb->watermark[0], _OVER)
SDB_GET_INT64(pRaw, dataPos, &pStb->watermark[1], _OVER)
SDB_GET_INT32(pRaw, dataPos, &pStb->ttl, _OVER)
SDB_GET_INT32(pRaw, dataPos, &pStb->numOfColumns, _OVER)
SDB_GET_INT32(pRaw, dataPos, &pStb->numOfTags, _OVER)
@ -399,18 +401,18 @@ static void *mndBuildVCreateStbReq(SMnode *pMnode, SVgObj *pVgroup, SStbObj *pSt
req.schemaTag.pSchema = pStb->pTags;
if (req.rollup) {
req.pRSmaParam.xFilesFactor = pStb->xFilesFactor;
req.pRSmaParam.delay = pStb->delay;
req.pRSmaParam.maxdelay[0] = pStb->maxdelay[0];
req.pRSmaParam.maxdelay[1] = pStb->maxdelay[1];
if (pStb->ast1Len > 0) {
if (mndConvertRsmaTask(&req.pRSmaParam.qmsg1, &req.pRSmaParam.qmsg1Len, pStb->pAst1, pStb->uid,
STREAM_TRIGGER_AT_ONCE, 0, req.pRSmaParam.xFilesFactor) != TSDB_CODE_SUCCESS) {
return NULL;
if (mndConvertRsmaTask(&req.pRSmaParam.qmsg[0], &req.pRSmaParam.qmsgLen[0], pStb->pAst1, pStb->uid,
STREAM_TRIGGER_WINDOW_CLOSE, req.pRSmaParam.watermark[0]) < 0) {
goto _err;
}
}
if (pStb->ast2Len > 0) {
if (mndConvertRsmaTask(&req.pRSmaParam.qmsg2, &req.pRSmaParam.qmsg2Len, pStb->pAst2, pStb->uid,
STREAM_TRIGGER_AT_ONCE, 0, req.pRSmaParam.xFilesFactor) != TSDB_CODE_SUCCESS) {
return NULL;
if (mndConvertRsmaTask(&req.pRSmaParam.qmsg[1], &req.pRSmaParam.qmsgLen[1], pStb->pAst2, pStb->uid,
STREAM_TRIGGER_WINDOW_CLOSE, req.pRSmaParam.watermark[1]) < 0) {
goto _err;
}
}
}
@ -418,17 +420,15 @@ static void *mndBuildVCreateStbReq(SMnode *pMnode, SVgObj *pVgroup, SStbObj *pSt
int32_t ret = 0;
tEncodeSize(tEncodeSVCreateStbReq, &req, contLen, ret);
if (ret < 0) {
return NULL;
goto _err;
}
contLen += sizeof(SMsgHead);
SMsgHead *pHead = taosMemoryMalloc(contLen);
if (pHead == NULL) {
taosMemoryFreeClear(req.pRSmaParam.qmsg1);
taosMemoryFreeClear(req.pRSmaParam.qmsg2);
terrno = TSDB_CODE_OUT_OF_MEMORY;
return NULL;
goto _err;
}
pHead->contLen = htonl(contLen);
@ -438,17 +438,19 @@ static void *mndBuildVCreateStbReq(SMnode *pMnode, SVgObj *pVgroup, SStbObj *pSt
tEncoderInit(&encoder, pBuf, contLen - sizeof(SMsgHead));
if (tEncodeSVCreateStbReq(&encoder, &req) < 0) {
taosMemoryFreeClear(pHead);
taosMemoryFreeClear(req.pRSmaParam.qmsg1);
taosMemoryFreeClear(req.pRSmaParam.qmsg2);
tEncoderClear(&encoder);
return NULL;
goto _err;
}
tEncoderClear(&encoder);
*pContLen = contLen;
taosMemoryFreeClear(req.pRSmaParam.qmsg1);
taosMemoryFreeClear(req.pRSmaParam.qmsg2);
taosMemoryFreeClear(req.pRSmaParam.qmsg[0]);
taosMemoryFreeClear(req.pRSmaParam.qmsg[1]);
return pHead;
_err:
taosMemoryFreeClear(req.pRSmaParam.qmsg[0]);
taosMemoryFreeClear(req.pRSmaParam.qmsg[1]);
return NULL;
}
static void *mndBuildVDropStbReq(SMnode *pMnode, SVgObj *pVgroup, SStbObj *pStb, int32_t *pContLen) {
@ -670,8 +672,10 @@ int32_t mndBuildStbFromReq(SMnode *pMnode, SStbObj *pDst, SMCreateStbReq *pCreat
pDst->tagVer = 1;
pDst->colVer = 1;
pDst->nextColId = 1;
// pDst->xFilesFactor = pCreate->xFilesFactor;
// pDst->delay = pCreate->delay;
pDst->maxdelay[0] = pCreate->delay1;
pDst->maxdelay[1] = pCreate->delay2;
pDst->watermark[0] = pCreate->watermark1;
pDst->watermark[1] = pCreate->watermark2;
pDst->ttl = pCreate->ttl;
pDst->numOfColumns = pCreate->numOfColumns;
pDst->numOfTags = pCreate->numOfTags;

20
source/dnode/mnode/impl/src/mndStream.c
View File

@ -518,7 +518,6 @@ static int32_t mndCreateStream(SMnode *pMnode, SRpcMsg *pReq, SCMCreateStreamReq
// TODO
streamObj.fixedSinkVgId = 0;
streamObj.smaId = 0;
/*streamObj.physicalPlan = "";*/
streamObj.trigger = pCreate->triggerType;
streamObj.watermark = pCreate->watermark;
streamObj.triggerParam = pCreate->maxDelay;
@ -607,17 +606,6 @@ static int32_t mndProcessCreateStreamReq(SRpcMsg *pReq) {
}
#endif
STrans *pTrans = mndTransCreate(pMnode, TRN_POLICY_ROLLBACK, TRN_CONFLICT_DB_INSIDE, pReq);
if (pTrans == NULL) {
mError("stream:%s, failed to create since %s", createStreamReq.name, terrstr());
goto _OVER;
}
mndTransSetDbName(pTrans, createStreamReq.sourceDB, NULL);
// TODO
/*mndTransSetDbName(pTrans, streamObj.targetDb, NULL);*/
mDebug("trans:%d, used to create stream:%s", pTrans->id, createStreamReq.name);
// build stream obj from request
SStreamObj streamObj = {0};
if (mndBuildStreamObjFromCreateReq(pMnode, &streamObj, &createStreamReq) < 0) {
@ -626,6 +614,14 @@ static int32_t mndProcessCreateStreamReq(SRpcMsg *pReq) {
goto _OVER;
}
STrans *pTrans = mndTransCreate(pMnode, TRN_POLICY_ROLLBACK, TRN_CONFLICT_DB_INSIDE, pReq);
if (pTrans == NULL) {
mError("stream:%s, failed to create since %s", createStreamReq.name, terrstr());
goto _OVER;
}
mndTransSetDbName(pTrans, createStreamReq.sourceDB, streamObj.targetDb);
mDebug("trans:%d, used to create stream:%s", pTrans->id, createStreamReq.name);
// create stb for stream
if (mndCreateStbForStream(pMnode, pTrans, &streamObj, pReq->info.conn.user) < 0) {
mError("trans:%d, failed to create stb for stream %s since %s", pTrans->id, createStreamReq.name, terrstr());

3
source/dnode/mnode/impl/src/mndSync.c
View File

@ -68,8 +68,9 @@ void mndSyncCommitMsg(struct SSyncFSM *pFsm, const SRpcMsg *pMsg, SFsmCbMeta cbM
mndTransExecute(pMnode, pTrans);
mndReleaseTrans(pMnode, pTrans);
}
#if 0
sdbWriteFile(pMnode->pSdb, SDB_WRITE_DELTA);
#endif
}
}

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

@ -804,7 +804,7 @@ static void mndTransSendRpcRsp(SMnode *pMnode, STrans *pTrans) {
sendRsp = true;
}
} else {
if (pTrans->stage == TRN_STAGE_REDO_ACTION && pTrans->failedTimes > 2) {
if (pTrans->stage == TRN_STAGE_REDO_ACTION && pTrans->failedTimes > 3) {
if (code == 0) code = TSDB_CODE_MND_TRANS_UNKNOW_ERROR;
sendRsp = true;
}
@ -896,7 +896,7 @@ static void mndTransResetActions(SMnode *pMnode, STrans *pTrans, SArray *pArray)
pAction->rawWritten = 0;
pAction->msgSent = 0;
pAction->msgReceived = 0;
if (pAction->errCode == TSDB_CODE_RPC_REDIRECT || pAction->errCode == TSDB_CODE_SYN_NOT_IN_NEW_CONFIG ||
if (pAction->errCode == TSDB_CODE_RPC_REDIRECT || pAction->errCode == TSDB_CODE_SYN_NEW_CONFIG_ERROR ||
pAction->errCode == TSDB_CODE_SYN_INTERNAL_ERROR || pAction->errCode == TSDB_CODE_SYN_NOT_LEADER) {
pAction->epSet.inUse = (pAction->epSet.inUse + 1) % pAction->epSet.numOfEps;
mDebug("trans:%d, %s:%d execute status is reset and set epset inuse:%d", pTrans->id, mndTransStr(pAction->stage),
@ -1127,6 +1127,7 @@ static int32_t mndTransExecuteRedoActionsSerial(SMnode *pMnode, STrans *pTrans)
}
if (code == 0) {
pTrans->failedTimes = 0;
pTrans->lastAction = action;
pTrans->lastMsgType = 0;
pTrans->lastErrorNo = 0;
@ -1430,8 +1431,7 @@ void mndTransPullup(SMnode *pMnode) {
mndReleaseTrans(pMnode, pTrans);
}
// todo, set to SDB_WRITE_DELTA
sdbWriteFile(pMnode->pSdb, 0);
sdbWriteFile(pMnode->pSdb, SDB_WRITE_DELTA);
taosArrayDestroy(pArray);
}

2
source/dnode/mnode/sdb/inc/sdb.h
View File

@ -37,7 +37,7 @@ extern "C" {
#define mTrace(...) { if (mDebugFlag & DEBUG_TRACE) { taosPrintLog("MND ", DEBUG_TRACE, mDebugFlag, __VA_ARGS__); }}
// clang-format on
#define SDB_WRITE_DELTA 100
#define SDB_WRITE_DELTA 20
#define SDB_GET_VAL(pData, dataPos, val, pos, func, type) \
{ \

6
source/dnode/mnode/sdb/src/sdb.c
View File

@ -161,9 +161,11 @@ static int32_t sdbCreateDir(SSdb *pSdb) {
}
void sdbSetApplyInfo(SSdb *pSdb, int64_t index, int64_t term, int64_t config) {
mTrace("mnode apply info changed, from index:%" PRId64 " term:%" PRId64 " config:%" PRId64 ", to index:%" PRId64
#if 1
mTrace("mnode apply info changed from index:%" PRId64 " term:%" PRId64 " config:%" PRId64 " to index:%" PRId64
" term:%" PRId64 " config:%" PRId64,
pSdb->applyIndex, pSdb->applyTerm, pSdb->applyConfig, index, term, config);
#endif
pSdb->applyIndex = index;
pSdb->applyTerm = term;
pSdb->applyConfig = config;
@ -173,7 +175,9 @@ void sdbGetCommitInfo(SSdb *pSdb, int64_t *index, int64_t *term, int64_t *config
*index = pSdb->commitIndex;
*term = pSdb->commitTerm;
*config = pSdb->commitConfig;
#if 0
mTrace("mnode current info, apply index:%" PRId64 " term:%" PRId64 " config:%" PRId64 ", commit index:%" PRId64
" term:%" PRId64 " config:%" PRId64,
pSdb->applyIndex, pSdb->applyTerm, pSdb->applyConfig, *index, *term, *config);
#endif
}

1
source/dnode/snode/inc/sndInt.h
View File

@ -56,7 +56,6 @@ SStreamTask* sndMetaGetTask(SStreamMeta* pMeta, int32_t taskId);
int32_t sndMetaRemoveTask(SStreamMeta* pMeta, int32_t taskId);
int32_t sndDropTaskOfStream(SStreamMeta* pMeta, int64_t streamId);
int32_t sndStopTaskOfStream(SStreamMeta* pMeta, int64_t streamId);
int32_t sndResumeTaskOfStream(SStreamMeta* pMeta, int64_t streamId);

172
source/dnode/snode/src/snode.c
View File

@ -76,45 +76,157 @@ int32_t sndMetaRemoveTask(SStreamMeta *pMeta, int32_t taskId) {
return taosHashRemove(pMeta->pHash, &taskId, sizeof(int32_t));
}
static int32_t sndProcessTaskExecReq(SSnode *pSnode, SRpcMsg *pMsg) {
/*SStreamExecMsgHead *pHead = pMsg->pCont;*/
/*int32_t taskId = pHead->streamTaskId;*/
/*SStreamTask *pTask = sndMetaGetTask(pSnode->pMeta, taskId);*/
/*if (pTask == NULL) {*/
/*return -1;*/
/*}*/
static int32_t sndProcessTaskDeployReq(SSnode *pNode, SRpcMsg *pMsg) {
SStreamMeta *pMeta = pNode->pMeta;
char *msg = POINTER_SHIFT(pMsg->pCont, sizeof(SMsgHead));
int32_t msgLen = pMsg->contLen - sizeof(SMsgHead);
SStreamTask *pTask = taosMemoryCalloc(1, sizeof(SStreamTask));
if (pTask == NULL) {
return -1;
}
SDecoder decoder;
tDecoderInit(&decoder, (uint8_t *)msg, msgLen);
if (tDecodeSStreamTask(&decoder, pTask) < 0) {
ASSERT(0);
}
tDecoderClear(&decoder);
pTask->status = TASK_STATUS__IDLE;
pTask->inputQueue = streamQueueOpen();
pTask->outputQueue = streamQueueOpen();
pTask->inputStatus = TASK_INPUT_STATUS__NORMAL;
pTask->outputStatus = TASK_INPUT_STATUS__NORMAL;
if (pTask->inputQueue == NULL || pTask->outputQueue == NULL) goto FAIL;
pTask->pMsgCb = &pNode->msgCb;
ASSERT(pTask->execType != TASK_EXEC__NONE);
ASSERT(pTask->dataScan == 0);
pTask->exec.executor = qCreateStreamExecTaskInfo(pTask->exec.qmsg, NULL);
ASSERT(pTask->exec.executor);
streamSetupTrigger(pTask);
qInfo("deploy stream: stream id %ld task id %d child id %d on snode", pTask->streamId, pTask->taskId, pTask->childId);
taosHashPut(pMeta->pHash, &pTask->taskId, sizeof(int32_t), &pTask, sizeof(void *));
return 0;
FAIL:
if (pTask->inputQueue) streamQueueClose(pTask->inputQueue);
if (pTask->outputQueue) streamQueueClose(pTask->outputQueue);
if (pTask) taosMemoryFree(pTask);
return -1;
}
static int32_t sndProcessTaskRunReq(SSnode *pNode, SRpcMsg *pMsg) {
SStreamMeta *pMeta = pNode->pMeta;
SStreamTaskRunReq *pReq = pMsg->pCont;
int32_t taskId = pReq->taskId;
SStreamTask *pTask = *(SStreamTask **)taosHashGet(pMeta->pHash, &taskId, sizeof(int32_t));
streamProcessRunReq(pTask);
return 0;
}
void sndProcessUMsg(SSnode *pSnode, SRpcMsg *pMsg) {
static int32_t sndProcessTaskDispatchReq(SSnode *pNode, SRpcMsg *pMsg) {
SStreamMeta *pMeta = pNode->pMeta;
char *msgStr = pMsg->pCont;
char *msgBody = POINTER_SHIFT(msgStr, sizeof(SMsgHead));
int32_t msgLen = pMsg->contLen - sizeof(SMsgHead);
SStreamDispatchReq req;
SDecoder decoder;
tDecoderInit(&decoder, msgBody, msgLen);
tDecodeStreamDispatchReq(&decoder, &req);
int32_t taskId = req.taskId;
SStreamTask *pTask = *(SStreamTask **)taosHashGet(pMeta->pHash, &taskId, sizeof(int32_t));
SRpcMsg rsp = {
.info = pMsg->info,
.code = 0,
};
streamProcessDispatchReq(pTask, &req, &rsp);
return 0;
}
static int32_t sndProcessTaskRecoverReq(SSnode *pNode, SRpcMsg *pMsg) {
SStreamMeta *pMeta = pNode->pMeta;
SStreamTaskRecoverReq *pReq = pMsg->pCont;
int32_t taskId = pReq->taskId;
SStreamTask *pTask = *(SStreamTask **)taosHashGet(pMeta->pHash, &taskId, sizeof(int32_t));
streamProcessRecoverReq(pTask, pReq, pMsg);
return 0;
}
static int32_t sndProcessTaskDispatchRsp(SSnode *pNode, SRpcMsg *pMsg) {
SStreamMeta *pMeta = pNode->pMeta;
SStreamDispatchRsp *pRsp = POINTER_SHIFT(pMsg->pCont, sizeof(SMsgHead));
int32_t taskId = pRsp->taskId;
SStreamTask *pTask = *(SStreamTask **)taosHashGet(pMeta->pHash, &taskId, sizeof(int32_t));
streamProcessDispatchRsp(pTask, pRsp);
return 0;
}
static int32_t sndProcessTaskRecoverRsp(SSnode *pNode, SRpcMsg *pMsg) {
SStreamMeta *pMeta = pNode->pMeta;
SStreamTaskRecoverRsp *pRsp = pMsg->pCont;
int32_t taskId = pRsp->taskId;
SStreamTask *pTask = *(SStreamTask **)taosHashGet(pMeta->pHash, &taskId, sizeof(int32_t));
streamProcessRecoverRsp(pTask, pRsp);
return 0;
}
static int32_t sndProcessTaskDropReq(SSnode *pNode, SRpcMsg *pMsg) {
SStreamMeta *pMeta = pNode->pMeta;
char *msg = pMsg->pCont;
int32_t msgLen = pMsg->contLen;
SVDropStreamTaskReq *pReq = (SVDropStreamTaskReq *)msg;
int32_t code = taosHashRemove(pMeta->pHash, &pReq->taskId, sizeof(int32_t));
ASSERT(code == 0);
if (code == 0) {
// sendrsp
}
return code;
}
int32_t sndProcessUMsg(SSnode *pSnode, SRpcMsg *pMsg) {
// stream deploy
// stream stop/resume
// operator exec
if (pMsg->msgType == TDMT_STREAM_TASK_DEPLOY) {
void *msg = POINTER_SHIFT(pMsg->pCont, sizeof(SMsgHead));
SStreamTask *pTask = taosMemoryMalloc(sizeof(SStreamTask));
if (pTask == NULL) {
ASSERT(0);
return;
}
SDecoder decoder;
tDecoderInit(&decoder, msg, pMsg->contLen - sizeof(SMsgHead));
tDecodeSStreamTask(&decoder, pTask);
tDecoderClear(&decoder);
sndMetaDeployTask(pSnode->pMeta, pTask);
/*} else if (pMsg->msgType == TDMT_SND_TASK_EXEC) {*/
/*sndProcessTaskExecReq(pSnode, pMsg);*/
} else {
switch (pMsg->msgType) {
case TDMT_STREAM_TASK_DEPLOY:
return sndProcessTaskDeployReq(pSnode, pMsg);
case TDMT_VND_STREAM_TASK_DROP:
return sndProcessTaskDropReq(pSnode, pMsg);
default:
ASSERT(0);
}
return 0;
}
void sndProcessSMsg(SSnode *pSnode, SRpcMsg *pMsg) {
// operator exec
/*if (pMsg->msgType == TDMT_SND_TASK_EXEC) {*/
/*sndProcessTaskExecReq(pSnode, pMsg);*/
/*} else {*/
int32_t sndProcessSMsg(SSnode *pSnode, SRpcMsg *pMsg) {
switch (pMsg->msgType) {
case TDMT_STREAM_TASK_RUN:
return sndProcessTaskRunReq(pSnode, pMsg);
case TDMT_STREAM_TASK_DISPATCH:
return sndProcessTaskDispatchReq(pSnode, pMsg);
case TDMT_STREAM_TASK_RECOVER:
return sndProcessTaskRecoverReq(pSnode, pMsg);
case TDMT_STREAM_TASK_DISPATCH_RSP:
return sndProcessTaskDispatchRsp(pSnode, pMsg);
case TDMT_STREAM_TASK_RECOVER_RSP:
return sndProcessTaskRecoverRsp(pSnode, pMsg);
default:
ASSERT(0);
/*}*/
}
return 0;
}

1
source/dnode/vnode/src/inc/sma.h
View File

@ -37,6 +37,7 @@ typedef struct SSmaStat SSmaStat;
typedef struct SSmaStatItem SSmaStatItem;
typedef struct SSmaKey SSmaKey;
typedef struct SRSmaInfo SRSmaInfo;
typedef struct SRSmaInfoItem SRSmaInfoItem;
struct SSmaEnv {
TdThreadRwlock lock;

2
source/dnode/vnode/src/inc/vnodeInt.h
View File

@ -141,7 +141,7 @@ int32_t tqProcessVgChangeReq(STQ* pTq, char* msg, int32_t msgLen);
int32_t tqProcessVgDeleteReq(STQ* pTq, char* msg, int32_t msgLen);
int32_t tqProcessOffsetCommitReq(STQ* pTq, char* msg, int32_t msgLen);
int32_t tqProcessPollReq(STQ* pTq, SRpcMsg* pMsg, int32_t workerId);
int32_t tqProcessTaskDeploy(STQ* pTq, char* msg, int32_t msgLen);
int32_t tqProcessTaskDeployReq(STQ* pTq, char* msg, int32_t msgLen);
int32_t tqProcessTaskDropReq(STQ* pTq, char* msg, int32_t msgLen);
int32_t tqProcessStreamTrigger(STQ* pTq, SSubmitReq* data);
int32_t tqProcessTaskRunReq(STQ* pTq, SRpcMsg* pMsg);

5
source/dnode/vnode/src/sma/sma.c
View File

@ -15,6 +15,8 @@
#include "sma.h"
// functions for external invocation
// TODO: Who is responsible for resource allocate and release?
int32_t tdProcessTSmaInsert(SSma* pSma, int64_t indexUid, const char* msg) {
int32_t code = TSDB_CODE_SUCCESS;
@ -45,6 +47,9 @@ int32_t smaGetTSmaDays(SVnodeCfg* pCfg, void* pCont, uint32_t contLen, int32_t*
return code;
}
// functions for internal invocation
#if 0
/**

32
source/dnode/vnode/src/sma/smaEnv.c
View File

@ -208,7 +208,6 @@ int32_t tdUnLockSma(SSma *pSma) {
int32_t tdCheckAndInitSmaEnv(SSma *pSma, int8_t smaType) {
SSmaEnv *pEnv = NULL;
// return if already init
switch (smaType) {
case TSDB_SMA_TYPE_TIME_RANGE:
if ((pEnv = (SSmaEnv *)atomic_load_ptr(&SMA_TSMA_ENV(pSma)))) {
@ -244,3 +243,34 @@ int32_t tdCheckAndInitSmaEnv(SSma *pSma, int8_t smaType) {
return TSDB_CODE_SUCCESS;
};
int32_t smaTimerInit(void **timer, int8_t *initFlag, const char *label) {
int8_t old;
while (1) {
old = atomic_val_compare_exchange_8(initFlag, 0, 2);
if (old != 2) break;
}
if (old == 0) {
*timer = taosTmrInit(10000, 100, 10000, label);
if (!(*timer)) {
atomic_store_8(initFlag, 0);
return -1;
}
atomic_store_8(initFlag, 1);
}
return 0;
}
void smaTimerCleanUp(void *timer, int8_t *initFlag) {
int8_t old;
while (1) {
old = atomic_val_compare_exchange_8(initFlag, 1, 2);
if (old != 2) break;
}
if (old == 1) {
taosTmrCleanUp(timer);
atomic_store_8(initFlag, 0);
}
}

216
source/dnode/vnode/src/sma/smaRollup.c
View File

@ -14,14 +14,61 @@
*/
#include "sma.h"
#include "tstream.h"
static FORCE_INLINE int32_t tdUidStorePut(STbUidStore *pStore, tb_uid_t suid, tb_uid_t *uid);
static FORCE_INLINE int32_t tdUpdateTbUidListImpl(SSma *pSma, tb_uid_t *suid, SArray *tbUids);
static FORCE_INLINE int32_t tdExecuteRSmaImpl(SSma *pSma, const void *pMsg, int32_t inputType, qTaskInfo_t *taskInfo,
STSchema *pTSchema, tb_uid_t suid, int8_t level);
static FORCE_INLINE int32_t tdExecuteRSmaImpl(SSma *pSma, const void *pMsg, int32_t inputType, SRSmaInfoItem *rsmaItem,
tb_uid_t suid, int8_t level);
#define SET_RSMA_INFO_ITEM_PARAMS(__idx, __level) \
if (param->qmsg[__idx]) { \
pRSmaInfo->items[__idx].pRsmaInfo = pRSmaInfo; \
pRSmaInfo->items[__idx].taskInfo = qCreateStreamExecTaskInfo(param->qmsg[0], &handle); \
if (!pRSmaInfo->items[__idx].taskInfo) { \
goto _err; \
} \
pRSmaInfo->items[__idx].triggerStatus = TASK_TRIGGER_STATUS__IN_ACTIVE; \
if (param->maxdelay[__idx] < 1) { \
int64_t msInterval = \
convertTimeFromPrecisionToUnit(pRetention[__level].freq, pTsdbCfg->precision, TIME_UNIT_MILLISECOND); \
pRSmaInfo->items[__idx].maxDelay = msInterval; \
} else { \
pRSmaInfo->items[__idx].maxDelay = param->maxdelay[__idx]; \
} \
if (pRSmaInfo->items[__idx].maxDelay > TSDB_MAX_ROLLUP_MAX_DELAY) { \
pRSmaInfo->items[__idx].maxDelay = TSDB_MAX_ROLLUP_MAX_DELAY; \
} \
pRSmaInfo->items[__idx].level = TSDB_RETENTION_L##__level; \
pRSmaInfo->items[__idx].tmrHandle = taosTmrInit(10000, 100, 10000, "RSMA"); \
if (!pRSmaInfo->items[__idx].tmrHandle) { \
goto _err; \
} \
}
struct SRSmaInfoItem {
SRSmaInfo *pRsmaInfo;
void *taskInfo; // qTaskInfo_t
void *tmrHandle;
tmr_h tmrId;
int8_t level;
int8_t tmrInitFlag;
int8_t triggerStatus; // TASK_TRIGGER_STATUS__IN_ACTIVE/TASK_TRIGGER_STATUS__ACTIVE
int32_t maxDelay;
};
typedef struct {
int64_t suid;
SRSmaInfoItem *pItem;
SSma *pSma;
STSchema *pTSchema;
} SRSmaTriggerParam;
struct SRSmaInfo {
void *taskInfo[TSDB_RETENTION_L2]; // qTaskInfo_t
STSchema *pTSchema;
SSma *pSma;
int64_t suid;
SRSmaInfoItem items[TSDB_RETENTION_L2];
};
static FORCE_INLINE void tdFreeTaskHandle(qTaskInfo_t *taskHandle) {
@ -33,11 +80,20 @@ static FORCE_INLINE void tdFreeTaskHandle(qTaskInfo_t *taskHandle) {
}
void *tdFreeRSmaInfo(SRSmaInfo *pInfo) {
if (pInfo) {
for (int32_t i = 0; i < TSDB_RETENTION_MAX; ++i) {
if (pInfo->taskInfo[i]) {
tdFreeTaskHandle(pInfo->taskInfo[i]);
SRSmaInfoItem *pItem = &pInfo->items[i];
if (pItem->taskInfo) {
tdFreeTaskHandle(pItem->taskInfo);
}
if (pItem->tmrHandle) {
taosTmrCleanUp(pItem->tmrHandle);
}
}
taosMemoryFree(pInfo->pTSchema);
taosMemoryFree(pInfo);
}
return NULL;
}
@ -69,20 +125,20 @@ static FORCE_INLINE int32_t tdUpdateTbUidListImpl(SSma *pSma, tb_uid_t *suid, SA
return TSDB_CODE_FAILED;
}
if (pRSmaInfo->taskInfo[0] && (qUpdateQualifiedTableId(pRSmaInfo->taskInfo[0], tbUids, true) != 0)) {
if (pRSmaInfo->items[0].taskInfo && (qUpdateQualifiedTableId(pRSmaInfo->items[0].taskInfo, tbUids, true) < 0)) {
smaError("vgId:%d, update tbUidList failed for uid:%" PRIi64 " since %s", SMA_VID(pSma), *suid, terrstr(terrno));
return TSDB_CODE_FAILED;
} else {
smaDebug("vgId:%d, update tbUidList succeed for qTaskInfo:%p with suid:%" PRIi64 ", uid:%" PRIi64, SMA_VID(pSma),
pRSmaInfo->taskInfo[0], *suid, *(int64_t *)taosArrayGet(tbUids, 0));
pRSmaInfo->items[0].taskInfo, *suid, *(int64_t *)taosArrayGet(tbUids, 0));
}
if (pRSmaInfo->taskInfo[1] && (qUpdateQualifiedTableId(pRSmaInfo->taskInfo[1], tbUids, true) != 0)) {
if (pRSmaInfo->items[1].taskInfo && (qUpdateQualifiedTableId(pRSmaInfo->items[1].taskInfo, tbUids, true) < 0)) {
smaError("vgId:%d, update tbUidList failed for uid:%" PRIi64 " since %s", SMA_VID(pSma), *suid, terrstr(terrno));
return TSDB_CODE_FAILED;
} else {
smaDebug("vgId:%d, update tbUidList succeed for qTaskInfo:%p with suid:%" PRIi64 ", uid:%" PRIi64, SMA_VID(pSma),
pRSmaInfo->taskInfo[1], *suid, *(int64_t *)taosArrayGet(tbUids, 0));
pRSmaInfo->items[1].taskInfo, *suid, *(int64_t *)taosArrayGet(tbUids, 0));
}
return TSDB_CODE_SUCCESS;
@ -144,12 +200,12 @@ int32_t tdFetchTbUidList(SSma *pSma, STbUidStore **ppStore, tb_uid_t suid, tb_ui
ASSERT(ppStore != NULL);
if (!(*ppStore)) {
if (tdUidStoreInit(ppStore) != 0) {
if (tdUidStoreInit(ppStore) < 0) {
return TSDB_CODE_FAILED;
}
}
if (tdUidStorePut(*ppStore, suid, &uid) != 0) {
if (tdUidStorePut(*ppStore, suid, &uid) < 0) {
*ppStore = tdUidStoreFree(*ppStore);
return TSDB_CODE_FAILED;
}
@ -176,7 +232,7 @@ int32_t tdProcessRSmaCreate(SVnode *pVnode, SVCreateStbReq *pReq) {
SMsgCb *pMsgCb = &pVnode->msgCb;
SRSmaParam *param = &pReq->pRSmaParam;
if ((param->qmsg1Len == 0) && (param->qmsg2Len == 0)) {
if ((param->qmsgLen[0] == 0) && (param->qmsgLen[1] == 0)) {
smaWarn("vgId:%d, no qmsg1/qmsg2 for rollup stable %s %" PRIi64, SMA_VID(pSma), pReq->name, pReq->suid);
return TSDB_CODE_SUCCESS;
}
@ -192,10 +248,12 @@ int32_t tdProcessRSmaCreate(SVnode *pVnode, SVCreateStbReq *pReq) {
pRSmaInfo = taosHashGet(SMA_STAT_INFO_HASH(pStat), &pReq->suid, sizeof(tb_uid_t));
if (pRSmaInfo) {
ASSERT(0); // TODO: free original pRSmaInfo is exists abnormally
smaWarn("vgId:%d, rsma info already exists for stb: %s, %" PRIi64, SMA_VID(pSma), pReq->name, pReq->suid);
return TSDB_CODE_SUCCESS;
}
// from write queue: single thead
pRSmaInfo = (SRSmaInfo *)taosMemoryCalloc(1, sizeof(SRSmaInfo));
if (!pRSmaInfo) {
terrno = TSDB_CODE_OUT_OF_MEMORY;
@ -204,9 +262,8 @@ int32_t tdProcessRSmaCreate(SVnode *pVnode, SVCreateStbReq *pReq) {
STqReadHandle *pReadHandle = tqInitSubmitMsgScanner(pMeta);
if (!pReadHandle) {
taosMemoryFree(pRSmaInfo);
terrno = TSDB_CODE_OUT_OF_MEMORY;
return TSDB_CODE_FAILED;
goto _err;
}
SReadHandle handle = {
@ -216,32 +273,33 @@ int32_t tdProcessRSmaCreate(SVnode *pVnode, SVCreateStbReq *pReq) {
.vnode = pVnode,
};
if (param->qmsg1) {
pRSmaInfo->taskInfo[0] = qCreateStreamExecTaskInfo(param->qmsg1, &handle);
if (!pRSmaInfo->taskInfo[0]) {
taosMemoryFree(pRSmaInfo);
taosMemoryFree(pReadHandle);
return TSDB_CODE_FAILED;
}
STSchema *pTSchema = metaGetTbTSchema(SMA_META(pSma), pReq->suid, -1);
if (!pTSchema) {
terrno = TSDB_CODE_TDB_IVD_TB_SCHEMA_VERSION;
goto _err;
}
pRSmaInfo->pTSchema = pTSchema;
pRSmaInfo->pSma = pSma;
pRSmaInfo->suid = pReq->suid;
if (param->qmsg2) {
pRSmaInfo->taskInfo[1] = qCreateStreamExecTaskInfo(param->qmsg2, &handle);
if (!pRSmaInfo->taskInfo[1]) {
taosMemoryFree(pRSmaInfo);
taosMemoryFree(pReadHandle);
return TSDB_CODE_FAILED;
}
}
SRetention *pRetention = SMA_RETENTION(pSma);
STsdbCfg *pTsdbCfg = SMA_TSDB_CFG(pSma);
SET_RSMA_INFO_ITEM_PARAMS(0, 1);
SET_RSMA_INFO_ITEM_PARAMS(1, 2);
if (taosHashPut(SMA_STAT_INFO_HASH(pStat), &pReq->suid, sizeof(tb_uid_t), &pRSmaInfo, sizeof(pRSmaInfo)) !=
TSDB_CODE_SUCCESS) {
return TSDB_CODE_FAILED;
goto _err;
} else {
smaDebug("vgId:%d, register rsma info succeed for suid:%" PRIi64, SMA_VID(pSma), pReq->suid);
}
return TSDB_CODE_SUCCESS;
_err:
tdFreeRSmaInfo(pRSmaInfo);
taosMemoryFree(pReadHandle);
return TSDB_CODE_FAILED;
}
/**
@ -291,12 +349,12 @@ static int32_t tdUidStorePut(STbUidStore *pStore, tb_uid_t suid, tb_uid_t *uid)
terrno = TSDB_CODE_OUT_OF_MEMORY;
return TSDB_CODE_FAILED;
}
if (taosHashPut(pStore->uidHash, &suid, sizeof(suid), &pUidArray, sizeof(pUidArray)) != 0) {
if (taosHashPut(pStore->uidHash, &suid, sizeof(suid), &pUidArray, sizeof(pUidArray)) < 0) {
return TSDB_CODE_FAILED;
}
}
} else {
if (taosHashPut(pStore->uidHash, &suid, sizeof(suid), NULL, 0) != 0) {
if (taosHashPut(pStore->uidHash, &suid, sizeof(suid), NULL, 0) < 0) {
return TSDB_CODE_FAILED;
}
}
@ -367,22 +425,15 @@ static int32_t tdFetchSubmitReqSuids(SSubmitReq *pMsg, STbUidStore *pStore) {
return 0;
}
static FORCE_INLINE int32_t tdExecuteRSmaImpl(SSma *pSma, const void *pMsg, int32_t inputType, qTaskInfo_t *taskInfo,
STSchema *pTSchema, tb_uid_t suid, int8_t level) {
static int32_t tdFetchAndSubmitRSmaResult(SRSmaInfoItem *pItem, int8_t blkType) {
SArray *pResult = NULL;
SRSmaInfo *pRSmaInfo = pItem->pRsmaInfo;
SSma *pSma = pRSmaInfo->pSma;
if (!taskInfo) {
smaDebug("vgId:%d, no qTaskInfo to execute rsma %" PRIi8 " task for suid:%" PRIu64, SMA_VID(pSma), level, suid);
return TSDB_CODE_SUCCESS;
}
smaDebug("vgId:%d, execute rsma %" PRIi8 " task for qTaskInfo:%p suid:%" PRIu64, SMA_VID(pSma), level, taskInfo, suid);
qSetStreamInput(taskInfo, pMsg, inputType, true);
while (1) {
SSDataBlock *output = NULL;
uint64_t ts;
if (qExecTask(taskInfo, &output, &ts) < 0) {
if (qExecTask(pItem->taskInfo, &output, &ts) < 0) {
ASSERT(false);
}
if (!output) {
@ -402,12 +453,12 @@ static FORCE_INLINE int32_t tdExecuteRSmaImpl(SSma *pSma, const void *pMsg, int3
if (taosArrayGetSize(pResult) > 0) {
#if 0
char flag[10] = {0};
snprintf(flag, 10, "level %" PRIi8, level);
snprintf(flag, 10, "level %" PRIi8, pItem->level);
blockDebugShowData(pResult, flag);
#endif
STsdb *sinkTsdb = (level == TSDB_RETENTION_L1 ? pSma->pRSmaTsdb1 : pSma->pRSmaTsdb2);
STsdb *sinkTsdb = (pItem->level == TSDB_RETENTION_L1 ? pSma->pRSmaTsdb1 : pSma->pRSmaTsdb2);
SSubmitReq *pReq = NULL;
if (buildSubmitReqFromDataBlock(&pReq, pResult, pTSchema, SMA_VID(pSma), suid) < 0) {
if (buildSubmitReqFromDataBlock(&pReq, pResult, pRSmaInfo->pTSchema, SMA_VID(pSma), pRSmaInfo->suid) < 0) {
taosArrayDestroy(pResult);
return TSDB_CODE_FAILED;
}
@ -420,10 +471,63 @@ static FORCE_INLINE int32_t tdExecuteRSmaImpl(SSma *pSma, const void *pMsg, int3
taosMemoryFreeClear(pReq);
} else {
smaDebug("vgId:%d, no rsma % " PRIi8 " data generated since %s", SMA_VID(pSma), level, tstrerror(terrno));
smaDebug("vgId:%d, no rsma % " PRIi8 " data generated since %s", SMA_VID(pSma), pItem->level, tstrerror(terrno));
}
if (blkType == STREAM_DATA_TYPE_SUBMIT_BLOCK) {
atomic_store_8(&pItem->triggerStatus, TASK_TRIGGER_STATUS__ACTIVE);
}
taosArrayDestroy(pResult);
return 0;
}
/**
* @brief trigger to get rsma result
*
* @param param
* @param tmrId
*/
static void rsmaTriggerByTimer(void *param, void *tmrId) {
// SRSmaTriggerParam *pParam = (SRSmaTriggerParam *)param;
// SRSmaInfoItem *pItem = pParam->pItem;
SRSmaInfoItem *pItem = param;
if (atomic_load_8(&pItem->triggerStatus) == TASK_TRIGGER_STATUS__ACTIVE) {
smaTrace("level %" PRIi8 " status is active for tb suid:%" PRIi64, pItem->level, pItem->pRsmaInfo->suid);
SSDataBlock dataBlock = {.info.type = STREAM_GET_ALL};
atomic_store_8(&pItem->triggerStatus, TASK_TRIGGER_STATUS__IN_ACTIVE);
qSetStreamInput(pItem->taskInfo, &dataBlock, STREAM_DATA_TYPE_SSDATA_BLOCK, false);
tdFetchAndSubmitRSmaResult(pItem, STREAM_DATA_TYPE_SSDATA_BLOCK);
} else {
smaTrace("level %" PRIi8 " status is inactive for tb suid:%" PRIi64, pItem->level, pItem->pRsmaInfo->suid);
}
// taosTmrReset(rsmaTriggerByTimer, pItem->maxDelay, pItem, pItem->tmrHandle, &pItem->tmrId);
}
static FORCE_INLINE int32_t tdExecuteRSmaImpl(SSma *pSma, const void *pMsg, int32_t inputType, SRSmaInfoItem *pItem,
tb_uid_t suid, int8_t level) {
if (!pItem || !pItem->taskInfo) {
smaDebug("vgId:%d, no qTaskInfo to execute rsma %" PRIi8 " task for suid:%" PRIu64, SMA_VID(pSma), level, suid);
return TSDB_CODE_SUCCESS;
}
smaDebug("vgId:%d, execute rsma %" PRIi8 " task for qTaskInfo:%p suid:%" PRIu64, SMA_VID(pSma), level,
pItem->taskInfo, suid);
// inputType = STREAM_DATA_TYPE_SUBMIT_BLOCK(1)
if (qSetStreamInput(pItem->taskInfo, pMsg, inputType, true) < 0) {
smaError("vgId:%d, rsma % " PRIi8 " qSetStreamInput failed since %s", SMA_VID(pSma), level, tstrerror(terrno));
return TSDB_CODE_FAILED;
}
// SRSmaTriggerParam triggerParam = {.suid = suid, .pItem = pItem, .pSma = pSma, .pTSchema = pTSchema};
tdFetchAndSubmitRSmaResult(pItem, STREAM_DATA_TYPE_SUBMIT_BLOCK);
atomic_store_8(&pItem->triggerStatus, TASK_TRIGGER_STATUS__ACTIVE);
taosTmrReset(rsmaTriggerByTimer, pItem->maxDelay, pItem, pItem->tmrHandle, &pItem->tmrId);
return TSDB_CODE_SUCCESS;
}
@ -441,24 +545,18 @@ static int32_t tdExecuteRSma(SSma *pSma, const void *pMsg, int32_t inputType, tb
pRSmaInfo = taosHashGet(SMA_STAT_INFO_HASH(pStat), &suid, sizeof(tb_uid_t));
if (!pRSmaInfo || !(pRSmaInfo = *(SRSmaInfo **)pRSmaInfo)) {
smaDebug("vgId:%d, no rsma info for suid:%" PRIu64, SMA_VID(pSma), suid);
smaDebug("vgId:%d, return as no rsma info for suid:%" PRIu64, SMA_VID(pSma), suid);
return TSDB_CODE_SUCCESS;
}
if (!pRSmaInfo->taskInfo[0]) {
smaDebug("vgId:%d, no rsma qTaskInfo for suid:%" PRIu64, SMA_VID(pSma), suid);
if (!pRSmaInfo->items[0].taskInfo) {
smaDebug("vgId:%d, return as no rsma qTaskInfo for suid:%" PRIu64, SMA_VID(pSma), suid);
return TSDB_CODE_SUCCESS;
}
if (inputType == STREAM_DATA_TYPE_SUBMIT_BLOCK) {
// TODO: cache STSchema
STSchema *pTSchema = metaGetTbTSchema(SMA_META(pSma), suid, -1);
if (!pTSchema) {
terrno = TSDB_CODE_TDB_IVD_TB_SCHEMA_VERSION;
return TSDB_CODE_FAILED;
}
tdExecuteRSmaImpl(pSma, pMsg, inputType, pRSmaInfo->taskInfo[0], pTSchema, suid, TSDB_RETENTION_L1);
tdExecuteRSmaImpl(pSma, pMsg, inputType, pRSmaInfo->taskInfo[1], pTSchema, suid, TSDB_RETENTION_L2);
taosMemoryFree(pTSchema);
tdExecuteRSmaImpl(pSma, pMsg, inputType, &pRSmaInfo->items[0], suid, TSDB_RETENTION_L1);
tdExecuteRSmaImpl(pSma, pMsg, inputType, &pRSmaInfo->items[1], suid, TSDB_RETENTION_L2);
}
return TSDB_CODE_SUCCESS;

46
source/dnode/vnode/src/tq/tq.c
View File

@ -125,10 +125,10 @@ int32_t tqProcessOffsetCommitReq(STQ* pTq, char* msg, int32_t msgLen) {
if (offset.type == TMQ_OFFSET__SNAPSHOT) {
tqDebug("receive offset commit msg to %s on vg %d, offset(type:snapshot) uid: %ld, ts: %ld", offset.subKey,
pTq->pVnode->config.vgId, offset.uid, offset.ts);
TD_VID(pTq->pVnode), offset.uid, offset.ts);
} else if (offset.type == TMQ_OFFSET__LOG) {
tqDebug("receive offset commit msg to %s on vg %d, offset(type:log) version: %ld", offset.subKey,
pTq->pVnode->config.vgId, offset.version);
TD_VID(pTq->pVnode), offset.version);
} else {
ASSERT(0);
}
@ -159,7 +159,7 @@ int32_t tqProcessPollReq(STQ* pTq, SRpcMsg* pMsg, int32_t workerId) {
if (pOffset != NULL) {
ASSERT(pOffset->type == TMQ_OFFSET__LOG);
tqDebug("consumer %ld, restore offset of %s on vg %d, offset(type:log) version: %ld", consumerId, pReq->subKey,
pTq->pVnode->config.vgId, pOffset->version);
TD_VID(pTq->pVnode), pOffset->version);
fetchOffset = pOffset->version + 1;
} else {
if (pReq->currentOffset == TMQ_CONF__RESET_OFFSET__EARLIEAST) {
@ -167,13 +167,13 @@ int32_t tqProcessPollReq(STQ* pTq, SRpcMsg* pMsg, int32_t workerId) {
} else if (pReq->currentOffset == TMQ_CONF__RESET_OFFSET__LATEST) {
fetchOffset = walGetCommittedVer(pTq->pWal);
} else if (pReq->currentOffset == TMQ_CONF__RESET_OFFSET__NONE) {
tqError("tmq poll: no offset committed for consumer %ld in vg %d, subkey %s", consumerId,
pTq->pVnode->config.vgId, pReq->subKey);
tqError("tmq poll: no offset committed for consumer %ld in vg %d, subkey %s", consumerId, TD_VID(pTq->pVnode),
pReq->subKey);
terrno = TSDB_CODE_TQ_NO_COMMITTED_OFFSET;
return -1;
}
tqDebug("consumer %ld, restore offset of %s on vg %d failed, config is %ld, set to %ld", consumerId, pReq->subKey,
pTq->pVnode->config.vgId, pReq->currentOffset, fetchOffset);
TD_VID(pTq->pVnode), pReq->currentOffset, fetchOffset);
}
}
@ -183,14 +183,14 @@ int32_t tqProcessPollReq(STQ* pTq, SRpcMsg* pMsg, int32_t workerId) {
STqHandle* pHandle = taosHashGet(pTq->handles, pReq->subKey, strlen(pReq->subKey));
/*ASSERT(pHandle);*/
if (pHandle == NULL) {
tqError("tmq poll: no consumer handle for consumer %ld in vg %d, subkey %s", consumerId, pTq->pVnode->config.vgId,
tqError("tmq poll: no consumer handle for consumer %ld in vg %d, subkey %s", consumerId, TD_VID(pTq->pVnode),
pReq->subKey);
return -1;
}
if (pHandle->consumerId != consumerId) {
tqError("tmq poll: consumer handle mismatch for consumer %ld in vg %d, subkey %s, handle consumer id %ld",
consumerId, pTq->pVnode->config.vgId, pReq->subKey, pHandle->consumerId);
consumerId, TD_VID(pTq->pVnode), pReq->subKey, pHandle->consumerId);
return -1;
}
@ -304,7 +304,6 @@ int32_t tqProcessVgDeleteReq(STQ* pTq, char* msg, int32_t msgLen) {
return 0;
}
// TODO: persist meta into tdb
int32_t tqProcessVgChangeReq(STQ* pTq, char* msg, int32_t msgLen) {
SMqRebVgReq req = {0};
tDecodeSMqRebVgReq(msg, &req);
@ -334,7 +333,6 @@ int32_t tqProcessVgChangeReq(STQ* pTq, char* msg, int32_t msgLen) {
SReadHandle handle = {
.reader = pHandle->execHandle.pExecReader[i],
.meta = pTq->pVnode->pMeta,
.pMsgCb = &pTq->pVnode->msgCb,
};
pHandle->execHandle.execCol.task[i] = qCreateStreamExecTaskInfo(pHandle->execHandle.execCol.qmsg, &handle);
ASSERT(pHandle->execHandle.execCol.task[i]);
@ -346,10 +344,10 @@ int32_t tqProcessVgChangeReq(STQ* pTq, char* msg, int32_t msgLen) {
pHandle->execHandle.execTb.suid = req.suid;
SArray* tbUidList = taosArrayInit(0, sizeof(int64_t));
tsdbGetCtbIdList(pTq->pVnode->pMeta, req.suid, tbUidList);
tqDebug("vg %d, tq try get suid: %ld", pTq->pVnode->config.vgId, req.suid);
tqDebug("vg %d, tq try get suid: %ld", TD_VID(pTq->pVnode), req.suid);
for (int32_t i = 0; i < taosArrayGetSize(tbUidList); i++) {
int64_t tbUid = *(int64_t*)taosArrayGet(tbUidList, i);
tqDebug("vg %d, idx %d, uid: %ld", pTq->pVnode->config.vgId, i, tbUid);
tqDebug("vg %d, idx %d, uid: %ld", TD_VID(pTq->pVnode), i, tbUid);
}
for (int32_t i = 0; i < 5; i++) {
tqReadHandleSetTbUidList(pHandle->execHandle.pExecReader[i], tbUidList);
@ -374,7 +372,7 @@ int32_t tqProcessVgChangeReq(STQ* pTq, char* msg, int32_t msgLen) {
return 0;
}
int32_t tqProcessTaskDeploy(STQ* pTq, char* msg, int32_t msgLen) {
int32_t tqProcessTaskDeployReq(STQ* pTq, char* msg, int32_t msgLen) {
SStreamTask* pTask = taosMemoryCalloc(1, sizeof(SStreamTask));
if (pTask == NULL) {
return -1;
@ -400,16 +398,20 @@ int32_t tqProcessTaskDeploy(STQ* pTq, char* msg, int32_t msgLen) {
// exec
if (pTask->execType != TASK_EXEC__NONE) {
// expand runners
if (pTask->dataScan) {
STqReadHandle* pStreamReader = tqInitSubmitMsgScanner(pTq->pVnode->pMeta);
SReadHandle handle = {
.reader = pStreamReader,
.meta = pTq->pVnode->pMeta,
.pMsgCb = &pTq->pVnode->msgCb,
.vnode = pTq->pVnode,
};
pTask->exec.inputHandle = pStreamReader;
/*pTask->exec.inputHandle = pStreamReader;*/
pTask->exec.executor = qCreateStreamExecTaskInfo(pTask->exec.qmsg, &handle);
ASSERT(pTask->exec.executor);
} else {
pTask->exec.executor = qCreateStreamExecTaskInfo(pTask->exec.qmsg, NULL);
ASSERT(pTask->exec.executor);
}
}
// sink
@ -431,7 +433,7 @@ int32_t tqProcessTaskDeploy(STQ* pTq, char* msg, int32_t msgLen) {
streamSetupTrigger(pTask);
tqInfo("deploy stream task id %d child id %d on vg %d", pTask->taskId, pTask->childId, pTq->pVnode->config.vgId);
tqInfo("deploy stream task id %d child id %d on vg %d", pTask->taskId, pTask->childId, TD_VID(pTq->pVnode));
taosHashPut(pTq->pStreamTasks, &pTask->taskId, sizeof(int32_t), &pTask, sizeof(void*));
@ -464,7 +466,7 @@ int32_t tqProcessStreamTrigger(STQ* pTq, SSubmitReq* pReq) {
continue;
}
if (streamLaunchByWrite(pTask, pTq->pVnode->config.vgId, &pTq->pVnode->msgCb) < 0) {
if (streamLaunchByWrite(pTask, TD_VID(pTq->pVnode)) < 0) {
continue;
}
} else {
@ -485,7 +487,7 @@ int32_t tqProcessTaskRunReq(STQ* pTq, SRpcMsg* pMsg) {
SStreamTaskRunReq* pReq = pMsg->pCont;
int32_t taskId = pReq->taskId;
SStreamTask* pTask = *(SStreamTask**)taosHashGet(pTq->pStreamTasks, &taskId, sizeof(int32_t));
streamTaskProcessRunReq(pTask, &pTq->pVnode->msgCb);
streamProcessRunReq(pTask);
return 0;
}
@ -503,7 +505,7 @@ int32_t tqProcessTaskDispatchReq(STQ* pTq, SRpcMsg* pMsg) {
.info = pMsg->info,
.code = 0,
};
streamProcessDispatchReq(pTask, &pTq->pVnode->msgCb, &req, &rsp);
streamProcessDispatchReq(pTask, &req, &rsp);
return 0;
}
@ -511,7 +513,7 @@ int32_t tqProcessTaskRecoverReq(STQ* pTq, SRpcMsg* pMsg) {
SStreamTaskRecoverReq* pReq = pMsg->pCont;
int32_t taskId = pReq->taskId;
SStreamTask* pTask = *(SStreamTask**)taosHashGet(pTq->pStreamTasks, &taskId, sizeof(int32_t));
streamProcessRecoverReq(pTask, &pTq->pVnode->msgCb, pReq, pMsg);
streamProcessRecoverReq(pTask, pReq, pMsg);
return 0;
}
@ -519,7 +521,7 @@ int32_t tqProcessTaskDispatchRsp(STQ* pTq, SRpcMsg* pMsg) {
SStreamDispatchRsp* pRsp = POINTER_SHIFT(pMsg->pCont, sizeof(SMsgHead));
int32_t taskId = pRsp->taskId;
SStreamTask* pTask = *(SStreamTask**)taosHashGet(pTq->pStreamTasks, &taskId, sizeof(int32_t));
streamProcessDispatchRsp(pTask, &pTq->pVnode->msgCb, pRsp);
streamProcessDispatchRsp(pTask, pRsp);
return 0;
}
@ -534,9 +536,9 @@ int32_t tqProcessTaskRecoverRsp(STQ* pTq, SRpcMsg* pMsg) {
int32_t tqProcessTaskDropReq(STQ* pTq, char* msg, int32_t msgLen) {
SVDropStreamTaskReq* pReq = (SVDropStreamTaskReq*)msg;
int32_t code = taosHashRemove(pTq->pStreamTasks, &pReq->taskId, sizeof(int32_t));
ASSERT(code == 0);
if (code == 0) {
// sendrsp
}
ASSERT(code == 0);
return code;
}

10
source/dnode/vnode/src/vnd/vnodeSvr.c
View File

@ -106,7 +106,7 @@ int32_t vnodeProcessWriteReq(SVnode *pVnode, SRpcMsg *pMsg, int64_t version, SRp
int32_t len;
int32_t ret;
vError("vgId:%d, start to process write request %s, index:%" PRId64, TD_VID(pVnode), TMSG_INFO(pMsg->msgType),
vTrace("vgId:%d, start to process write request %s, index:%" PRId64, TD_VID(pVnode), TMSG_INFO(pMsg->msgType),
version);
pVnode->state.applied = version;
@ -167,7 +167,7 @@ int32_t vnodeProcessWriteReq(SVnode *pVnode, SRpcMsg *pMsg, int64_t version, SRp
}
break;
case TDMT_STREAM_TASK_DEPLOY: {
if (tqProcessTaskDeploy(pVnode->pTq, POINTER_SHIFT(pMsg->pCont, sizeof(SMsgHead)),
if (tqProcessTaskDeployReq(pVnode->pTq, POINTER_SHIFT(pMsg->pCont, sizeof(SMsgHead)),
pMsg->contLen - sizeof(SMsgHead)) < 0) {
goto _err;
}
@ -305,7 +305,6 @@ void vnodeUpdateMetaRsp(SVnode *pVnode, STableMetaRsp *pMetaRsp) {
}
static int32_t vnodeProcessDropTtlTbReq(SVnode *pVnode, int64_t version, void *pReq, int32_t len, SRpcMsg *pRsp) {
SArray *tbUids = taosArrayInit(8, sizeof(int64_t));
if (tbUids == NULL) return TSDB_CODE_OUT_OF_MEMORY;
@ -346,7 +345,10 @@ static int32_t vnodeProcessCreateStbReq(SVnode *pVnode, int64_t version, void *p
goto _err;
}
tdProcessRSmaCreate(pVnode, &req);
if (tdProcessRSmaCreate(pVnode, &req) < 0) {
pRsp->code = terrno;
goto _err;
}
tDecoderClear(&coder);
return 0;

20
source/dnode/vnode/src/vnd/vnodeSync.c
View File

@ -211,21 +211,23 @@ int32_t vnodeProcessSyncReq(SVnode *pVnode, SRpcMsg *pMsg, SRpcMsg **pRsp) {
SSyncNode *pSyncNode = syncNodeAcquire(pVnode->sync);
assert(pSyncNode != NULL);
ESyncState state = syncGetMyRole(pVnode->sync);
SyncTerm currentTerm = syncGetMyTerm(pVnode->sync);
SMsgHead *pHead = pMsg->pCont;
char logBuf[512] = {0};
char *syncNodeStr = sync2SimpleStr(pVnode->sync);
snprintf(logBuf, sizeof(logBuf), "==vnodeProcessSyncReq== msgType:%d, syncNode: %s", pMsg->msgType, syncNodeStr);
static int64_t vndTick = 0;
STraceId *trace = &pMsg->info.traceId;
do {
char *syncNodeStr = sync2SimpleStr(pVnode->sync);
static int64_t vndTick = 0;
if (++vndTick % 10 == 1) {
vGTrace("sync trace msg:%s, %s", TMSG_INFO(pMsg->msgType), syncNodeStr);
vGTrace("vgId:%d, sync heartbeat msg:%s, %s", syncGetVgId(pVnode->sync), TMSG_INFO(pMsg->msgType), syncNodeStr);
}
if (gRaftDetailLog) {
char logBuf[512] = {0};
snprintf(logBuf, sizeof(logBuf), "==vnodeProcessSyncReq== msgType:%d, syncNode: %s", pMsg->msgType,
syncNodeStr);
syncRpcMsgLog2(logBuf, pMsg);
}
taosMemoryFree(syncNodeStr);
} while (0);
SRpcMsg *pRpcMsg = pMsg;

2
source/libs/executor/inc/executil.h
View File

@ -114,7 +114,7 @@ SArray* extractColMatchInfo(SNodeList* pNodeList, SDataBlockDescNode* pOutputNod
SExprInfo* createExprInfo(SNodeList* pNodeList, SNodeList* pGroupKeys, int32_t* numOfExprs);
SqlFunctionCtx* createSqlFunctionCtx(SExprInfo* pExprInfo, int32_t numOfOutput, int32_t** rowEntryInfoOffset);
void relocateColumnData(SSDataBlock* pBlock, const SArray* pColMatchInfo, SArray* pCols);
void relocateColumnData(SSDataBlock* pBlock, const SArray* pColMatchInfo, SArray* pCols, bool outputEveryColumn);
void initExecTimeWindowInfo(SColumnInfoData* pColData, STimeWindow* pQueryWindow);
SInterval extractIntervalInfo(const STableScanPhysiNode* pTableScanNode);

3
source/libs/executor/inc/executorimpl.h
View File

@ -506,7 +506,8 @@ typedef struct SPartitionOperatorInfo {
SDiskbasedBuf* pBuf; // query result buffer based on blocked-wised disk file
int32_t rowCapacity; // maximum number of rows for each buffer page
int32_t* columnOffset; // start position for each column data
void* pGroupIter; // group iterator
SArray* sortedGroupArray; // SDataGroupInfo sorted by group id
int32_t groupIndex; // group index
int32_t pageIndex; // page index of current group
SSDataBlock* pUpdateRes;
SExprSupp scalarSup;

12
source/libs/executor/src/dataDispatcher.c
View File

@ -75,12 +75,18 @@ static bool needCompress(const SSDataBlock* pData, int32_t numOfCols) {
// The length of bitmap is decided by number of rows of this data block, and the length of each column data is
// recorded in the first segment, next to the struct header
static void toDataCacheEntry(SDataDispatchHandle* pHandle, const SInputData* pInput, SDataDispatchBuf* pBuf) {
int32_t numOfCols = LIST_LENGTH(pHandle->pSchema->pSlots);
int32_t numOfCols = 0;
SNode* pNode;
FOREACH(pNode, pHandle->pSchema->pSlots) {
SSlotDescNode* pSlotDesc = (SSlotDescNode*)pNode;
if (pSlotDesc->output) {
++numOfCols;
}
}
SDataCacheEntry* pEntry = (SDataCacheEntry*)pBuf->pData;
pEntry->compressed = (int8_t)needCompress(pInput->pData, numOfCols);
pEntry->numOfRows = pInput->pData->info.rows;
pEntry->numOfCols = pInput->pData->info.numOfCols;
pEntry->numOfCols = numOfCols;
pEntry->dataLen = 0;
pBuf->useSize = sizeof(SDataCacheEntry);

16
source/libs/executor/src/executil.c
View File

@ -235,7 +235,7 @@ int32_t getTableList(void* metaHandle, SScanPhysiNode* pScanNode, STableListInfo
terrno = code;
return code;
} else {
qDebug("sucess to get tableIds, size: %d, suid: %" PRIu64 "", (int)taosArrayGetSize(res), tableUid);
qDebug("success to get tableIds, size: %d, suid: %" PRIu64 "", (int)taosArrayGetSize(res), tableUid);
}
for (int i = 0; i < taosArrayGetSize(res); i++) {
@ -319,7 +319,14 @@ SArray* extractColMatchInfo(SNodeList* pNodeList, SDataBlockDescNode* pOutputNod
continue;
}
SColMatchInfo* info = taosArrayGet(pList, pNode->slotId);
SColMatchInfo* info = NULL;
for (int32_t j = 0; j < taosArrayGetSize(pList); ++j) {
info = taosArrayGet(pList, j);
if (info->targetSlotId == pNode->slotId) {
break;
}
}
if (pNode->output) {
(*numOfOutputCols) += 1;
} else {
@ -578,14 +585,15 @@ SqlFunctionCtx* createSqlFunctionCtx(SExprInfo* pExprInfo, int32_t numOfOutput,
}
// NOTE: sources columns are more than the destination SSDatablock columns.
void relocateColumnData(SSDataBlock* pBlock, const SArray* pColMatchInfo, SArray* pCols) {
// doFilter in table scan needs every column even its output is false
void relocateColumnData(SSDataBlock* pBlock, const SArray* pColMatchInfo, SArray* pCols, bool outputEveryColumn) {
size_t numOfSrcCols = taosArrayGetSize(pCols);
int32_t i = 0, j = 0;
while (i < numOfSrcCols && j < taosArrayGetSize(pColMatchInfo)) {
SColumnInfoData* p = taosArrayGet(pCols, i);
SColMatchInfo* pmInfo = taosArrayGet(pColMatchInfo, j);
if (!pmInfo->output) {
if (!outputEveryColumn && !pmInfo->output) {
j++;
continue;
}

8
source/libs/executor/src/executor.c
View File

@ -41,12 +41,18 @@ static int32_t doSetStreamBlock(SOperatorInfo* pOperator, void* input, size_t nu
pInfo->assignBlockUid = assignUid;
// the block type can not be changed in the streamscan operators
#if 0
if (pInfo->blockType == 0) {
pInfo->blockType = type;
} else if (pInfo->blockType != type) {
ASSERT(0);
return TSDB_CODE_QRY_APP_ERROR;
}
#endif
// rollup sma, the same qTaskInfo is used to insert data by SubmitReq and fetch result by SSDataBlock
if (pInfo->blockType != type) {
pInfo->blockType = type;
}
if (type == STREAM_DATA_TYPE_SUBMIT_BLOCK) {
if (tqReadHandleSetMsg(pInfo->streamBlockReader, input, 0) < 0) {
@ -99,7 +105,7 @@ int32_t qSetMultiStreamInput(qTaskInfo_t tinfo, const void* pBlocks, size_t numO
}
qTaskInfo_t qCreateStreamExecTaskInfo(void* msg, void* streamReadHandle) {
if (msg == NULL || streamReadHandle == NULL) {
if (msg == NULL) {
return NULL;
}

13
source/libs/executor/src/executorMain.c
View File

@ -13,10 +13,10 @@
* along with this program. If not, see <http://www.gnu.org/licenses/>.
*/
#include "os.h"
#include "tref.h"
#include "dataSinkMgt.h"
#include "os.h"
#include "tmsg.h"
#include "tref.h"
#include "tudf.h"
#include "executor.h"
@ -26,13 +26,11 @@
static TdThreadOnce initPoolOnce = PTHREAD_ONCE_INIT;
int32_t exchangeObjRefPool = -1;
static void initRefPool() {
exchangeObjRefPool = taosOpenRef(1024, doDestroyExchangeOperatorInfo);
}
static void initRefPool() { exchangeObjRefPool = taosOpenRef(1024, doDestroyExchangeOperatorInfo); }
int32_t qCreateExecTask(SReadHandle* readHandle, int32_t vgId, uint64_t taskId, SSubplan* pSubplan,
qTaskInfo_t* pTaskInfo, DataSinkHandle* handle, const char* sql, EOPTR_EXEC_MODEL model) {
assert(readHandle != NULL && pSubplan != NULL);
assert(pSubplan != NULL);
SExecTaskInfo** pTask = (SExecTaskInfo**)pTaskInfo;
taosThreadOnce(&initPoolOnce, initRefPool);
@ -115,8 +113,7 @@ int32_t qExecTask(qTaskInfo_t tinfo, SSDataBlock** pRes, uint64_t *useconds) {
*pRes = NULL;
int64_t curOwner = 0;
if ((curOwner = atomic_val_compare_exchange_64(&pTaskInfo->owner, 0, threadId)) != 0) {
qError("%s-%p execTask is now executed by thread:%p", GET_TASKID(pTaskInfo), pTaskInfo,
(void*)curOwner);
qError("%s-%p execTask is now executed by thread:%p", GET_TASKID(pTaskInfo), pTaskInfo, (void*)curOwner);
pTaskInfo->code = TSDB_CODE_QRY_IN_EXEC;
return pTaskInfo->code;
}

43
source/libs/executor/src/executorimpl.c
View File

@ -141,8 +141,8 @@ static int32_t doCopyToSDataBlock(SExecTaskInfo* taskInfo, SSDataBlock* pBlock,
SqlFunctionCtx* pCtx, int32_t numOfExprs);
static void initCtxOutputBuffer(SqlFunctionCtx* pCtx, int32_t size);
static void doSetTableGroupOutputBuf(SOperatorInfo* pOperator, SAggOperatorInfo* pAggInfo, int32_t numOfOutput, uint64_t groupId);
static void doSetTableGroupOutputBuf(SOperatorInfo* pOperator, SAggOperatorInfo* pAggInfo, int32_t numOfOutput,
uint64_t groupId);
// setup the output buffer for each operator
static bool hasNull(SColumn* pColumn, SColumnDataAgg* pStatis) {
@ -1230,7 +1230,8 @@ void initResultRow(SResultRow* pResultRow) {
* offset[0] offset[1] offset[2]
*/
// TODO refactor: some function move away
void setFunctionResultOutput(SOperatorInfo *pOperator, SOptrBasicInfo *pInfo, SAggSupporter* pSup, int32_t stage, int32_t numOfExprs) {
void setFunctionResultOutput(SOperatorInfo* pOperator, SOptrBasicInfo* pInfo, SAggSupporter* pSup, int32_t stage,
int32_t numOfExprs) {
SExecTaskInfo* pTaskInfo = pOperator->pTaskInfo;
SqlFunctionCtx* pCtx = pOperator->exprSupp.pCtx;
int32_t* rowEntryInfoOffset = pOperator->exprSupp.rowEntryInfoOffset;
@ -1380,7 +1381,8 @@ void extractQualifiedTupleByFilterResult(SSDataBlock* pBlock, const int8_t* rowR
}
}
void doSetTableGroupOutputBuf(SOperatorInfo* pOperator, SAggOperatorInfo* pAggInfo, int32_t numOfOutput, uint64_t groupId) {
void doSetTableGroupOutputBuf(SOperatorInfo* pOperator, SAggOperatorInfo* pAggInfo, int32_t numOfOutput,
uint64_t groupId) {
// for simple group by query without interval, all the tables belong to one group result.
SExecTaskInfo* pTaskInfo = pOperator->pTaskInfo;
SResultRowInfo* pResultRowInfo = &pAggInfo->binfo.resultRowInfo;
@ -2084,7 +2086,7 @@ int32_t extractDataBlockFromFetchRsp(SSDataBlock* pRes, SLoadRemoteDataInfo* pLo
// data from mnode
pRes->info.rows = numOfRows;
relocateColumnData(pRes, pColList, pBlock->pDataBlock);
relocateColumnData(pRes, pColList, pBlock->pDataBlock, false);
taosArrayDestroy(pBlock->pDataBlock);
taosMemoryFree(pBlock);
// blockDataDestroy(pBlock);
@ -2165,7 +2167,8 @@ static SSDataBlock* concurrentlyLoadRemoteDataImpl(SOperatorInfo* pOperator, SEx
}
SRetrieveTableRsp* pTableRsp = pDataInfo->pRsp;
code = extractDataBlockFromFetchRsp(pExchangeInfo->pResult, pLoadInfo, pTableRsp->numOfRows, pTableRsp->data,
code =
extractDataBlockFromFetchRsp(pExchangeInfo->pResult, pLoadInfo, pTableRsp->numOfRows, pTableRsp->data,
pTableRsp->compLen, pTableRsp->numOfCols, startTs, &pDataInfo->totalRows, NULL);
if (code != 0) {
taosMemoryFreeClear(pDataInfo->pRsp);
@ -2673,8 +2676,8 @@ static SSDataBlock* doSortedMerge(SOperatorInfo* pOperator) {
}
int32_t numOfBufPage = pInfo->sortBufSize / pInfo->bufPageSize;
pInfo->pSortHandle = tsortCreateSortHandle(pInfo->pSortInfo, SORT_MULTISOURCE_MERGE, pInfo->bufPageSize,
numOfBufPage, pInfo->binfo.pRes, "GET_TASKID(pTaskInfo)");
pInfo->pSortHandle = tsortCreateSortHandle(pInfo->pSortInfo, SORT_MULTISOURCE_MERGE, pInfo->bufPageSize, numOfBufPage,
pInfo->binfo.pRes, "GET_TASKID(pTaskInfo)");
tsortSetFetchRawDataFp(pInfo->pSortHandle, loadNextDataBlock, NULL, NULL);
@ -3404,8 +3407,8 @@ void cleanupAggSup(SAggSupporter* pAggSup) {
destroyDiskbasedBuf(pAggSup->pResultBuf);
}
int32_t initAggInfo(SExprSupp* pSup, SAggSupporter* pAggSup, SExprInfo* pExprInfo, int32_t numOfCols,
size_t keyBufSize, const char* pkey) {
int32_t initAggInfo(SExprSupp* pSup, SAggSupporter* pAggSup, SExprInfo* pExprInfo, int32_t numOfCols, size_t keyBufSize,
const char* pkey) {
int32_t code = initExprSupp(pSup, pExprInfo, numOfCols);
if (code != TSDB_CODE_SUCCESS) {
return code;
@ -3591,7 +3594,8 @@ static SArray* setRowTsColumnOutputInfo(SqlFunctionCtx* pCtx, int32_t numOfCols)
return pList;
}
SOperatorInfo* createProjectOperatorInfo(SOperatorInfo* downstream, SProjectPhysiNode* pProjPhyNode, SExecTaskInfo* pTaskInfo) {
SOperatorInfo* createProjectOperatorInfo(SOperatorInfo* downstream, SProjectPhysiNode* pProjPhyNode,
SExecTaskInfo* pTaskInfo) {
SProjectOperatorInfo* pInfo = taosMemoryCalloc(1, sizeof(SProjectOperatorInfo));
SOperatorInfo* pOperator = taosMemoryCalloc(1, sizeof(SOperatorInfo));
if (pInfo == NULL || pOperator == NULL) {
@ -3699,8 +3703,8 @@ static SSDataBlock* doApplyIndefinitFunction(SOperatorInfo* pOperator) {
setInputDataBlock(pOperator, pSup->pCtx, pBlock, order, scanFlag, false);
blockDataEnsureCapacity(pInfo->pRes, pInfo->pRes->info.rows + pBlock->info.rows);
code = projectApplyFunctions(pOperator->exprSupp.pExprInfo, pInfo->pRes, pBlock, pSup->pCtx, pOperator->exprSupp.numOfExprs,
pIndefInfo->pPseudoColInfo);
code = projectApplyFunctions(pOperator->exprSupp.pExprInfo, pInfo->pRes, pBlock, pSup->pCtx,
pOperator->exprSupp.numOfExprs, pIndefInfo->pPseudoColInfo);
if (code != TSDB_CODE_SUCCESS) {
longjmp(pTaskInfo->env, code);
}
@ -4034,13 +4038,22 @@ SOperatorInfo* createOperatorTree(SPhysiNode* pPhyNode, SExecTaskInfo* pTaskInfo
SScanPhysiNode* pScanPhyNode = (SScanPhysiNode*)pPhyNode; // simple child table.
STableScanPhysiNode* pTableScanNode = (STableScanPhysiNode*)pPhyNode;
STimeWindowAggSupp twSup = {
.waterMark = pTableScanNode->watermark, .calTrigger = pTableScanNode->triggerType, .maxTs = INT64_MIN};
.waterMark = pTableScanNode->watermark,
.calTrigger = pTableScanNode->triggerType,
.maxTs = INT64_MIN,
};
tsdbReaderT pDataReader = NULL;
if (pHandle) {
if (pHandle->vnode) {
pDataReader = doCreateDataReader(pTableScanNode, pHandle, pTableListInfo, (uint64_t)queryId, taskId, pTagCond);
// for stram
pDataReader =
doCreateDataReader(pTableScanNode, pHandle, pTableListInfo, (uint64_t)queryId, taskId, pTagCond);
} else {
// for tq
getTableList(pHandle->meta, pScanPhyNode, pTableListInfo, pTagCond);
}
}
if (pDataReader == NULL && terrno != 0) {
qDebug("%s pDataReader is NULL", GET_TASKID(pTaskInfo));

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

@ -587,24 +587,30 @@ static void clearPartitionOperator(SPartitionOperatorInfo* pInfo) {
while( (ite = taosHashIterate(pInfo->pGroupSet, ite)) != NULL ) {
taosArrayDestroy( ((SDataGroupInfo *)ite)->pPageList);
}
taosHashClear(pInfo->pGroupSet);
taosArrayClear(pInfo->sortedGroupArray);
clearDiskbasedBuf(pInfo->pBuf);
}
static int compareDataGroupInfo(const void* group1, const void* group2) {
const SDataGroupInfo* pGroupInfo1 = group1;
const SDataGroupInfo* pGroupInfo2 = group2;
return pGroupInfo1->groupId - pGroupInfo2->groupId;
}
static SSDataBlock* buildPartitionResult(SOperatorInfo* pOperator) {
SPartitionOperatorInfo* pInfo = pOperator->info;
SDataGroupInfo* pGroupInfo = pInfo->pGroupIter;
if (pInfo->pGroupIter == NULL || pInfo->pageIndex >= taosArrayGetSize(pGroupInfo->pPageList)) {
SDataGroupInfo* pGroupInfo = (pInfo->groupIndex != -1) ? taosArrayGet(pInfo->sortedGroupArray, pInfo->groupIndex) : NULL;
if (pInfo->groupIndex == -1 || pInfo->pageIndex >= taosArrayGetSize(pGroupInfo->pPageList)) {
// try next group data
pInfo->pGroupIter = taosHashIterate(pInfo->pGroupSet, pInfo->pGroupIter);
if (pInfo->pGroupIter == NULL) {
++pInfo->groupIndex;
if (pInfo->groupIndex >= taosArrayGetSize(pInfo->sortedGroupArray)) {
doSetOperatorCompleted(pOperator);
clearPartitionOperator(pInfo);
return NULL;
}
pGroupInfo = pInfo->pGroupIter;
pGroupInfo = taosArrayGet(pInfo->sortedGroupArray, pInfo->groupIndex);
pInfo->pageIndex = 0;
}
@ -658,6 +664,20 @@ static SSDataBlock* hashPartition(SOperatorInfo* pOperator) {
doHashPartition(pOperator, pBlock);
}
SArray* groupArray = taosArrayInit(taosHashGetSize(pInfo->pGroupSet), sizeof(SDataGroupInfo));
void* pGroupIter = NULL;
pGroupIter = taosHashIterate(pInfo->pGroupSet, NULL);
while (pGroupIter != NULL) {
SDataGroupInfo* pGroupInfo = pGroupIter;
taosArrayPush(groupArray, pGroupInfo);
pGroupIter = taosHashIterate(pInfo->pGroupSet, pGroupIter);
}
taosArraySort(groupArray, compareDataGroupInfo);
pInfo->sortedGroupArray = groupArray;
pInfo->groupIndex = -1;
taosHashClear(pInfo->pGroupSet);
pOperator->cost.openCost = (taosGetTimestampUs() - st) / 1000.0;
pOperator->status = OP_RES_TO_RETURN;
@ -677,6 +697,7 @@ static void destroyPartitionOperatorInfo(void* param, int32_t numOfOutput) {
taosArrayDestroy(pInfo->pGroupColVals);
taosMemoryFree(pInfo->keyBuf);
taosArrayDestroy(pInfo->sortedGroupArray);
taosHashCleanup(pInfo->pGroupSet);
taosMemoryFree(pInfo->columnOffset);

68
source/libs/executor/src/scanoperator.c
View File

@ -259,7 +259,7 @@ static int32_t loadDataBlock(SOperatorInfo* pOperator, STableScanInfo* pTableSca
return terrno;
}
relocateColumnData(pBlock, pTableScanInfo->pColMatchInfo, pCols);
relocateColumnData(pBlock, pTableScanInfo->pColMatchInfo, pCols, true);
// currently only the tbname pseudo column
if (pTableScanInfo->pseudoSup.numOfExprs > 0) {
@ -744,8 +744,7 @@ static bool prepareDataScan(SStreamBlockScanInfo* pInfo) {
int64_t gap = pInfo->sessionSup.gap;
int32_t winIndex = 0;
SResultWindowInfo* pCurWin =
getSessionTimeWindow(pAggSup, tsCols[pInfo->updateResIndex], INT64_MIN,
pSDB->info.groupId, gap, &winIndex);
getSessionTimeWindow(pAggSup, tsCols[pInfo->updateResIndex], INT64_MIN, pSDB->info.groupId, gap, &winIndex);
win = pCurWin->win;
pInfo->updateResIndex +=
updateSessionWindowInfo(pCurWin, tsCols, NULL, pSDB->info.rows, pInfo->updateResIndex, gap, NULL);
@ -1087,9 +1086,6 @@ SOperatorInfo* createStreamScanOperatorInfo(void* pDataReader, SReadHandle* pHan
SScanPhysiNode* pScanPhyNode = &pTableScanNode->scan;
SDataBlockDescNode* pDescNode = pScanPhyNode->node.pOutputDataBlockDesc;
SOperatorInfo* pTableScanDummy = createTableScanOperatorInfo(pTableScanNode, pDataReader, pHandle, pTaskInfo);
STableScanInfo* pSTInfo = (STableScanInfo*)pTableScanDummy->info;
int32_t numOfCols = 0;
pInfo->pColMatchInfo = extractColMatchInfo(pScanPhyNode->pScanCols, pDescNode, &numOfCols, COL_MATCH_FROM_COL_ID);
@ -1106,16 +1102,6 @@ SOperatorInfo* createStreamScanOperatorInfo(void* pDataReader, SReadHandle* pHan
}
}
// set the extract column id to streamHandle
tqReadHandleSetColIdList((STqReadHandle*)pHandle->reader, pColIds);
SArray* tableIdList = extractTableIdList(&pTaskInfo->tableqinfoList);
int32_t code = tqReadHandleSetTbUidList(pHandle->reader, tableIdList);
if (code != 0) {
taosArrayDestroy(tableIdList);
goto _error;
}
taosArrayDestroy(tableIdList);
pInfo->pBlockLists = taosArrayInit(4, POINTER_BYTES);
if (pInfo->pBlockLists == NULL) {
terrno = TSDB_CODE_OUT_OF_MEMORY;
@ -1127,27 +1113,43 @@ SOperatorInfo* createStreamScanOperatorInfo(void* pDataReader, SReadHandle* pHan
goto _error;
}
if (pSTInfo->interval.interval > 0 && pDataReader) {
if (pHandle) {
SOperatorInfo* pTableScanDummy = createTableScanOperatorInfo(pTableScanNode, pDataReader, pHandle, pTaskInfo);
STableScanInfo* pSTInfo = (STableScanInfo*)pTableScanDummy->info;
if (pSTInfo->interval.interval > 0) {
pInfo->pUpdateInfo = updateInfoInitP(&pSTInfo->interval, pTwSup->waterMark);
} else {
pInfo->pUpdateInfo = NULL;
}
pInfo->pOperatorDumy = pTableScanDummy;
pInfo->interval = pSTInfo->interval;
pInfo->readHandle = *pHandle;
ASSERT(pHandle->reader);
pInfo->streamBlockReader = pHandle->reader;
pInfo->tableUid = pScanPhyNode->uid;
// set the extract column id to streamHandle
tqReadHandleSetColIdList((STqReadHandle*)pHandle->reader, pColIds);
SArray* tableIdList = extractTableIdList(&pTaskInfo->tableqinfoList);
int32_t code = tqReadHandleSetTbUidList(pHandle->reader, tableIdList);
if (code != 0) {
taosArrayDestroy(tableIdList);
goto _error;
}
taosArrayDestroy(tableIdList);
}
// create the pseduo columns info
if (pTableScanNode->scan.pScanPseudoCols != NULL) {
pInfo->pPseudoExpr = createExprInfo(pTableScanNode->scan.pScanPseudoCols, NULL, &pInfo->numOfPseudoExpr);
}
pInfo->readHandle = *pHandle;
pInfo->tableUid = pScanPhyNode->uid;
pInfo->streamBlockReader = pHandle->reader;
pInfo->pRes = createResDataBlock(pDescNode);
pInfo->pUpdateRes = createResDataBlock(pDescNode);
pInfo->pCondition = pScanPhyNode->node.pConditions;
pInfo->pDataReader = pDataReader;
pInfo->scanMode = STREAM_SCAN_FROM_READERHANDLE;
pInfo->pOperatorDumy = pTableScanDummy;
pInfo->interval = pSTInfo->interval;
pInfo->sessionSup = (SessionWindowSupporter){.pStreamAggSup = NULL, .gap = -1};
pInfo->groupId = 0;
@ -1506,7 +1508,7 @@ static SSDataBlock* doSysTableScan(SOperatorInfo* pOperator) {
p->info.rows = numOfRows;
pInfo->pRes->info.rows = numOfRows;
relocateColumnData(pInfo->pRes, pInfo->scanCols, p->pDataBlock);
relocateColumnData(pInfo->pRes, pInfo->scanCols, p->pDataBlock, false);
doFilterResult(pInfo);
blockDataDestroy(p);
@ -1598,7 +1600,7 @@ int32_t buildSysDbTableInfo(const SSysTableScanInfo* pInfo, int32_t capacity) {
getPerfDbMeta(&pSysDbTableMeta, &size);
p->info.rows = buildDbTableInfoBlock(p, pSysDbTableMeta, size, TSDB_PERFORMANCE_SCHEMA_DB);
relocateColumnData(pInfo->pRes, pInfo->scanCols, p->pDataBlock);
relocateColumnData(pInfo->pRes, pInfo->scanCols, p->pDataBlock, false);
pInfo->pRes->info.rows = p->info.rows;
blockDataDestroy(p);
@ -1872,6 +1874,7 @@ SOperatorInfo* createTagScanOperatorInfo(SReadHandle* pReadHandle, STagScanPhysi
SExprInfo* pExprInfo = createExprInfo(pPhyNode->pScanPseudoCols, NULL, &numOfExprs);
SArray* colList = extractColMatchInfo(pPhyNode->pScanPseudoCols, pDescNode, &num, COL_MATCH_FROM_COL_ID);
int32_t code = initExprSupp(&pOperator->exprSupp, pExprInfo, numOfExprs);
if (code != TSDB_CODE_SUCCESS) {
goto _error;
@ -1886,6 +1889,7 @@ SOperatorInfo* createTagScanOperatorInfo(SReadHandle* pReadHandle, STagScanPhysi
pOperator->name = "TagScanOperator";
pOperator->operatorType = QUERY_NODE_PHYSICAL_PLAN_TAG_SCAN;
pOperator->blocking = false;
pOperator->status = OP_NOT_OPENED;
pOperator->info = pInfo;
@ -1958,8 +1962,7 @@ typedef struct STableMergeScanInfo {
int32_t createMultipleDataReaders(STableScanPhysiNode* pTableScanNode, SReadHandle* pHandle,
STableListInfo* pTableListInfo, SArray* arrayReader, uint64_t queryId,
uint64_t taskId, SNode* pTagCond) {
int32_t code =
getTableList(pHandle->meta, &pTableScanNode->scan, pTableListInfo, pTagCond);
int32_t code = getTableList(pHandle->meta, &pTableScanNode->scan, pTableListInfo, pTagCond);
if (code != TSDB_CODE_SUCCESS) {
goto _error;
}
@ -2082,7 +2085,7 @@ static int32_t loadDataBlockFromOneTable(SOperatorInfo* pOperator, STableMergeSc
return terrno;
}
relocateColumnData(pBlock, pTableScanInfo->pColMatchInfo, pCols);
relocateColumnData(pBlock, pTableScanInfo->pColMatchInfo, pCols, true);
// currently only the tbname pseudo column
if (pTableScanInfo->numOfPseudoExpr > 0) {
@ -2183,9 +2186,8 @@ int32_t doOpenTableMergeScanOperator(SOperatorInfo* pOperator) {
int32_t numOfBufPage = pInfo->sortBufSize / pInfo->bufPageSize;
pInfo->pSortHandle =
tsortCreateSortHandle(pInfo->pSortInfo, SORT_MULTISOURCE_MERGE, pInfo->bufPageSize,
numOfBufPage, pInfo->pSortInputBlock, pTaskInfo->id.str);
pInfo->pSortHandle = tsortCreateSortHandle(pInfo->pSortInfo, SORT_MULTISOURCE_MERGE, pInfo->bufPageSize, numOfBufPage,
pInfo->pSortInputBlock, pTaskInfo->id.str);
tsortSetFetchRawDataFp(pInfo->pSortHandle, getTableDataBlock, NULL, NULL);
@ -2249,8 +2251,7 @@ SSDataBlock* doTableMergeScan(SOperatorInfo* pOperator) {
longjmp(pTaskInfo->env, code);
}
SSDataBlock* pBlock =
getSortedTableMergeScanBlockData(pInfo->pSortHandle, pOperator->resultInfo.capacity, pOperator);
SSDataBlock* pBlock = getSortedTableMergeScanBlockData(pInfo->pSortHandle, pOperator->resultInfo.capacity, pOperator);
if (pBlock != NULL) {
pOperator->resultInfo.totalRows += pBlock->info.rows;
@ -2311,8 +2312,7 @@ SOperatorInfo* createTableMergeScanOperatorInfo(STableScanPhysiNode* pTableScanN
SDataBlockDescNode* pDescNode = pTableScanNode->scan.node.pOutputDataBlockDesc;
int32_t numOfCols = 0;
SArray* pColList =
extractColMatchInfo(pTableScanNode->scan.pScanCols, pDescNode, &numOfCols, COL_MATCH_FROM_COL_ID);
SArray* pColList = extractColMatchInfo(pTableScanNode->scan.pScanCols, pDescNode, &numOfCols, COL_MATCH_FROM_COL_ID);
int32_t code = initQueryTableDataCond(&pInfo->cond, pTableScanNode);
if (code != TSDB_CODE_SUCCESS) {

120
source/libs/function/src/builtins.c
View File

@ -419,7 +419,7 @@ static int32_t translateTopBot(SFunctionNode* pFunc, char* pErrBuf, int32_t len)
return TSDB_CODE_SUCCESS;
}
int32_t topCreateMergePara(SNodeList* pRawParameters, SNode* pPartialRes, SNodeList** pParameters) {
int32_t topBotCreateMergePara(SNodeList* pRawParameters, SNode* pPartialRes, SNodeList** pParameters) {
int32_t code = nodesListMakeAppend(pParameters, pPartialRes);
if (TSDB_CODE_SUCCESS == code) {
code = nodesListStrictAppend(*pParameters, nodesCloneNode(nodesListGetNode(pRawParameters, 1)));
@ -427,65 +427,6 @@ int32_t topCreateMergePara(SNodeList* pRawParameters, SNode* pPartialRes, SNodeL
return TSDB_CODE_SUCCESS;
}
static int32_t translateTopBotImpl(SFunctionNode* pFunc, char* pErrBuf, int32_t len, bool isPartial) {
int32_t numOfParams = LIST_LENGTH(pFunc->pParameterList);
if (isPartial) {
if (2 != numOfParams) {
return invaildFuncParaNumErrMsg(pErrBuf, len, pFunc->functionName);
}
uint8_t para1Type = ((SExprNode*)nodesListGetNode(pFunc->pParameterList, 0))->resType.type;
uint8_t para2Type = ((SExprNode*)nodesListGetNode(pFunc->pParameterList, 1))->resType.type;
if (!IS_NUMERIC_TYPE(para1Type) || !IS_INTEGER_TYPE(para2Type)) {
return invaildFuncParaTypeErrMsg(pErrBuf, len, pFunc->functionName);
}
// param1
SNode* pParamNode1 = nodesListGetNode(pFunc->pParameterList, 1);
if (nodeType(pParamNode1) != QUERY_NODE_VALUE) {
return invaildFuncParaTypeErrMsg(pErrBuf, len, pFunc->functionName);
}
SValueNode* pValue = (SValueNode*)pParamNode1;
if (pValue->node.resType.type != TSDB_DATA_TYPE_BIGINT) {
return invaildFuncParaTypeErrMsg(pErrBuf, len, pFunc->functionName);
}
if (pValue->datum.i < 1 || pValue->datum.i > 100) {
return invaildFuncParaValueErrMsg(pErrBuf, len, pFunc->functionName);
}
pValue->notReserved = true;
// set result type
pFunc->node.resType =
(SDataType){.bytes = getTopBotInfoSize(pValue->datum.i) + VARSTR_HEADER_SIZE, .type = TSDB_DATA_TYPE_BINARY};
} else {
if (1 != numOfParams) {
return invaildFuncParaNumErrMsg(pErrBuf, len, pFunc->functionName);
}
uint8_t para1Type = ((SExprNode*)nodesListGetNode(pFunc->pParameterList, 0))->resType.type;
if (TSDB_DATA_TYPE_BINARY != para1Type) {
return invaildFuncParaTypeErrMsg(pErrBuf, len, pFunc->functionName);
}
// Do nothing. We can only access output of partial functions as input,
// so original input type cannot be obtained, resType will be set same
// as original function input type after merge function created.
}
return TSDB_CODE_SUCCESS;
}
static int32_t translateTopBotPartial(SFunctionNode* pFunc, char* pErrBuf, int32_t len) {
return translateTopBotImpl(pFunc, pErrBuf, len, true);
}
static int32_t translateTopBotMerge(SFunctionNode* pFunc, char* pErrBuf, int32_t len) {
return translateTopBotImpl(pFunc, pErrBuf, len, false);
}
static int32_t translateSpread(SFunctionNode* pFunc, char* pErrBuf, int32_t len) {
if (1 != LIST_LENGTH(pFunc->pParameterList)) {
return invaildFuncParaNumErrMsg(pErrBuf, len, pFunc->functionName);
@ -1735,31 +1676,9 @@ const SBuiltinFuncDefinition funcMgtBuiltins[] = {
.processFunc = topFunction,
.finalizeFunc = topBotFinalize,
.combineFunc = topCombine,
.pPartialFunc = "_top_partial",
.pMergeFunc = "_top_merge",
// .createMergeParaFuc = topCreateMergePara
},
{
.name = "_top_partial",
.type = FUNCTION_TYPE_TOP_PARTIAL,
.classification = FUNC_MGT_AGG_FUNC | FUNC_MGT_SELECT_FUNC | FUNC_MGT_INDEFINITE_ROWS_FUNC,
.translateFunc = translateTopBotPartial,
.getEnvFunc = getTopBotFuncEnv,
.initFunc = topBotFunctionSetup,
.processFunc = topFunction,
.finalizeFunc = topBotPartialFinalize,
.combineFunc = topCombine,
},
{
.name = "_top_merge",
.type = FUNCTION_TYPE_TOP_MERGE,
.classification = FUNC_MGT_AGG_FUNC | FUNC_MGT_SELECT_FUNC | FUNC_MGT_INDEFINITE_ROWS_FUNC,
.translateFunc = translateTopBotMerge,
.getEnvFunc = getTopBotMergeFuncEnv,
.initFunc = functionSetup,
.processFunc = topFunctionMerge,
.finalizeFunc = topBotMergeFinalize,
.combineFunc = topCombine,
.pPartialFunc = "top",
.pMergeFunc = "top",
.createMergeParaFuc = topBotCreateMergePara
},
{
.name = "bottom",
@ -1771,30 +1690,9 @@ const SBuiltinFuncDefinition funcMgtBuiltins[] = {
.processFunc = bottomFunction,
.finalizeFunc = topBotFinalize,
.combineFunc = bottomCombine,
.pPartialFunc = "_bottom_partial",
.pMergeFunc = "_bottom_merge"
},
{
.name = "_bottom_partial",
.type = FUNCTION_TYPE_BOTTOM_PARTIAL,
.classification = FUNC_MGT_AGG_FUNC | FUNC_MGT_SELECT_FUNC | FUNC_MGT_INDEFINITE_ROWS_FUNC,
.translateFunc = translateTopBotPartial,
.getEnvFunc = getTopBotFuncEnv,
.initFunc = topBotFunctionSetup,
.processFunc = bottomFunction,
.finalizeFunc = topBotPartialFinalize,
.combineFunc = bottomCombine,
},
{
.name = "_bottom_merge",
.type = FUNCTION_TYPE_BOTTOM_MERGE,
.classification = FUNC_MGT_AGG_FUNC | FUNC_MGT_SELECT_FUNC | FUNC_MGT_INDEFINITE_ROWS_FUNC,
.translateFunc = translateTopBotMerge,
.getEnvFunc = getTopBotMergeFuncEnv,
.initFunc = functionSetup,
.processFunc = bottomFunctionMerge,
.finalizeFunc = topBotMergeFinalize,
.combineFunc = bottomCombine,
.pPartialFunc = "bottom",
.pMergeFunc = "bottom",
.createMergeParaFuc = topBotCreateMergePara
},
{
.name = "spread",
@ -2524,7 +2422,9 @@ const SBuiltinFuncDefinition funcMgtBuiltins[] = {
.getEnvFunc = getSelectivityFuncEnv, // todo remove this function later.
.initFunc = functionSetup,
.processFunc = NULL,
.finalizeFunc = NULL
.finalizeFunc = NULL,
.pPartialFunc = "_select_value",
.pMergeFunc = "_select_value"
},
{
.name = "_block_dist",

93
source/libs/function/src/builtinsimpl.c
View File

@ -67,8 +67,7 @@ typedef struct STopBotResItem {
typedef struct STopBotRes {
int32_t maxSize;
int16_t type; // store the original input type, used in merge function
int32_t numOfItems;
int16_t type;
STopBotResItem* pItems;
} STopBotRes;
@ -2563,9 +2562,6 @@ int32_t lastFunction(SqlFunctionCtx* pCtx) {
}
static void firstLastTransferInfo(SFirstLastRes* pInput, SFirstLastRes* pOutput, bool isFirst) {
if (!pInput->hasResult) {
return;
}
pOutput->bytes = pInput->bytes;
TSKEY* tsIn = (TSKEY*)(pInput->buf + pInput->bytes);
TSKEY* tsOut = (TSKEY*)(pOutput->buf + pInput->bytes);
@ -2599,7 +2595,9 @@ static int32_t firstLastFunctionMergeImpl(SqlFunctionCtx* pCtx, bool isFirstQuer
firstLastTransferInfo(pInputInfo, pInfo, isFirstQuery);
SET_VAL(GET_RES_INFO(pCtx), 1, 1);
int32_t numOfElems = pInputInfo->hasResult ? 1 : 0;
SET_VAL(GET_RES_INFO(pCtx), numOfElems, 1);
return TSDB_CODE_SUCCESS;
}
@ -2624,6 +2622,7 @@ int32_t firstLastFinalize(SqlFunctionCtx* pCtx, SSDataBlock* pBlock) {
int32_t firstLastPartialFinalize(SqlFunctionCtx* pCtx, SSDataBlock* pBlock) {
SResultRowEntryInfo* pEntryInfo = GET_RES_INFO(pCtx);
SFirstLastRes* pRes = GET_ROWCELL_INTERBUF(GET_RES_INFO(pCtx));
int32_t resultBytes = getFirstLastInfoSize(pRes->bytes);
char* res = taosMemoryCalloc(resultBytes + VARSTR_HEADER_SIZE, sizeof(char));
@ -2872,12 +2871,6 @@ bool getTopBotFuncEnv(SFunctionNode* pFunc, SFuncExecEnv* pEnv) {
return true;
}
bool getTopBotMergeFuncEnv(SFunctionNode* pFunc, SFuncExecEnv* pEnv) {
// intermediate result is binary and length contains VAR header size
pEnv->calcMemSize = pFunc->node.resType.bytes;
return true;
}
bool topBotFunctionSetup(SqlFunctionCtx* pCtx, SResultRowEntryInfo* pResInfo) {
if (!functionSetup(pCtx, pResInfo)) {
return false;
@ -2952,50 +2945,6 @@ int32_t bottomFunction(SqlFunctionCtx* pCtx) {
return TSDB_CODE_SUCCESS;
}
static void topBotTransferInfo(SqlFunctionCtx* pCtx, STopBotRes* pInput, bool isTopQuery) {
for (int32_t i = 0; i < pInput->numOfItems; i++) {
addResult(pCtx, &pInput->pItems[i], pInput->type, isTopQuery);
}
}
int32_t topFunctionMerge(SqlFunctionCtx* pCtx) {
SResultRowEntryInfo* pEntryInfo = GET_RES_INFO(pCtx);
SInputColumnInfoData* pInput = &pCtx->input;
SColumnInfoData* pCol = pInput->pData[0];
ASSERT(pCol->info.type == TSDB_DATA_TYPE_BINARY);
int32_t start = pInput->startRowIndex;
char* data = colDataGetData(pCol, start);
STopBotRes* pInputInfo = (STopBotRes*)varDataVal(data);
STopBotRes* pInfo = GET_ROWCELL_INTERBUF(GET_RES_INFO(pCtx));
pInfo->maxSize = pInputInfo->maxSize;
pInfo->type = pInputInfo->type;
topBotTransferInfo(pCtx, pInputInfo, true);
SET_VAL(GET_RES_INFO(pCtx), pEntryInfo->numOfRes, pEntryInfo->numOfRes);
return TSDB_CODE_SUCCESS;
}
int32_t bottomFunctionMerge(SqlFunctionCtx* pCtx) {
SResultRowEntryInfo* pEntryInfo = GET_RES_INFO(pCtx);
SInputColumnInfoData* pInput = &pCtx->input;
SColumnInfoData* pCol = pInput->pData[0];
ASSERT(pCol->info.type == TSDB_DATA_TYPE_BINARY);
int32_t start = pInput->startRowIndex;
char* data = colDataGetData(pCol, start);
STopBotRes* pInputInfo = (STopBotRes*)varDataVal(data);
STopBotRes* pInfo = GET_ROWCELL_INTERBUF(GET_RES_INFO(pCtx));
pInfo->maxSize = pInputInfo->maxSize;
pInfo->type = pInputInfo->type;
topBotTransferInfo(pCtx, pInputInfo, false);
SET_VAL(GET_RES_INFO(pCtx), pEntryInfo->numOfRes, pEntryInfo->numOfRes);
return TSDB_CODE_SUCCESS;
}
static int32_t topBotResComparFn(const void* p1, const void* p2, const void* param) {
uint16_t type = *(uint16_t*)param;
@ -3044,8 +2993,6 @@ void doAddIntoResult(SqlFunctionCtx* pCtx, void* pData, int32_t rowIndex, SSData
// allocate the buffer and keep the data of this row into the new allocated buffer
pEntryInfo->numOfRes++;
// accumulate number of items for each vgroup, this info is needed for merge
pRes->numOfItems++;
taosheapsort((void*)pItems, sizeof(STopBotResItem), pEntryInfo->numOfRes, (const void*)&type, topBotResComparFn,
!isTopQuery);
} else { // replace the minimum value in the result
@ -3144,7 +3091,7 @@ void copyTupleData(SqlFunctionCtx* pCtx, int32_t rowIndex, const SSDataBlock* pS
releaseBufPage(pCtx->pBuf, pPage);
}
int32_t topBotFinalizeImpl(SqlFunctionCtx* pCtx, SSDataBlock* pBlock, bool isMerge) {
int32_t topBotFinalize(SqlFunctionCtx* pCtx, SSDataBlock* pBlock) {
SResultRowEntryInfo* pEntryInfo = GET_RES_INFO(pCtx);
STopBotRes* pRes = GET_ROWCELL_INTERBUF(pEntryInfo);
@ -3164,39 +3111,13 @@ int32_t topBotFinalizeImpl(SqlFunctionCtx* pCtx, SSDataBlock* pBlock, bool isMer
colDataAppend(pCol, currentRow, (const char*)&pItem->v.i, false);
}
if (!isMerge) {
setSelectivityValue(pCtx, pBlock, &pRes->pItems[i].tuplePos, currentRow);
}
currentRow += 1;
}
return pEntryInfo->numOfRes;
}
int32_t topBotFinalize(SqlFunctionCtx* pCtx, SSDataBlock* pBlock) { return topBotFinalizeImpl(pCtx, pBlock, false); }
int32_t topBotMergeFinalize(SqlFunctionCtx* pCtx, SSDataBlock* pBlock) {
return topBotFinalizeImpl(pCtx, pBlock, true);
}
int32_t topBotPartialFinalize(SqlFunctionCtx* pCtx, SSDataBlock* pBlock) {
SResultRowEntryInfo* pEntryInfo = GET_RES_INFO(pCtx);
STopBotRes* pRes = GET_ROWCELL_INTERBUF(GET_RES_INFO(pCtx));
int32_t resultBytes = getTopBotInfoSize(pRes->maxSize);
char* res = taosMemoryCalloc(resultBytes + VARSTR_HEADER_SIZE, sizeof(char));
memcpy(varDataVal(res), pRes, 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 1;
}
void addResult(SqlFunctionCtx* pCtx, STopBotResItem* pSourceItem, int16_t type, bool isTopQuery) {
SResultRowEntryInfo* pEntryInfo = GET_RES_INFO(pCtx);
STopBotRes* pRes = getTopBotOutputInfo(pCtx);
@ -3211,8 +3132,6 @@ void addResult(SqlFunctionCtx* pCtx, STopBotResItem* pSourceItem, int16_t type,
pItem->tuplePos.pageId = -1;
replaceTupleData(&pItem->tuplePos, &pSourceItem->tuplePos);
pEntryInfo->numOfRes++;
// accumulate number of items for each vgroup, this info is needed for merge
pRes->numOfItems++;
taosheapsort((void*)pItems, sizeof(STopBotResItem), pEntryInfo->numOfRes, (const void*)&type, topBotResComparFn,
!isTopQuery);
} else { // replace the minimum value in the result

5
source/libs/function/src/tudf.c
View File

@ -972,6 +972,11 @@ void releaseUdfFuncHandle(char* udfName) {
}
int32_t cleanUpUdfs() {
int8_t initialized = atomic_load_8(&gUdfdProxy.initialized);
if (!initialized) {
return TSDB_CODE_SUCCESS;
}
uv_mutex_lock(&gUdfdProxy.udfStubsMutex);
int32_t i = 0;
SArray* udfStubs = taosArrayInit(16, sizeof(SUdfcFuncStub));

78
source/libs/planner/src/planOptimizer.c
View File

@ -1082,13 +1082,89 @@ static int32_t partTagsOptimize(SOptimizeContext* pCxt, SLogicSubplan* pLogicSub
return code;
}
static bool eliminateProjOptMayBeOptimized(SLogicNode* pNode) {
//TODO: enable this optimization after new mechanising that map projection and targets of project node
if (NULL != pNode->pParent) {
return false;
}
if (QUERY_NODE_LOGIC_PLAN_PROJECT != nodeType(pNode) || 1 != LIST_LENGTH(pNode->pChildren)) {
return false;
}
SProjectLogicNode* pProjectNode = (SProjectLogicNode*)pNode;
if (-1 != pProjectNode->limit || -1 != pProjectNode->slimit || -1 != pProjectNode->offset || -1 != pProjectNode->soffset) {
return false;
}
SHashObj* pProjColNameHash = taosHashInit(16, taosGetDefaultHashFunction(TSDB_DATA_TYPE_BINARY), true, HASH_NO_LOCK);
SNode* pProjection;
FOREACH(pProjection, pProjectNode->pProjections) {
SExprNode* pExprNode = (SExprNode*)pProjection;
if (QUERY_NODE_COLUMN != nodeType(pExprNode)) {
taosHashCleanup(pProjColNameHash);
return false;
}
char* projColumnName = ((SColumnNode*)pProjection)->colName;
int32_t* pExist = taosHashGet(pProjColNameHash, projColumnName, strlen(projColumnName));
if (NULL != pExist) {
taosHashCleanup(pProjColNameHash);
return false;
} else {
int32_t exist = 1;
taosHashPut(pProjColNameHash, projColumnName, strlen(projColumnName), &exist, sizeof(exist));
}
}
taosHashCleanup(pProjColNameHash);
return true;
}
static int32_t eliminateProjOptimizeImpl(SOptimizeContext* pCxt, SLogicSubplan* pLogicSubplan, SProjectLogicNode* pProjectNode) {
SLogicNode* pChild = (SLogicNode*)nodesListGetNode(pProjectNode->node.pChildren, 0);
SNodeList* pNewChildTargets = nodesMakeList();
SNode* pProjection = NULL;
FOREACH(pProjection, pProjectNode->pProjections) {
SNode* pChildTarget = NULL;
FOREACH(pChildTarget, pChild->pTargets) {
if (strcmp(((SColumnNode*)pProjection)->colName, ((SColumnNode*)pChildTarget)->colName) == 0) {
nodesListAppend(pNewChildTargets, nodesCloneNode(pChildTarget));
break;
}
}
}
nodesDestroyList(pChild->pTargets);
pChild->pTargets = pNewChildTargets;
int32_t code = replaceLogicNode(pLogicSubplan, (SLogicNode*)pProjectNode, pChild);
if (TSDB_CODE_SUCCESS == code) {
NODES_CLEAR_LIST(pProjectNode->node.pChildren);
nodesDestroyNode((SNode*)pProjectNode);
}
return code;
}
static int32_t eliminateProjOptimize(SOptimizeContext* pCxt, SLogicSubplan* pLogicSubplan) {
SProjectLogicNode* pProjectNode =
(SProjectLogicNode*)optFindPossibleNode(pLogicSubplan->pNode, eliminateProjOptMayBeOptimized);
if (NULL == pProjectNode) {
return TSDB_CODE_SUCCESS;
}
return eliminateProjOptimizeImpl(pCxt, pLogicSubplan, pProjectNode);
}
// clang-format off
static const SOptimizeRule optimizeRuleSet[] = {
{.pName = "OptimizeScanData", .optimizeFunc = osdOptimize},
{.pName = "ConditionPushDown", .optimizeFunc = cpdOptimize},
{.pName = "OrderByPrimaryKey", .optimizeFunc = opkOptimize},
{.pName = "SmaIndex", .optimizeFunc = smaOptimize},
{.pName = "PartitionByTags", .optimizeFunc = partTagsOptimize}
{.pName = "PartitionByTags", .optimizeFunc = partTagsOptimize},
{.pName = "EliminateProject", .optimizeFunc = eliminateProjOptimize}
};
// clang-format on

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

@ -107,6 +107,7 @@ int32_t createColumnByRewriteExpr(SNode* pExpr, SNodeList** pList) {
int32_t replaceLogicNode(SLogicSubplan* pSubplan, SLogicNode* pOld, SLogicNode* pNew) {
if (NULL == pOld->pParent) {
pSubplan->pNode = (SLogicNode*)pNew;
pNew->pParent = NULL;
return TSDB_CODE_SUCCESS;
}

10
source/libs/planner/test/planOptimizeTest.cpp
View File

@ -52,3 +52,13 @@ TEST_F(PlanOptimizeTest, orderByPrimaryKey) {
run("SELECT COUNT(*) FROM t1 INTERVAL(10S) ORDER BY _WSTARTTS DESC");
}
TEST_F(PlanOptimizeTest, eliminateProjection) {
useDb("root", "test");
run("SELECT c1, sum(c3) FROM t1 GROUP BY c1");
run("SELECT c1 FROM t1");
run("SELECT * FROM st1");
run("SELECT c1 FROM st1s3");
//run("select 1-abs(c1) from (select unique(c1) c1 from st1s3) order by 1 nulls first");
}

2
source/libs/qworker/src/qworker.c
View File

@ -533,7 +533,7 @@ int32_t qwProcessQuery(QW_FPARAMS_DEF, SQWMsg *qwMsg, int8_t taskType, int8_t ex
atomic_store_8(&ctx->taskType, taskType);
atomic_store_8(&ctx->explain, explain);
/*QW_TASK_DLOGL("subplan json string, len:%d, %s", qwMsg->msgLen, qwMsg->msg);*/
QW_TASK_DLOGL("subplan json string, len:%d, %s", qwMsg->msgLen, qwMsg->msg);
code = qStringToSubplan(qwMsg->msg, &plan);
if (TSDB_CODE_SUCCESS != code) {

30
source/libs/stream/src/stream.c
View File

@ -64,7 +64,7 @@ void streamTriggerByTimer(void* param, void* tmrId) {
atomic_store_8(&pTask->triggerStatus, TASK_TRIGGER_STATUS__IN_ACTIVE);
streamTaskInput(pTask, (SStreamQueueItem*)trigger);
streamLaunchByWrite(pTask, pTask->nodeId, pTask->pMsgCb);
streamLaunchByWrite(pTask, pTask->nodeId);
}
taosTmrReset(streamTriggerByTimer, (int32_t)pTask->triggerParam, pTask, streamEnv.timer, &pTask->timer);
@ -81,7 +81,7 @@ int32_t streamSetupTrigger(SStreamTask* pTask) {
return 0;
}
int32_t streamLaunchByWrite(SStreamTask* pTask, int32_t vgId, SMsgCb* pMsgCb) {
int32_t streamLaunchByWrite(SStreamTask* pTask, int32_t vgId) {
int8_t execStatus = atomic_load_8(&pTask->status);
if (execStatus == TASK_STATUS__IDLE || execStatus == TASK_STATUS__CLOSING) {
SStreamTaskRunReq* pRunReq = rpcMallocCont(sizeof(SStreamTaskRunReq));
@ -96,7 +96,7 @@ int32_t streamLaunchByWrite(SStreamTask* pTask, int32_t vgId, SMsgCb* pMsgCb) {
.pCont = pRunReq,
.contLen = sizeof(SStreamTaskRunReq),
};
tmsgPutToQueue(pMsgCb, FETCH_QUEUE, &msg);
tmsgPutToQueue(pTask->pMsgCb, FETCH_QUEUE, &msg);
}
return 0;
}
@ -136,7 +136,9 @@ int32_t streamTaskEnqueue(SStreamTask* pTask, SStreamDispatchReq* pReq, SRpcMsg*
return status == TASK_INPUT_STATUS__NORMAL ? 0 : -1;
}
int32_t streamProcessDispatchReq(SStreamTask* pTask, SMsgCb* pMsgCb, SStreamDispatchReq* pReq, SRpcMsg* pRsp) {
int32_t streamProcessDispatchReq(SStreamTask* pTask, SStreamDispatchReq* pReq, SRpcMsg* pRsp) {
qInfo("task %d receive dispatch req from node %d task %d", pTask->taskId, pReq->upstreamNodeId, pReq->sourceTaskId);
// 1. handle input
streamTaskEnqueue(pTask, pReq, pRsp);
@ -145,7 +147,7 @@ int32_t streamProcessDispatchReq(SStreamTask* pTask, SMsgCb* pMsgCb, SStreamDisp
// 2.2. executing: return
// 2.3. closing: keep trying
if (pTask->execType != TASK_EXEC__NONE) {
streamExec(pTask, pMsgCb);
streamExec(pTask, pTask->pMsgCb);
} else {
ASSERT(pTask->sinkType != TASK_SINK__NONE);
while (1) {
@ -161,34 +163,38 @@ int32_t streamProcessDispatchReq(SStreamTask* pTask, SMsgCb* pMsgCb, SStreamDisp
// 3.1 check and set status
// 3.2 dispatch / sink
if (pTask->dispatchType != TASK_DISPATCH__NONE) {
streamDispatch(pTask, pMsgCb);
streamDispatch(pTask, pTask->pMsgCb);
}
return 0;
}
int32_t streamProcessDispatchRsp(SStreamTask* pTask, SMsgCb* pMsgCb, SStreamDispatchRsp* pRsp) {
int32_t streamProcessDispatchRsp(SStreamTask* pTask, SStreamDispatchRsp* pRsp) {
ASSERT(pRsp->inputStatus == TASK_OUTPUT_STATUS__NORMAL || pRsp->inputStatus == TASK_OUTPUT_STATUS__BLOCKED);
qInfo("task %d receive dispatch rsp", pTask->taskId);
int8_t old = atomic_exchange_8(&pTask->outputStatus, pRsp->inputStatus);
ASSERT(old == TASK_OUTPUT_STATUS__WAIT);
if (pRsp->inputStatus == TASK_INPUT_STATUS__BLOCKED) {
// TODO: init recover timer
ASSERT(0);
return 0;
}
// continue dispatch
streamDispatch(pTask, pMsgCb);
streamDispatch(pTask, pTask->pMsgCb);
return 0;
}
int32_t streamTaskProcessRunReq(SStreamTask* pTask, SMsgCb* pMsgCb) {
streamExec(pTask, pMsgCb);
int32_t streamProcessRunReq(SStreamTask* pTask) {
streamExec(pTask, pTask->pMsgCb);
if (pTask->dispatchType != TASK_DISPATCH__NONE) {
streamDispatch(pTask, pMsgCb);
streamDispatch(pTask, pTask->pMsgCb);
}
return 0;
}
int32_t streamProcessRecoverReq(SStreamTask* pTask, SMsgCb* pMsgCb, SStreamTaskRecoverReq* pReq, SRpcMsg* pMsg) {
int32_t streamProcessRecoverReq(SStreamTask* pTask, SStreamTaskRecoverReq* pReq, SRpcMsg* pMsg) {
//
return 0;
}

4
source/libs/stream/src/streamDispatch.c
View File

@ -144,7 +144,7 @@ int32_t streamBuildDispatchMsg(SStreamTask* pTask, SStreamDataBlock* data, SRpcM
}
}
ASSERT(vgId != 0);
ASSERT(vgId > 0 || vgId == SNODE_HANDLE);
req.taskId = downstreamTaskId;
qInfo("dispatch from task %d (child id %d) to down stream task %d in vnode %d", pTask->taskId, pTask->childId,
@ -199,6 +199,8 @@ int32_t streamDispatch(SStreamTask* pTask, SMsgCb* pMsgCb) {
}
ASSERT(pBlock->type == STREAM_DATA_TYPE_SSDATA_BLOCK);
qInfo("stream continue dispatching: task %d", pTask->taskId);
SRpcMsg dispatchMsg = {0};
SEpSet* pEpSet = NULL;
if (streamBuildDispatchMsg(pTask, pBlock, &dispatchMsg, &pEpSet) < 0) {

11
source/libs/stream/src/streamExec.c
View File

@ -107,18 +107,19 @@ int32_t streamExec(SStreamTask* pTask, SMsgCb* pMsgCb) {
pRes = streamExecForQall(pTask, pRes);
if (pRes == NULL) goto FAIL;
break;
taosArrayDestroy(pRes);
atomic_store_8(&pTask->status, TASK_STATUS__IDLE);
return 0;
} else if (execStatus == TASK_STATUS__CLOSING) {
continue;
} else if (execStatus == TASK_STATUS__EXECUTING) {
break;
ASSERT(taosArrayGetSize(pRes) == 0);
taosArrayDestroy(pRes);
return 0;
} else {
ASSERT(0);
}
}
if (pRes) taosArrayDestroy(pRes);
atomic_store_8(&pTask->status, TASK_STATUS__IDLE);
return 0;
FAIL:
if (pRes) taosArrayDestroy(pRes);
atomic_store_8(&pTask->status, TASK_STATUS__IDLE);

4
source/libs/stream/src/streamTask.c
View File

@ -36,11 +36,11 @@ int32_t tEncodeSStreamTask(SEncoder* pEncoder, const SStreamTask* pTask) {
if (tEncodeI32(pEncoder, pTask->taskId) < 0) return -1;
if (tEncodeI8(pEncoder, pTask->inputType) < 0) return -1;
if (tEncodeI8(pEncoder, pTask->status) < 0) return -1;
if (tEncodeI8(pEncoder, pTask->sourceType) < 0) return -1;
if (tEncodeI8(pEncoder, pTask->execType) < 0) return -1;
if (tEncodeI8(pEncoder, pTask->sinkType) < 0) return -1;
if (tEncodeI8(pEncoder, pTask->dispatchType) < 0) return -1;
if (tEncodeI16(pEncoder, pTask->dispatchMsgType) < 0) return -1;
if (tEncodeI8(pEncoder, pTask->dataScan) < 0) return -1;
if (tEncodeI32(pEncoder, pTask->childId) < 0) return -1;
if (tEncodeI32(pEncoder, pTask->nodeId) < 0) return -1;
@ -84,11 +84,11 @@ int32_t tDecodeSStreamTask(SDecoder* pDecoder, SStreamTask* pTask) {
if (tDecodeI32(pDecoder, &pTask->taskId) < 0) return -1;
if (tDecodeI8(pDecoder, &pTask->inputType) < 0) return -1;
if (tDecodeI8(pDecoder, &pTask->status) < 0) return -1;
if (tDecodeI8(pDecoder, &pTask->sourceType) < 0) return -1;
if (tDecodeI8(pDecoder, &pTask->execType) < 0) return -1;
if (tDecodeI8(pDecoder, &pTask->sinkType) < 0) return -1;
if (tDecodeI8(pDecoder, &pTask->dispatchType) < 0) return -1;
if (tDecodeI16(pDecoder, &pTask->dispatchMsgType) < 0) return -1;
if (tDecodeI8(pDecoder, &pTask->dataScan) < 0) return -1;
if (tDecodeI32(pDecoder, &pTask->childId) < 0) return -1;
if (tDecodeI32(pDecoder, &pTask->nodeId) < 0) return -1;

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

@ -159,7 +159,8 @@ typedef struct SSyncNode {
SSyncSnapshotSender* senders[TSDB_MAX_REPLICA];
SSyncSnapshotReceiver* pNewNodeReceiver;
// SSnapshotMeta sMeta;
// is config changing
bool changing;
} SSyncNode;
@ -198,7 +199,7 @@ char* syncNode2Str(const SSyncNode* pSyncNode);
void syncNodeEventLog(const SSyncNode* pSyncNode, char* str);
char* syncNode2SimpleStr(const SSyncNode* pSyncNode);
bool syncNodeInConfig(SSyncNode* pSyncNode, const SSyncCfg* config);
void syncNodeUpdateConfig(SSyncNode* pSyncNode, SSyncCfg* newConfig, SyncIndex lastConfigChangeIndex, bool* isDrop);
void syncNodeDoConfigChange(SSyncNode* pSyncNode, SSyncCfg* newConfig, SyncIndex lastConfigChangeIndex);
SSyncNode* syncNodeAcquire(int64_t rid);
void syncNodeRelease(SSyncNode* pNode);
@ -244,6 +245,9 @@ int32_t syncGetSnapshotMetaByIndex(int64_t rid, SyncIndex snapshotIndex, str
void syncStartNormal(int64_t rid);
void syncStartStandBy(int64_t rid);
bool syncNodeCanChange(SSyncNode* pSyncNode);
bool syncNodeCheckNewConfig(SSyncNode* pSyncNode, const SSyncCfg* pNewCfg);
// for debug --------------
void syncNodePrint(SSyncNode* pObj);
void syncNodePrint2(char* s, SSyncNode* pObj);

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

@ -70,6 +70,7 @@ typedef struct SSyncSnapshotReceiver {
void *pWriter;
SyncTerm term;
SyncTerm privateTerm;
SSnapshot snapshot;
SSyncNode *pSyncNode;
SRaftId fromId;
@ -78,7 +79,7 @@ typedef struct SSyncSnapshotReceiver {
SSyncSnapshotReceiver *snapshotReceiverCreate(SSyncNode *pSyncNode, SRaftId fromId);
void snapshotReceiverDestroy(SSyncSnapshotReceiver *pReceiver);
void snapshotReceiverStart(SSyncSnapshotReceiver *pReceiver, SyncTerm privateTerm, SRaftId fromId);
void snapshotReceiverStart(SSyncSnapshotReceiver *pReceiver, SyncTerm privateTerm, SyncSnapshotSend *pBeginMsg);
bool snapshotReceiverIsStart(SSyncSnapshotReceiver *pReceiver);
void snapshotReceiverStop(SSyncSnapshotReceiver *pReceiver, bool apply);
cJSON *snapshotReceiver2Json(SSyncSnapshotReceiver *pReceiver);

370
source/libs/sync/src/syncMain.c
View File

@ -154,10 +154,10 @@ int32_t syncSetStandby(int64_t rid) {
return -1;
}
if (pSyncNode->state == TAOS_SYNC_STATE_LEADER) {
if (pSyncNode->state != TAOS_SYNC_STATE_FOLLOWER) {
taosReleaseRef(tsNodeRefId, pSyncNode->rid);
terrno = TSDB_CODE_SYN_IS_LEADER;
sError("failed to set standby since it is leader, rid:%" PRId64, rid);
sError("failed to set standby since it is not follower, rid:%" PRId64, rid);
return -1;
}
@ -178,6 +178,23 @@ int32_t syncSetStandby(int64_t rid) {
return 0;
}
bool syncNodeCheckNewConfig(SSyncNode* pSyncNode, const SSyncCfg* pNewCfg) {
bool IamInNew = syncNodeInConfig(pSyncNode, pNewCfg);
if (!IamInNew) {
return false;
}
if (pNewCfg->replicaNum > pSyncNode->replicaNum + 1) {
return false;
}
if (pNewCfg->replicaNum < pSyncNode->replicaNum - 1) {
return false;
}
return true;
}
int32_t syncReconfigBuild(int64_t rid, const SSyncCfg* pNewCfg, SRpcMsg* pRpcMsg) {
SSyncNode* pSyncNode = (SSyncNode*)taosAcquireRef(tsNodeRefId, rid);
if (pSyncNode == NULL) {
@ -185,13 +202,12 @@ int32_t syncReconfigBuild(int64_t rid, const SSyncCfg* pNewCfg, SRpcMsg* pRpcMsg
return -1;
}
ASSERT(rid == pSyncNode->rid);
int32_t ret = 0;
bool IamInNew = syncNodeInConfig(pSyncNode, pNewCfg);
if (!IamInNew) {
if (!syncNodeCheckNewConfig(pSyncNode, pNewCfg)) {
taosReleaseRef(tsNodeRefId, pSyncNode->rid);
terrno = TSDB_CODE_SYN_NOT_IN_NEW_CONFIG;
terrno = TSDB_CODE_SYN_NEW_CONFIG_ERROR;
sError("syncNodeCheckNewConfig error");
return -1;
}
@ -215,12 +231,10 @@ int32_t syncReconfig(int64_t rid, const SSyncCfg* pNewCfg) {
}
ASSERT(rid == pSyncNode->rid);
bool IamInNew = syncNodeInConfig(pSyncNode, pNewCfg);
if (!IamInNew) {
sError("sync reconfig error, not in new config");
if (!syncNodeCheckNewConfig(pSyncNode, pNewCfg)) {
taosReleaseRef(tsNodeRefId, pSyncNode->rid);
terrno = TSDB_CODE_SYN_NOT_IN_NEW_CONFIG;
terrno = TSDB_CODE_SYN_NEW_CONFIG_ERROR;
sError("syncNodeCheckNewConfig error");
return -1;
}
@ -425,18 +439,6 @@ const char* syncGetMyRoleStr(int64_t rid) {
return s;
}
int32_t syncGetVgId(int64_t rid) {
SSyncNode* pSyncNode = (SSyncNode*)taosAcquireRef(tsNodeRefId, rid);
if (pSyncNode == NULL) {
return TAOS_SYNC_STATE_ERROR;
}
assert(rid == pSyncNode->rid);
int32_t vgId = pSyncNode->vgId;
taosReleaseRef(tsNodeRefId, pSyncNode->rid);
return vgId;
}
SyncTerm syncGetMyTerm(int64_t rid) {
SSyncNode* pSyncNode = (SSyncNode*)taosAcquireRef(tsNodeRefId, rid);
if (pSyncNode == NULL) {
@ -449,6 +451,18 @@ SyncTerm syncGetMyTerm(int64_t rid) {
return term;
}
SyncGroupId syncGetVgId(int64_t rid) {
SSyncNode* pSyncNode = (SSyncNode*)taosAcquireRef(tsNodeRefId, rid);
if (pSyncNode == NULL) {
return TAOS_SYNC_STATE_ERROR;
}
assert(rid == pSyncNode->rid);
SyncGroupId vgId = pSyncNode->vgId;
taosReleaseRef(tsNodeRefId, pSyncNode->rid);
return vgId;
}
void syncGetEpSet(int64_t rid, SEpSet* pEpSet) {
SSyncNode* pSyncNode = (SSyncNode*)taosAcquireRef(tsNodeRefId, rid);
if (pSyncNode == NULL) {
@ -589,6 +603,26 @@ int32_t syncNodePropose(SSyncNode* pSyncNode, const SRpcMsg* pMsg, bool isWeak)
syncNodeEventLog(pSyncNode, eventLog);
if (pSyncNode->state == TAOS_SYNC_STATE_LEADER) {
if (pSyncNode->changing && pMsg->msgType != TDMT_SYNC_CONFIG_CHANGE_FINISH) {
ret = -1;
terrno = TSDB_CODE_SYN_PROPOSE_NOT_READY;
sError("sync propose not ready, type:%s,%d", TMSG_INFO(pMsg->msgType), pMsg->msgType);
goto _END;
}
// config change
if (pMsg->msgType == TDMT_SYNC_CONFIG_CHANGE) {
if (!syncNodeCanChange(pSyncNode)) {
ret = -1;
terrno = TSDB_CODE_SYN_RECONFIG_NOT_READY;
sError("sync reconfig not ready, type:%s,%d", TMSG_INFO(pMsg->msgType), pMsg->msgType);
goto _END;
}
ASSERT(!pSyncNode->changing);
pSyncNode->changing = true;
}
SRespStub stub;
stub.createTime = taosGetTimestampMs();
stub.rpcMsg = *pMsg;
@ -606,12 +640,16 @@ int32_t syncNodePropose(SSyncNode* pSyncNode, const SRpcMsg* pMsg, bool isWeak)
sError("syncPropose pSyncNode->FpEqMsg is NULL");
}
syncClientRequestDestroy(pSyncMsg);
goto _END;
} else {
ret = -1;
terrno = TSDB_CODE_SYN_NOT_LEADER;
sError("syncPropose not leader, %s", syncUtilState2String(pSyncNode->state));
goto _END;
}
_END:
return ret;
}
@ -825,6 +863,9 @@ SSyncNode* syncNodeOpen(const SSyncInfo* pOldSyncInfo) {
// snapshot receivers
pSyncNode->pNewNodeReceiver = snapshotReceiverCreate(pSyncNode, EMPTY_RAFT_ID);
// is config changing
pSyncNode->changing = false;
// start in syncNodeStart
// start raft
// syncNodeBecomeFollower(pSyncNode);
@ -1253,20 +1294,32 @@ char* syncNode2Str(const SSyncNode* pSyncNode) {
void syncNodeEventLog(const SSyncNode* pSyncNode, char* str) {
int32_t userStrLen = strlen(str);
SSnapshot snapshot = {.data = NULL, .lastApplyIndex = -1, .lastApplyTerm = 0};
if (pSyncNode->pFsm->FpGetSnapshotInfo != NULL) {
pSyncNode->pFsm->FpGetSnapshotInfo(pSyncNode->pFsm, &snapshot);
}
SyncIndex logLastIndex = pSyncNode->pLogStore->syncLogLastIndex(pSyncNode->pLogStore);
if (userStrLen < 256) {
char logBuf[128 + 256];
snprintf(logBuf, sizeof(logBuf),
"vgId:%d %s term:%lu commit:%ld standby:%d replica-num:%d lconfig:%ld sync event %s", pSyncNode->vgId,
syncUtilState2String(pSyncNode->state), pSyncNode->pRaftStore->currentTerm, pSyncNode->commitIndex,
pSyncNode->pRaftCfg->isStandBy, pSyncNode->replicaNum, pSyncNode->pRaftCfg->lastConfigIndex, str);
"vgId:%d, sync %s %s, term:%lu, commit:%ld, lastlog:%ld, lastsnapshot:%ld, standby:%d, replica-num:%d, "
"lconfig:%ld, changing:%d",
pSyncNode->vgId, syncUtilState2String(pSyncNode->state), str, pSyncNode->pRaftStore->currentTerm,
pSyncNode->commitIndex, logLastIndex, snapshot.lastApplyIndex, pSyncNode->pRaftCfg->isStandBy,
pSyncNode->replicaNum, pSyncNode->pRaftCfg->lastConfigIndex, pSyncNode->changing);
sDebug("%s", logBuf);
} else {
int len = 128 + userStrLen;
char* s = (char*)taosMemoryMalloc(len);
snprintf(s, len, "vgId:%d %s term:%lu commit:%ld standby:%d replica-num:%d lconfig:%ld sync event %s",
pSyncNode->vgId, syncUtilState2String(pSyncNode->state), pSyncNode->pRaftStore->currentTerm,
pSyncNode->commitIndex, pSyncNode->pRaftCfg->isStandBy, pSyncNode->replicaNum,
pSyncNode->pRaftCfg->lastConfigIndex, str);
snprintf(s, len,
"vgId:%d, sync %s %s, term:%lu, commit:%ld, lastlog:%ld, lastsnapshot:%ld, standby:%d, replica-num:%d, "
"lconfig:%ld, changing:%d",
pSyncNode->vgId, syncUtilState2String(pSyncNode->state), str, pSyncNode->pRaftStore->currentTerm,
pSyncNode->commitIndex, logLastIndex, snapshot.lastApplyIndex, pSyncNode->pRaftCfg->isStandBy,
pSyncNode->replicaNum, pSyncNode->pRaftCfg->lastConfigIndex, pSyncNode->changing);
sDebug("%s", s);
taosMemoryFree(s);
}
@ -1313,11 +1366,41 @@ bool syncNodeInConfig(SSyncNode* pSyncNode, const SSyncCfg* config) {
return b1;
}
void syncNodeUpdateConfig(SSyncNode* pSyncNode, SSyncCfg* pNewConfig, SyncIndex lastConfigChangeIndex, bool* isDrop) {
void syncNodeDoConfigChange(SSyncNode* pSyncNode, SSyncCfg* pNewConfig, SyncIndex lastConfigChangeIndex) {
SSyncCfg oldConfig = pSyncNode->pRaftCfg->cfg;
pSyncNode->pRaftCfg->cfg = *pNewConfig;
pSyncNode->pRaftCfg->lastConfigIndex = lastConfigChangeIndex;
bool IamInOld = syncNodeInConfig(pSyncNode, &oldConfig);
bool IamInNew = syncNodeInConfig(pSyncNode, pNewConfig);
bool isDrop = false;
bool isAdd = false;
if (IamInOld && !IamInNew) {
isDrop = true;
} else {
isDrop = false;
}
if (!IamInOld && IamInNew) {
isAdd = true;
} else {
isAdd = false;
}
if (IamInNew) {
pSyncNode->pRaftCfg->isStandBy = 0; // change isStandBy to normal
}
if (isDrop) {
pSyncNode->pRaftCfg->isStandBy = 1; // set standby
}
// persist last config index
raftCfgAddConfigIndex(pSyncNode->pRaftCfg, lastConfigChangeIndex);
if (IamInNew) {
//-----------------------------------------
int32_t ret = 0;
// save snapshot senders
@ -1397,8 +1480,9 @@ void syncNodeUpdateConfig(SSyncNode* pSyncNode, SSyncCfg* pNewConfig, SyncIndex
do {
char eventLog[256];
snprintf(eventLog, sizeof(eventLog), "snapshot sender udpate replicaIndex from %d to %d, %s:%d, %p, reset:%d",
oldreplicaIndex, i, host, port, (pSyncNode->senders)[i], reset);
snprintf(eventLog, sizeof(eventLog),
"snapshot sender udpate replicaIndex from %d to %d, %s:%d, %p, reset:%d", oldreplicaIndex, i, host,
port, (pSyncNode->senders)[i], reset);
syncNodeEventLog(pSyncNode, eventLog);
} while (0);
}
@ -1431,28 +1515,39 @@ void syncNodeUpdateConfig(SSyncNode* pSyncNode, SSyncCfg* pNewConfig, SyncIndex
}
}
bool IamInOld = syncNodeInConfig(pSyncNode, &oldConfig);
bool IamInNew = syncNodeInConfig(pSyncNode, pNewConfig);
*isDrop = true;
if (IamInOld && !IamInNew) {
*isDrop = true;
} else {
*isDrop = false;
}
// may be add me to a new raft group
if (IamInOld && IamInNew && oldConfig.replicaNum == 1) {
}
if (IamInNew) {
pSyncNode->pRaftCfg->isStandBy = 0; // change isStandBy to normal
}
// persist cfg
raftCfgPersist(pSyncNode->pRaftCfg);
if (gRaftDetailLog) {
syncNodeLog2("==syncNodeUpdateConfig==", pSyncNode);
char tmpbuf[512];
char* oldStr = syncCfg2SimpleStr(&oldConfig);
char* newStr = syncCfg2SimpleStr(pNewConfig);
snprintf(tmpbuf, sizeof(tmpbuf), "config change from %d to %d, index:%ld, %s --> %s", oldConfig.replicaNum,
pNewConfig->replicaNum, lastConfigChangeIndex, oldStr, newStr);
taosMemoryFree(oldStr);
taosMemoryFree(newStr);
// change isStandBy to normal (election timeout)
if (pSyncNode->state == TAOS_SYNC_STATE_LEADER) {
syncNodeBecomeLeader(pSyncNode, tmpbuf);
} else {
syncNodeBecomeFollower(pSyncNode, tmpbuf);
}
} else {
// persist cfg
raftCfgPersist(pSyncNode->pRaftCfg);
char tmpbuf[512];
char* oldStr = syncCfg2SimpleStr(&oldConfig);
char* newStr = syncCfg2SimpleStr(pNewConfig);
snprintf(tmpbuf, sizeof(tmpbuf), "do not config change from %d to %d, index:%ld, %s --> %s", oldConfig.replicaNum,
pNewConfig->replicaNum, lastConfigChangeIndex, oldStr, newStr);
taosMemoryFree(oldStr);
taosMemoryFree(newStr);
syncNodeEventLog(pSyncNode, tmpbuf);
}
_END:
return;
}
SSyncNode* syncNodeAcquire(int64_t rid) {
@ -2170,9 +2265,54 @@ int32_t syncNodeUpdateNewConfigIndex(SSyncNode* ths, SSyncCfg* pNewCfg) {
return -1;
}
static int32_t syncNodeConfigChange(SSyncNode* ths, SRpcMsg* pRpcMsg, SSyncRaftEntry* pEntry) {
static int32_t syncNodeConfigChangeFinish(SSyncNode* ths, SRpcMsg* pRpcMsg, SSyncRaftEntry* pEntry) {
SyncReconfigFinish* pFinish = syncReconfigFinishFromRpcMsg2(pRpcMsg);
ASSERT(pFinish);
if (ths->pFsm->FpReConfigCb != NULL) {
SReConfigCbMeta cbMeta = {0};
cbMeta.code = 0;
cbMeta.index = pEntry->index;
cbMeta.term = pEntry->term;
cbMeta.seqNum = pEntry->seqNum;
cbMeta.lastConfigIndex = syncNodeGetSnapshotConfigIndex(ths, pEntry->index);
cbMeta.state = ths->state;
cbMeta.currentTerm = ths->pRaftStore->currentTerm;
cbMeta.isWeak = pEntry->isWeak;
cbMeta.flag = 0;
cbMeta.oldCfg = pFinish->oldCfg;
cbMeta.newCfg = pFinish->newCfg;
cbMeta.newCfgIndex = pFinish->newCfgIndex;
cbMeta.newCfgTerm = pFinish->newCfgTerm;
cbMeta.newCfgSeqNum = pFinish->newCfgSeqNum;
ths->pFsm->FpReConfigCb(ths->pFsm, pRpcMsg, cbMeta);
}
// update changing
ths->changing = false;
char tmpbuf[512];
char* oldStr = syncCfg2SimpleStr(&(pFinish->oldCfg));
char* newStr = syncCfg2SimpleStr(&(pFinish->newCfg));
snprintf(tmpbuf, sizeof(tmpbuf), "config change finish from %d to %d, index:%ld, %s --> %s",
pFinish->oldCfg.replicaNum, pFinish->newCfg.replicaNum, pFinish->newCfgIndex, oldStr, newStr);
taosMemoryFree(oldStr);
taosMemoryFree(newStr);
syncNodeEventLog(ths, tmpbuf);
syncReconfigFinishDestroy(pFinish);
return 0;
}
static int32_t syncNodeConfigChange(SSyncNode* ths, SRpcMsg* pRpcMsg, SSyncRaftEntry* pEntry,
SyncReconfigFinish* pFinish) {
// old config
SSyncCfg oldSyncCfg = ths->pRaftCfg->cfg;
// new config
SSyncCfg newSyncCfg;
int32_t ret = syncCfgFromStr(pRpcMsg->pCont, &newSyncCfg);
ASSERT(ret == 0);
@ -2180,77 +2320,28 @@ static int32_t syncNodeConfigChange(SSyncNode* ths, SRpcMsg* pRpcMsg, SSyncRaftE
// update new config myIndex
syncNodeUpdateNewConfigIndex(ths, &newSyncCfg);
bool IamInNew = syncNodeInConfig(ths, &newSyncCfg);
// do config change
syncNodeDoConfigChange(ths, &newSyncCfg, pEntry->index);
/*
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;
IamInNew = true;
break;
}
}
*/
// set pFinish
pFinish->oldCfg = oldSyncCfg;
pFinish->newCfg = newSyncCfg;
pFinish->newCfgIndex = pEntry->index;
pFinish->newCfgTerm = pEntry->term;
pFinish->newCfgSeqNum = pEntry->seqNum;
bool isDrop;
if (IamInNew) {
syncNodeUpdateConfig(ths, &newSyncCfg, pEntry->index, &isDrop);
// change isStandBy to normal
if (!isDrop) {
char tmpbuf[512];
char* oldStr = syncCfg2SimpleStr(&oldSyncCfg);
char* newStr = syncCfg2SimpleStr(&newSyncCfg);
snprintf(tmpbuf, sizeof(tmpbuf), "config change from %d to %d, index:%ld, %s --> %s", oldSyncCfg.replicaNum,
newSyncCfg.replicaNum, pEntry->index, oldStr, newStr);
taosMemoryFree(oldStr);
taosMemoryFree(newStr);
if (ths->state == TAOS_SYNC_STATE_LEADER) {
syncNodeBecomeLeader(ths, tmpbuf);
} else {
syncNodeBecomeFollower(ths, tmpbuf);
}
}
} else {
char tmpbuf[512];
char* oldStr = syncCfg2SimpleStr(&oldSyncCfg);
char* newStr = syncCfg2SimpleStr(&newSyncCfg);
snprintf(tmpbuf, sizeof(tmpbuf), "config change2 from %d to %d, index:%ld, %s --> %s", oldSyncCfg.replicaNum,
newSyncCfg.replicaNum, pEntry->index, oldStr, newStr);
taosMemoryFree(oldStr);
taosMemoryFree(newStr);
syncNodeBecomeFollower(ths, tmpbuf);
return 0;
}
if (gRaftDetailLog) {
char* sOld = syncCfg2Str(&oldSyncCfg);
char* sNew = syncCfg2Str(&newSyncCfg);
sInfo("==config change== 0x11 old:%s new:%s isDrop:%d index:%ld IamInNew:%d \n", sOld, sNew, isDrop, pEntry->index,
IamInNew);
taosMemoryFree(sOld);
taosMemoryFree(sNew);
}
static int32_t syncNodeProposeConfigChangeFinish(SSyncNode* ths, SyncReconfigFinish* pFinish) {
SRpcMsg rpcMsg;
syncReconfigFinish2RpcMsg(pFinish, &rpcMsg);
// always call FpReConfigCb
if (ths->pFsm->FpReConfigCb != NULL) {
SReConfigCbMeta cbMeta = {0};
cbMeta.code = 0;
cbMeta.currentTerm = ths->pRaftStore->currentTerm;
cbMeta.index = pEntry->index;
cbMeta.lastConfigIndex = syncNodeGetSnapshotConfigIndex(ths, pEntry->index);
cbMeta.term = pEntry->term;
cbMeta.newCfg = newSyncCfg;
cbMeta.oldCfg = oldSyncCfg;
cbMeta.seqNum = pEntry->seqNum;
cbMeta.flag = 0x11;
cbMeta.isDrop = isDrop;
ths->pFsm->FpReConfigCb(ths->pFsm, pRpcMsg, cbMeta);
int32_t code = syncNodePropose(ths, &rpcMsg, false);
if (code != 0) {
sError("syncNodeProposeConfigChangeFinish error");
ths->changing = false;
}
return 0;
}
@ -2292,9 +2383,21 @@ int32_t syncNodeCommit(SSyncNode* ths, SyncIndex beginIndex, SyncIndex endIndex,
// config change
if (pEntry->originalRpcType == TDMT_SYNC_CONFIG_CHANGE) {
raftCfgAddConfigIndex(ths->pRaftCfg, pEntry->index);
raftCfgPersist(ths->pRaftCfg);
code = syncNodeConfigChange(ths, &rpcMsg, pEntry);
SyncReconfigFinish* pFinish = syncReconfigFinishBuild(ths->vgId);
ASSERT(pFinish != NULL);
code = syncNodeConfigChange(ths, &rpcMsg, pEntry, pFinish);
ASSERT(code == 0);
if (ths->state == TAOS_SYNC_STATE_LEADER) {
syncNodeProposeConfigChangeFinish(ths, pFinish);
}
syncReconfigFinishDestroy(pFinish);
}
// config change finish
if (pEntry->originalRpcType == TDMT_SYNC_CONFIG_CHANGE_FINISH) {
code = syncNodeConfigChangeFinish(ths, &rpcMsg, pEntry);
ASSERT(code == 0);
}
@ -2345,3 +2448,28 @@ SSyncSnapshotSender* syncNodeGetSnapshotSender(SSyncNode* ths, SRaftId* pDestId)
}
return pSender;
}
bool syncNodeCanChange(SSyncNode* pSyncNode) {
if (pSyncNode->changing) {
sError("sync cannot change");
return false;
}
if ((pSyncNode->commitIndex >= SYNC_INDEX_BEGIN)) {
SyncIndex lastIndex = syncNodeGetLastIndex(pSyncNode);
if (pSyncNode->commitIndex != lastIndex) {
sError("sync cannot change2");
return false;
}
}
for (int i = 0; i < pSyncNode->peersNum; ++i) {
SSyncSnapshotSender* pSender = syncNodeGetSnapshotSender(pSyncNode, &(pSyncNode->peersId)[i]);
if (pSender->start) {
sError("sync cannot change3");
return false;
}
}
return true;
}

129
source/libs/sync/src/syncMessage.c
View File

@ -2228,3 +2228,132 @@ void syncLeaderTransferLog2(char* s, const SyncLeaderTransfer* pMsg) {
taosMemoryFree(serialized);
}
}
// ---------------------------------------------
SyncReconfigFinish* syncReconfigFinishBuild(int32_t vgId) {
uint32_t bytes = sizeof(SyncReconfigFinish);
SyncReconfigFinish* pMsg = taosMemoryMalloc(bytes);
memset(pMsg, 0, bytes);
pMsg->bytes = bytes;
pMsg->vgId = vgId;
pMsg->msgType = TDMT_SYNC_CONFIG_CHANGE_FINISH;
return pMsg;
}
void syncReconfigFinishDestroy(SyncReconfigFinish* pMsg) {
if (pMsg != NULL) {
taosMemoryFree(pMsg);
}
}
void syncReconfigFinishSerialize(const SyncReconfigFinish* pMsg, char* buf, uint32_t bufLen) {
assert(pMsg->bytes <= bufLen);
memcpy(buf, pMsg, pMsg->bytes);
}
void syncReconfigFinishDeserialize(const char* buf, uint32_t len, SyncReconfigFinish* pMsg) {
memcpy(pMsg, buf, len);
assert(len == pMsg->bytes);
}
char* syncReconfigFinishSerialize2(const SyncReconfigFinish* pMsg, uint32_t* len) {
char* buf = taosMemoryMalloc(pMsg->bytes);
assert(buf != NULL);
syncReconfigFinishSerialize(pMsg, buf, pMsg->bytes);
if (len != NULL) {
*len = pMsg->bytes;
}
return buf;
}
SyncReconfigFinish* syncReconfigFinishDeserialize2(const char* buf, uint32_t len) {
uint32_t bytes = *((uint32_t*)buf);
SyncReconfigFinish* pMsg = taosMemoryMalloc(bytes);
assert(pMsg != NULL);
syncReconfigFinishDeserialize(buf, len, pMsg);
assert(len == pMsg->bytes);
return pMsg;
}
void syncReconfigFinish2RpcMsg(const SyncReconfigFinish* pMsg, SRpcMsg* pRpcMsg) {
memset(pRpcMsg, 0, sizeof(*pRpcMsg));
pRpcMsg->msgType = pMsg->msgType;
pRpcMsg->contLen = pMsg->bytes;
pRpcMsg->pCont = rpcMallocCont(pRpcMsg->contLen);
syncReconfigFinishSerialize(pMsg, pRpcMsg->pCont, pRpcMsg->contLen);
}
void syncReconfigFinishFromRpcMsg(const SRpcMsg* pRpcMsg, SyncReconfigFinish* pMsg) {
syncReconfigFinishDeserialize(pRpcMsg->pCont, pRpcMsg->contLen, pMsg);
}
SyncReconfigFinish* syncReconfigFinishFromRpcMsg2(const SRpcMsg* pRpcMsg) {
SyncReconfigFinish* pMsg = syncReconfigFinishDeserialize2(pRpcMsg->pCont, pRpcMsg->contLen);
assert(pMsg != NULL);
return pMsg;
}
cJSON* syncReconfigFinish2Json(const SyncReconfigFinish* 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* pOldCfg = syncCfg2Json((SSyncCfg*)(&(pMsg->oldCfg)));
cJSON* pNewCfg = syncCfg2Json((SSyncCfg*)(&(pMsg->newCfg)));
cJSON_AddItemToObject(pRoot, "oldCfg", pOldCfg);
cJSON_AddItemToObject(pRoot, "newCfg", pNewCfg);
snprintf(u64buf, sizeof(u64buf), "%ld", pMsg->newCfgIndex);
cJSON_AddStringToObject(pRoot, "newCfgIndex", u64buf);
snprintf(u64buf, sizeof(u64buf), "%lu", pMsg->newCfgTerm);
cJSON_AddStringToObject(pRoot, "newCfgTerm", u64buf);
snprintf(u64buf, sizeof(u64buf), "%lu", pMsg->newCfgSeqNum);
cJSON_AddStringToObject(pRoot, "newCfgSeqNum", u64buf);
}
cJSON* pJson = cJSON_CreateObject();
cJSON_AddItemToObject(pJson, "SyncReconfigFinish", pRoot);
return pJson;
}
char* syncReconfigFinish2Str(const SyncReconfigFinish* pMsg) {
cJSON* pJson = syncReconfigFinish2Json(pMsg);
char* serialized = cJSON_Print(pJson);
cJSON_Delete(pJson);
return serialized;
}
// for debug ----------------------
void syncReconfigFinishPrint(const SyncReconfigFinish* pMsg) {
char* serialized = syncReconfigFinish2Str(pMsg);
printf("syncReconfigFinishPrint | len:%lu | %s \n", strlen(serialized), serialized);
fflush(NULL);
taosMemoryFree(serialized);
}
void syncReconfigFinishPrint2(char* s, const SyncReconfigFinish* pMsg) {
char* serialized = syncReconfigFinish2Str(pMsg);
printf("syncReconfigFinishPrint2 | len:%lu | %s | %s \n", strlen(serialized), s, serialized);
fflush(NULL);
taosMemoryFree(serialized);
}
void syncReconfigFinishLog(const SyncReconfigFinish* pMsg) {
char* serialized = syncReconfigFinish2Str(pMsg);
sTrace("syncReconfigFinishLog | len:%lu | %s", strlen(serialized), serialized);
taosMemoryFree(serialized);
}
void syncReconfigFinishLog2(char* s, const SyncReconfigFinish* pMsg) {
if (gRaftDetailLog) {
char* serialized = syncReconfigFinish2Str(pMsg);
sTrace("syncReconfigFinishLog2 | len:%lu | %s | %s", strlen(serialized), s, serialized);
taosMemoryFree(serialized);
}
}

75
source/libs/sync/src/syncSnapshot.c
View File

@ -21,7 +21,8 @@
#include "syncUtil.h"
#include "wal.h"
static void snapshotReceiverDoStart(SSyncSnapshotReceiver *pReceiver, SyncTerm privateTerm, SRaftId fromId);
static void snapshotReceiverDoStart(SSyncSnapshotReceiver *pReceiver, SyncTerm privateTerm,
SyncSnapshotSend *pBeginMsg);
//----------------------------------
SSyncSnapshotSender *snapshotSenderCreate(SSyncNode *pSyncNode, int32_t replicaIndex) {
@ -341,6 +342,10 @@ SSyncSnapshotReceiver *snapshotReceiverCreate(SSyncNode *pSyncNode, SRaftId from
pReceiver->fromId = fromId;
pReceiver->term = pSyncNode->pRaftStore->currentTerm;
pReceiver->privateTerm = 0;
pReceiver->snapshot.data = NULL;
pReceiver->snapshot.lastApplyIndex = -1;
pReceiver->snapshot.lastApplyTerm = 0;
pReceiver->snapshot.lastConfigIndex = -1;
} else {
sInfo("snapshotReceiverCreate cannot create receiver");
@ -358,11 +363,16 @@ void snapshotReceiverDestroy(SSyncSnapshotReceiver *pReceiver) {
bool snapshotReceiverIsStart(SSyncSnapshotReceiver *pReceiver) { return pReceiver->start; }
// begin receive snapshot msg (current term, seq begin)
static void snapshotReceiverDoStart(SSyncSnapshotReceiver *pReceiver, SyncTerm privateTerm, SRaftId fromId) {
static void snapshotReceiverDoStart(SSyncSnapshotReceiver *pReceiver, SyncTerm privateTerm,
SyncSnapshotSend *pBeginMsg) {
pReceiver->term = pReceiver->pSyncNode->pRaftStore->currentTerm;
pReceiver->privateTerm = privateTerm;
pReceiver->ack = SYNC_SNAPSHOT_SEQ_BEGIN;
pReceiver->fromId = fromId;
pReceiver->fromId = pBeginMsg->srcId;
pReceiver->snapshot.lastApplyIndex = pBeginMsg->lastIndex;
pReceiver->snapshot.lastApplyTerm = pBeginMsg->lastTerm;
pReceiver->snapshot.lastConfigIndex = pBeginMsg->lastConfigIndex;
ASSERT(pReceiver->pWriter == NULL);
int32_t ret = pReceiver->pSyncNode->pFsm->FpSnapshotStartWrite(pReceiver->pSyncNode->pFsm, &(pReceiver->pWriter));
@ -371,10 +381,10 @@ static void snapshotReceiverDoStart(SSyncSnapshotReceiver *pReceiver, SyncTerm p
// 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, SRaftId fromId) {
void snapshotReceiverStart(SSyncSnapshotReceiver *pReceiver, SyncTerm privateTerm, SyncSnapshotSend *pBeginMsg) {
if (!snapshotReceiverIsStart(pReceiver)) {
// start
snapshotReceiverDoStart(pReceiver, privateTerm, fromId);
snapshotReceiverDoStart(pReceiver, privateTerm, pBeginMsg);
pReceiver->start = true;
} else {
@ -388,7 +398,7 @@ void snapshotReceiverStart(SSyncSnapshotReceiver *pReceiver, SyncTerm privateTer
pReceiver->pWriter = NULL;
// start again
snapshotReceiverDoStart(pReceiver, privateTerm, fromId);
snapshotReceiverDoStart(pReceiver, privateTerm, pBeginMsg);
pReceiver->start = true;
}
@ -449,6 +459,15 @@ cJSON *snapshotReceiver2Json(SSyncSnapshotReceiver *pReceiver) {
cJSON_AddNumberToObject(pFromId, "vgId", pReceiver->fromId.vgId);
cJSON_AddItemToObject(pRoot, "fromId", pFromId);
snprintf(u64buf, sizeof(u64buf), "%lu", pReceiver->snapshot.lastApplyIndex);
cJSON_AddStringToObject(pRoot, "snapshot.lastApplyIndex", u64buf);
snprintf(u64buf, sizeof(u64buf), "%lu", pReceiver->snapshot.lastApplyTerm);
cJSON_AddStringToObject(pRoot, "snapshot.lastApplyTerm", u64buf);
snprintf(u64buf, sizeof(u64buf), "%lu", pReceiver->snapshot.lastConfigIndex);
cJSON_AddStringToObject(pRoot, "snapshot.lastConfigIndex", u64buf);
snprintf(u64buf, sizeof(u64buf), "%lu", pReceiver->term);
cJSON_AddStringToObject(pRoot, "term", u64buf);
@ -477,8 +496,9 @@ char *snapshotReceiver2SimpleStr(SSyncSnapshotReceiver *pReceiver, char *event)
uint16_t port;
syncUtilU642Addr(fromId.addr, host, sizeof(host), &port);
snprintf(s, len, "%s %p start:%d ack:%d term:%lu pterm:%lu %s:%d ", event, pReceiver, pReceiver->start,
pReceiver->ack, pReceiver->term, pReceiver->privateTerm, host, port);
snprintf(s, len, "%s %p start:%d ack:%d term:%lu pterm:%lu from:%s:%d laindex:%ld laterm:%lu lcindex:%ld", event,
pReceiver, pReceiver->start, pReceiver->ack, pReceiver->term, pReceiver->privateTerm, host, port,
pReceiver->snapshot.lastApplyIndex, pReceiver->snapshot.lastApplyTerm, pReceiver->snapshot.lastConfigIndex);
return s;
}
@ -495,7 +515,7 @@ int32_t syncNodeOnSnapshotSendCb(SSyncNode *pSyncNode, SyncSnapshotSend *pMsg) {
if (pMsg->term == pSyncNode->pRaftStore->currentTerm) {
if (pMsg->seq == SYNC_SNAPSHOT_SEQ_BEGIN) {
// begin
snapshotReceiverStart(pReceiver, pMsg->privateTerm, pMsg->srcId);
snapshotReceiverStart(pReceiver, pMsg->privateTerm, pMsg);
pReceiver->ack = pMsg->seq;
needRsp = true;
@ -519,42 +539,9 @@ int32_t syncNodeOnSnapshotSendCb(SSyncNode *pSyncNode, SyncSnapshotSend *pMsg) {
// update new config myIndex
SSyncCfg newSyncCfg = pMsg->lastConfig;
syncNodeUpdateNewConfigIndex(pSyncNode, &newSyncCfg);
bool IamInNew = syncNodeInConfig(pSyncNode, &newSyncCfg);
bool isDrop = false;
if (IamInNew) {
char eventLog[128];
snprintf(eventLog, sizeof(eventLog),
"update config by snapshot, lastIndex:%ld, lastTerm:%lu, lastConfigIndex:%ld", pMsg->lastIndex,
pMsg->lastTerm, pMsg->lastConfigIndex);
syncNodeEventLog(pSyncNode, eventLog);
syncNodeUpdateConfig(pSyncNode, &newSyncCfg, pMsg->lastConfigIndex, &isDrop);
} else {
char eventLog[128];
snprintf(eventLog, sizeof(eventLog),
"do not update config by snapshot, not in new, lastIndex:%ld, lastTerm:%lu, lastConfigIndex:%ld",
pMsg->lastIndex, pMsg->lastTerm, pMsg->lastConfigIndex);
syncNodeEventLog(pSyncNode, eventLog);
}
// change isStandBy to normal
if (!isDrop) {
char tmpbuf[512];
char *oldStr = syncCfg2SimpleStr(&oldSyncCfg);
char *newStr = syncCfg2SimpleStr(&newSyncCfg);
snprintf(tmpbuf, sizeof(tmpbuf), "config change3 from %d to %d, index:%ld, %s --> %s",
oldSyncCfg.replicaNum, newSyncCfg.replicaNum, pMsg->lastConfigIndex, oldStr, newStr);
taosMemoryFree(oldStr);
taosMemoryFree(newStr);
if (pSyncNode->state == TAOS_SYNC_STATE_LEADER) {
syncNodeBecomeLeader(pSyncNode, tmpbuf);
} else {
syncNodeBecomeFollower(pSyncNode, tmpbuf);
}
}
// do config change
syncNodeDoConfigChange(pSyncNode, &newSyncCfg, pMsg->lastConfigIndex);
}
SSnapshot snapshot;

12
source/libs/sync/src/syncUtil.c
View File

@ -261,23 +261,29 @@ bool syncUtilIsData(tmsg_t msgType) {
#endif
bool syncUtilUserPreCommit(tmsg_t msgType) {
if (msgType != TDMT_SYNC_NOOP && msgType != TDMT_SYNC_CONFIG_CHANGE && msgType != TDMT_SYNC_LEADER_TRANSFER) {
if (msgType != TDMT_SYNC_NOOP && msgType != TDMT_SYNC_CONFIG_CHANGE && msgType != TDMT_SYNC_CONFIG_CHANGE_FINISH &&
msgType != TDMT_SYNC_LEADER_TRANSFER) {
return true;
}
return false;
}
bool syncUtilUserCommit(tmsg_t msgType) {
if (msgType != TDMT_SYNC_NOOP && msgType != TDMT_SYNC_CONFIG_CHANGE && msgType != TDMT_SYNC_LEADER_TRANSFER) {
if (msgType != TDMT_SYNC_NOOP && msgType != TDMT_SYNC_CONFIG_CHANGE && msgType != TDMT_SYNC_CONFIG_CHANGE_FINISH &&
msgType != TDMT_SYNC_LEADER_TRANSFER) {
return true;
}
return false;
}
bool syncUtilUserRollback(tmsg_t msgType) {
if (msgType != TDMT_SYNC_NOOP && msgType != TDMT_SYNC_CONFIG_CHANGE && msgType != TDMT_SYNC_LEADER_TRANSFER) {
if (msgType != TDMT_SYNC_NOOP && msgType != TDMT_SYNC_CONFIG_CHANGE && msgType != TDMT_SYNC_CONFIG_CHANGE_FINISH &&
msgType != TDMT_SYNC_LEADER_TRANSFER) {
return true;
}
return false;
}

14
source/libs/sync/test/CMakeLists.txt
View File

@ -49,6 +49,7 @@ add_executable(syncRaftLogTest "")
add_executable(syncRaftLogTest2 "")
add_executable(syncRaftLogTest3 "")
add_executable(syncLeaderTransferTest "")
add_executable(syncReconfigFinishTest "")
target_sources(syncTest
@ -255,6 +256,10 @@ target_sources(syncLeaderTransferTest
PRIVATE
"syncLeaderTransferTest.cpp"
)
target_sources(syncReconfigFinishTest
PRIVATE
"syncReconfigFinishTest.cpp"
)
target_include_directories(syncTest
@ -512,6 +517,11 @@ target_include_directories(syncLeaderTransferTest
"${TD_SOURCE_DIR}/include/libs/sync"
"${CMAKE_CURRENT_SOURCE_DIR}/../inc"
)
target_include_directories(syncReconfigFinishTest
PUBLIC
"${TD_SOURCE_DIR}/include/libs/sync"
"${CMAKE_CURRENT_SOURCE_DIR}/../inc"
)
target_link_libraries(syncTest
@ -718,6 +728,10 @@ target_link_libraries(syncLeaderTransferTest
sync
gtest_main
)
target_link_libraries(syncReconfigFinishTest
sync
gtest_main
)
enable_testing()

4
source/libs/sync/test/syncConfigChangeSnapshotTest.cpp
View File

@ -147,8 +147,8 @@ int32_t SnapshotDoWrite(struct SSyncFSM* pFsm, void* pWriter, void* pBuf, int32_
void RestoreFinishCb(struct SSyncFSM* pFsm) { sTrace("==callback== ==RestoreFinishCb=="); }
void ReConfigCb(struct SSyncFSM* pFsm, const SRpcMsg* pMsg, 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);
sTrace("==callback== ==ReConfigCb== flag:0x%lX, index:%ld, code:%d, currentTerm:%lu, term:%lu", cbMeta.flag,
cbMeta.index, cbMeta.code, cbMeta.currentTerm, cbMeta.term);
}
SSyncFSM* createFsm() {

4
source/libs/sync/test/syncConfigChangeTest.cpp
View File

@ -78,8 +78,8 @@ int32_t GetSnapshotCb(struct SSyncFSM* pFsm, SSnapshot* pSnapshot) {
void RestoreFinishCb(struct SSyncFSM* pFsm) { sTrace("==callback== ==RestoreFinishCb=="); }
void ReConfigCb(struct SSyncFSM* pFsm, const SRpcMsg* pMsg, 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);
sTrace("==callback== ==ReConfigCb== flag:0x%lX, index:%ld, code:%d, currentTerm:%lu, term:%lu", cbMeta.flag,
cbMeta.index, cbMeta.code, cbMeta.currentTerm, cbMeta.term);
}
SSyncFSM* createFsm() {

135
source/libs/sync/test/syncReconfigFinishTest.cpp Normal file
View File

@ -0,0 +1,135 @@
#include <gtest/gtest.h>
#include <stdio.h>
#include "syncIO.h"
#include "syncInt.h"
#include "syncMessage.h"
#include "syncUtil.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");
}
SSyncCfg* createSyncOldCfg() {
SSyncCfg* pCfg = (SSyncCfg*)taosMemoryMalloc(sizeof(SSyncCfg));
memset(pCfg, 0, sizeof(SSyncCfg));
pCfg->replicaNum = 3;
pCfg->myIndex = 1;
for (int i = 0; i < pCfg->replicaNum; ++i) {
((pCfg->nodeInfo)[i]).nodePort = i * 100;
snprintf(((pCfg->nodeInfo)[i]).nodeFqdn, sizeof(((pCfg->nodeInfo)[i]).nodeFqdn), "100.200.300.%d", i);
}
return pCfg;
}
SSyncCfg* createSyncNewCfg() {
SSyncCfg* pCfg = (SSyncCfg*)taosMemoryMalloc(sizeof(SSyncCfg));
memset(pCfg, 0, sizeof(SSyncCfg));
pCfg->replicaNum = 3;
pCfg->myIndex = 1;
for (int i = 0; i < pCfg->replicaNum; ++i) {
((pCfg->nodeInfo)[i]).nodePort = i * 100;
snprintf(((pCfg->nodeInfo)[i]).nodeFqdn, sizeof(((pCfg->nodeInfo)[i]).nodeFqdn), "500.600.700.%d", i);
}
return pCfg;
}
SyncReconfigFinish *createMsg() {
SyncReconfigFinish *pMsg = syncReconfigFinishBuild(1234);
SSyncCfg* pOld = createSyncOldCfg();
SSyncCfg* pNew = createSyncNewCfg();
pMsg->oldCfg = *pOld;
pMsg->newCfg = *pNew;
pMsg->newCfgIndex = 11;
pMsg->newCfgTerm = 22;
pMsg->newCfgSeqNum = 33;
taosMemoryFree(pOld);
taosMemoryFree(pNew);
return pMsg;
}
void test1() {
SyncReconfigFinish *pMsg = createMsg();
syncReconfigFinishLog2((char *)"test1:", pMsg);
syncReconfigFinishDestroy(pMsg);
}
void test2() {
SyncReconfigFinish *pMsg = createMsg();
uint32_t len = pMsg->bytes;
char * serialized = (char *)taosMemoryMalloc(len);
syncReconfigFinishSerialize(pMsg, serialized, len);
SyncReconfigFinish *pMsg2 = syncReconfigFinishBuild(1000);
syncReconfigFinishDeserialize(serialized, len, pMsg2);
syncReconfigFinishLog2((char *)"test2: syncReconfigFinishSerialize -> syncReconfigFinishDeserialize ", pMsg2);
taosMemoryFree(serialized);
syncReconfigFinishDestroy(pMsg);
syncReconfigFinishDestroy(pMsg2);
}
void test3() {
SyncReconfigFinish *pMsg = createMsg();
uint32_t len;
char * serialized = syncReconfigFinishSerialize2(pMsg, &len);
SyncReconfigFinish *pMsg2 = syncReconfigFinishDeserialize2(serialized, len);
syncReconfigFinishLog2((char *)"test3: SyncReconfigFinishSerialize2 -> syncReconfigFinishDeserialize2 ", pMsg2);
taosMemoryFree(serialized);
syncReconfigFinishDestroy(pMsg);
syncReconfigFinishDestroy(pMsg2);
}
void test4() {
SyncReconfigFinish *pMsg = createMsg();
SRpcMsg rpcMsg;
syncReconfigFinish2RpcMsg(pMsg, &rpcMsg);
SyncReconfigFinish *pMsg2 = (SyncReconfigFinish *)taosMemoryMalloc(rpcMsg.contLen);
syncReconfigFinishFromRpcMsg(&rpcMsg, pMsg2);
syncReconfigFinishLog2((char *)"test4: syncReconfigFinish2RpcMsg -> syncReconfigFinishFromRpcMsg ", pMsg2);
rpcFreeCont(rpcMsg.pCont);
syncReconfigFinishDestroy(pMsg);
syncReconfigFinishDestroy(pMsg2);
}
void test5() {
SyncReconfigFinish *pMsg = createMsg();
SRpcMsg rpcMsg;
syncReconfigFinish2RpcMsg(pMsg, &rpcMsg);
SyncReconfigFinish *pMsg2 = syncReconfigFinishFromRpcMsg2(&rpcMsg);
syncReconfigFinishLog2((char *)"test5: syncReconfigFinish2RpcMsg -> syncReconfigFinishFromRpcMsg2 ", pMsg2);
rpcFreeCont(rpcMsg.pCont);
syncReconfigFinishDestroy(pMsg);
syncReconfigFinishDestroy(pMsg2);
}
int main() {
gRaftDetailLog = true;
tsAsyncLog = 0;
sDebugFlag = DEBUG_TRACE + DEBUG_SCREEN + DEBUG_FILE;
logTest();
test1();
test2();
test3();
test4();
test5();
return 0;
}

4
source/libs/sync/test/syncTestTool.cpp
View File

@ -148,8 +148,8 @@ void RestoreFinishCb(struct SSyncFSM* pFsm) { sTrace("==callback== ==RestoreFini
void ReConfigCb(struct SSyncFSM* pFsm, const SRpcMsg* pMsg, SReConfigCbMeta cbMeta) {
char* s = syncCfg2Str(&(cbMeta.newCfg));
sTrace("==callback== ==ReConfigCb== flag:0x%lX, isDrop:%d, index:%ld, code:%d, currentTerm:%lu, term:%lu, newCfg:%s",
cbMeta.flag, cbMeta.isDrop, cbMeta.index, cbMeta.code, cbMeta.currentTerm, cbMeta.term, s);
sTrace("==callback== ==ReConfigCb== flag:0x%lX, index:%ld, code:%d, currentTerm:%lu, term:%lu, newCfg:%s",
cbMeta.flag, cbMeta.index, cbMeta.code, cbMeta.currentTerm, cbMeta.term, s);
taosMemoryFree(s);
}

516
source/libs/tdb/src/db/tdbBtree.c
View File

@ -17,6 +17,7 @@
#define TDB_BTREE_ROOT 0x1
#define TDB_BTREE_LEAF 0x2
#define TDB_BTREE_OVFL 0x4
struct SBTree {
SPgno root;
@ -38,9 +39,11 @@ struct SBTree {
#define TDB_BTREE_PAGE_SET_FLAGS(PAGE, flags) ((PAGE)->pData[0] = (flags))
#define TDB_BTREE_PAGE_IS_ROOT(PAGE) (TDB_BTREE_PAGE_GET_FLAGS(PAGE) & TDB_BTREE_ROOT)
#define TDB_BTREE_PAGE_IS_LEAF(PAGE) (TDB_BTREE_PAGE_GET_FLAGS(PAGE) & TDB_BTREE_LEAF)
#define TDB_BTREE_PAGE_IS_OVFL(PAGE) (TDB_BTREE_PAGE_GET_FLAGS(PAGE) & TDB_BTREE_OVFL)
#define TDB_BTREE_ASSERT_FLAG(flags) \
ASSERT(TDB_FLAG_IS(flags, TDB_BTREE_ROOT) || TDB_FLAG_IS(flags, TDB_BTREE_LEAF) || \
TDB_FLAG_IS(flags, TDB_BTREE_ROOT | TDB_BTREE_LEAF) || TDB_FLAG_IS(flags, 0))
TDB_FLAG_IS(flags, TDB_BTREE_ROOT | TDB_BTREE_LEAF) || TDB_FLAG_IS(flags, 0) || \
TDB_FLAG_IS(flags, TDB_BTREE_OVFL))
#pragma pack(push, 1)
typedef struct {
@ -62,10 +65,10 @@ static int tdbDefaultKeyCmprFn(const void *pKey1, int keyLen1, const void *pKey2
static int tdbBtreeOpenImpl(SBTree *pBt);
static int tdbBtreeInitPage(SPage *pPage, void *arg, int init);
static int tdbBtreeEncodeCell(SPage *pPage, const void *pKey, int kLen, const void *pVal, int vLen, SCell *pCell,
int *szCell);
static int tdbBtreeDecodeCell(SPage *pPage, const SCell *pCell, SCellDecoder *pDecoder);
int *szCell, TXN *pTxn, SBTree *pBt);
static int tdbBtreeDecodeCell(SPage *pPage, const SCell *pCell, SCellDecoder *pDecoder, TXN *pTxn, SBTree *pBt);
static int tdbBtreeBalance(SBTC *pBtc);
static int tdbBtreeCellSize(const SPage *pPage, SCell *pCell);
static int tdbBtreeCellSize(const SPage *pPage, SCell *pCell, int dropOfp, TXN *pTxn, SBTree *pBt);
static int tdbBtcMoveDownward(SBTC *pBtc);
static int tdbBtcMoveUpward(SBTC *pBtc);
@ -255,7 +258,7 @@ int tdbBtreePGet(SBTree *pBt, const void *pKey, int kLen, void **ppKey, int *pkL
}
pCell = tdbPageGetCell(btc.pPage, btc.idx);
tdbBtreeDecodeCell(btc.pPage, pCell, &cd);
tdbBtreeDecodeCell(btc.pPage, pCell, &cd, btc.pTxn, pBt);
if (ppKey) {
pTKey = tdbRealloc(*ppKey, cd.kLen);
@ -281,6 +284,14 @@ int tdbBtreePGet(SBTree *pBt, const void *pKey, int kLen, void **ppKey, int *pkL
memcpy(*ppVal, cd.pVal, cd.vLen);
}
if (TDB_CELLDECODER_FREE_KEY(&cd)) {
tdbFree(cd.pKey);
}
if (TDB_CELLDECODER_FREE_VAL(&cd)) {
tdbFree(cd.pVal);
}
tdbBtcClose(&btc);
return 0;
@ -375,6 +386,11 @@ static int tdbBtreeInitPage(SPage *pPage, void *arg, int init) {
pPage->vLen = pBt->valLen;
pPage->maxLocal = pBt->maxLeaf;
pPage->minLocal = pBt->minLeaf;
} else if (TDB_BTREE_PAGE_IS_OVFL(pPage)) {
pPage->kLen = pBt->keyLen;
pPage->vLen = pBt->valLen;
pPage->maxLocal = tdbPageCapacity(pBt->pageSize, sizeof(SIntHdr));
pPage->minLocal = pBt->minLocal;
} else {
pPage->kLen = pBt->keyLen;
pPage->vLen = sizeof(SPgno);
@ -499,7 +515,7 @@ static int tdbBtreeBalanceNonRoot(SBTree *pBt, SPage *pParent, int idx, TXN *pTx
for (int i = 0; i < nOlds; i++) {
if (sIdx + i < TDB_PAGE_TOTAL_CELLS(pParent)) {
pCell = tdbPageGetCell(pParent, sIdx + i);
szDivCell[i] = tdbBtreeCellSize(pParent, pCell);
szDivCell[i] = tdbBtreeCellSize(pParent, pCell, 0, NULL, NULL);
pDivCell[i] = tdbOsMalloc(szDivCell[i]);
memcpy(pDivCell[i], pCell, szDivCell[i]);
}
@ -524,7 +540,7 @@ static int tdbBtreeBalanceNonRoot(SBTree *pBt, SPage *pParent, int idx, TXN *pTx
for (int i = 0; i < nOlds; i++) {
nCells = TDB_PAGE_TOTAL_CELLS(pParent);
if (sIdx < nCells) {
tdbPageDropCell(pParent, sIdx);
tdbPageDropCell(pParent, sIdx, pTxn, pBt);
} else {
((SIntHdr *)pParent->pData)->pgno = 0;
}
@ -582,7 +598,7 @@ static int tdbBtreeBalanceNonRoot(SBTree *pBt, SPage *pParent, int idx, TXN *pTx
for (;;) {
pCell = tdbPageGetCell(pOlds[infoNews[iNew - 1].iPage], infoNews[iNew - 1].oIdx);
szLCell = tdbBtreeCellSize(pOlds[infoNews[iNew - 1].iPage], pCell);
szLCell = tdbBtreeCellSize(pOlds[infoNews[iNew - 1].iPage], pCell, 0, NULL, NULL);
if (!childNotLeaf) {
szRCell = szLCell;
} else {
@ -600,7 +616,7 @@ static int tdbBtreeBalanceNonRoot(SBTree *pBt, SPage *pParent, int idx, TXN *pTx
}
pCell = tdbPageGetCell(pPage, oIdx);
szRCell = tdbBtreeCellSize(pPage, pCell);
szRCell = tdbBtreeCellSize(pPage, pCell, 0, NULL, NULL);
}
ASSERT(infoNews[iNew - 1].cnt > 0);
@ -687,7 +703,7 @@ static int tdbBtreeBalanceNonRoot(SBTree *pBt, SPage *pParent, int idx, TXN *pTx
for (int oIdx = 0; oIdx < TDB_PAGE_TOTAL_CELLS(pPage); oIdx++) {
pCell = tdbPageGetCell(pPage, oIdx);
szCell = tdbBtreeCellSize(pPage, pCell);
szCell = tdbBtreeCellSize(pPage, pCell, 0, NULL, NULL);
ASSERT(nNewCells <= infoNews[iNew].cnt);
ASSERT(iNew < nNews);
@ -703,14 +719,14 @@ static int tdbBtreeBalanceNonRoot(SBTree *pBt, SPage *pParent, int idx, TXN *pTx
if (iNew == nNews - 1 && pIntHdr->pgno == 0) {
pIntHdr->pgno = TDB_PAGE_PGNO(pNews[iNew]);
} else {
tdbBtreeDecodeCell(pPage, pCell, &cd);
tdbBtreeDecodeCell(pPage, pCell, &cd, pTxn, pBt);
// TODO: pCell here may be inserted as an overflow cell, handle it
SCell *pNewCell = tdbOsMalloc(cd.kLen + 9);
int szNewCell;
SPgno pgno;
pgno = TDB_PAGE_PGNO(pNews[iNew]);
tdbBtreeEncodeCell(pParent, cd.pKey, cd.kLen, (void *)&pgno, sizeof(SPgno), pNewCell, &szNewCell);
tdbBtreeEncodeCell(pParent, cd.pKey, cd.kLen, (void *)&pgno, sizeof(SPgno), pNewCell, &szNewCell, pTxn, pBt);
tdbPageInsertCell(pParent, sIdx++, pNewCell, szNewCell, 0);
tdbOsFree(pNewCell);
}
@ -846,13 +862,50 @@ static int tdbBtreeBalance(SBTC *pBtc) {
}
// TDB_BTREE_BALANCE
static int tdbFetchOvflPage(SPager *pPager, SPgno *pPgno, SPage **ppOfp, TXN *pTxn, SBTree *pBt) {
int ret = 0;
*pPgno = 0;
SBtreeInitPageArg iArg;
iArg.pBt = pBt;
iArg.flags = TDB_FLAG_ADD(0, TDB_BTREE_OVFL);
ret = tdbPagerFetchPage(pPager, pPgno, ppOfp, tdbBtreeInitPage, &iArg, pTxn);
if (ret < 0) {
return -1;
}
// mark dirty
ret = tdbPagerWrite(pPager, *ppOfp);
if (ret < 0) {
ASSERT(0);
return -1;
}
return ret;
}
static int tdbLoadOvflPage(SPager *pPager, SPgno *pPgno, SPage **ppOfp, TXN *pTxn, SBTree *pBt) {
int ret = 0;
SBtreeInitPageArg iArg;
iArg.pBt = pBt;
iArg.flags = TDB_FLAG_ADD(0, TDB_BTREE_OVFL);
ret = tdbPagerFetchPage(pPager, pPgno, ppOfp, tdbBtreeInitPage, &iArg, pTxn);
if (ret < 0) {
return -1;
}
return ret;
}
// TDB_BTREE_CELL =====================
static int tdbBtreeEncodePayload(SPage *pPage, SCell *pCell, int nHeader, const void *pKey, int kLen, const void *pVal,
int vLen, int *szPayload) {
int nPayload;
int vLen, int *szPayload, TXN *pTxn, SBTree *pBt) {
int ret = 0;
int nPayload = kLen + vLen;
int maxLocal = pPage->maxLocal;
nPayload = kLen + vLen;
if (nPayload + nHeader <= pPage->maxLocal) {
if (nPayload + nHeader <= maxLocal) {
// no overflow page is needed
memcpy(pCell + nHeader, pKey, kLen);
if (pVal) {
@ -861,18 +914,190 @@ static int tdbBtreeEncodePayload(SPage *pPage, SCell *pCell, int nHeader, const
*szPayload = nPayload;
return 0;
} else {
// handle overflow case
// calc local storage size
int minLocal = pPage->minLocal;
int surplus = minLocal + (nPayload + nHeader - minLocal) % (maxLocal - sizeof(SPgno));
int nLocal = surplus <= maxLocal ? surplus : minLocal;
//int ofpCap = tdbPageCapacity(pBt->pageSize, sizeof(SIntHdr));
// fetch a new ofp and make it dirty
SPgno pgno = 0;
SPage *ofp, *nextOfp;
ret = tdbFetchOvflPage(pPage->pPager, &pgno, &ofp, pTxn, pBt);
if (ret < 0) {
return -1;
}
{
// TODO: handle overflow case
ASSERT(0);
// local buffer for cell
SCell *pBuf = tdbRealloc(NULL, pBt->pageSize);
if (pBuf == NULL) {
return -1;
}
int nLeft = nPayload;
int bytes;
int lastPage = 0;
if (nLocal >= kLen + 4) {
// pack key to local
memcpy(pCell + nHeader, pKey, kLen);
nLeft -= kLen;
// pack partial val to local if any space left
if (nLocal > kLen + 4) {
memcpy(pCell + nHeader + kLen, pVal, nLocal - kLen - sizeof(SPgno));
nLeft -= nLocal - kLen - sizeof(SPgno);
}
// pack nextPgno
memcpy(pCell + nHeader + nPayload - nLeft, &pgno, sizeof(pgno));
// pack left val data to ovpages
do {
lastPage = 0;
if (nLeft <= ofp->maxLocal - sizeof(SPgno)) {
bytes = nLeft;
lastPage = 1;
} else {
bytes = ofp->maxLocal - sizeof(SPgno);
}
// fetch next ofp if not last page
if (!lastPage) {
// fetch a new ofp and make it dirty
ret = tdbFetchOvflPage(pPage->pPager, &pgno, &nextOfp, pTxn, pBt);
if (ret < 0) {
tdbFree(pBuf);
return -1;
}
} else {
pgno = 0;
}
memcpy(pBuf, ((SCell *)pVal) + vLen - nLeft, bytes);
memcpy(pBuf + bytes, &pgno, sizeof(pgno));
ret = tdbPageInsertCell(ofp, 0, pBuf, bytes + sizeof(pgno), 0);
if (ret < 0) {
tdbFree(pBuf);
return -1;
}
ofp = nextOfp;
nLeft -= bytes;
} while (nLeft > 0);
} else {
int nLeftKey = kLen;
// pack partial key and nextPgno
memcpy(pCell + nHeader, pKey, nLocal - 4);
nLeft -= nLocal - 4;
nLeftKey -= nLocal -4;
memcpy(pCell + nHeader + nLocal - 4, &pgno, sizeof(pgno));
int lastKeyPageSpace = 0;
// pack left key & val to ovpages
do {
// cal key to cpy
int lastKeyPage = 0;
if (nLeftKey <= ofp->maxLocal - sizeof(SPgno)) {
bytes = nLeftKey;
lastKeyPage = 1;
lastKeyPageSpace = ofp->maxLocal - sizeof(SPgno) - nLeftKey;
} else {
bytes = ofp->maxLocal - sizeof(SPgno);
}
// cpy key
memcpy(pBuf, ((SCell *)pKey) + kLen - nLeftKey, bytes);
if (lastKeyPage) {
if (lastKeyPageSpace >= vLen) {
memcpy(pBuf + kLen -nLeftKey, pVal, vLen);
nLeft -= vLen;
pgno = 0;
} else {
memcpy(pBuf + kLen -nLeftKey, pVal, lastKeyPageSpace);
nLeft -= lastKeyPageSpace;
// fetch next ofp, a new ofp and make it dirty
ret = tdbFetchOvflPage(pPage->pPager, &pgno, &nextOfp, pTxn, pBt);
if (ret < 0) {
return -1;
}
}
} else {
// fetch next ofp, a new ofp and make it dirty
ret = tdbFetchOvflPage(pPage->pPager, &pgno, &nextOfp, pTxn, pBt);
if (ret < 0) {
return -1;
}
}
memcpy(pBuf + kLen - nLeft, &pgno, sizeof(pgno));
ret = tdbPageInsertCell(ofp, 0, pBuf, bytes + sizeof(pgno), 0);
if (ret < 0) {
return -1;
}
ofp = nextOfp;
nLeftKey -= bytes;
nLeft -= bytes;
} while (nLeftKey > 0);
while (nLeft > 0) {
// pack left val data to ovpages
lastPage = 0;
if (nLeft <= maxLocal - sizeof(SPgno)) {
bytes = nLeft;
lastPage = 1;
} else {
bytes = maxLocal - sizeof(SPgno);
}
// fetch next ofp if not last page
if (!lastPage) {
// fetch a new ofp and make it dirty
ret = tdbFetchOvflPage(pPage->pPager, &pgno, &nextOfp, pTxn, pBt);
if (ret < 0) {
tdbFree(pBuf);
return -1;
}
} else {
pgno = 0;
}
memcpy(pBuf, ((SCell *)pVal) + vLen - nLeft, bytes);
memcpy(pBuf + bytes, &pgno, sizeof(pgno));
ret = tdbPageInsertCell(ofp, 0, pBuf, bytes + sizeof(pgno), 0);
if (ret < 0) {
tdbFree(pBuf);
return -1;
}
ofp = nextOfp;
nLeft -= bytes;
}
}
// free local buffer
tdbFree(pBuf);
*szPayload = nLocal;
// ASSERT(0);
}
return 0;
}
static int tdbBtreeEncodeCell(SPage *pPage, const void *pKey, int kLen, const void *pVal, int vLen, SCell *pCell,
int *szCell) {
int *szCell, TXN *pTxn, SBTree *pBt) {
u8 leaf;
int nHeader;
int nPayload;
@ -911,7 +1136,7 @@ static int tdbBtreeEncodeCell(SPage *pPage, const void *pKey, int kLen, const vo
vLen = 0;
}
ret = tdbBtreeEncodePayload(pPage, pCell, nHeader, pKey, kLen, pVal, vLen, &nPayload);
ret = tdbBtreeEncodePayload(pPage, pCell, nHeader, pKey, kLen, pVal, vLen, &nPayload, pTxn, pBt);
if (ret < 0) {
// TODO
ASSERT(0);
@ -922,8 +1147,13 @@ static int tdbBtreeEncodeCell(SPage *pPage, const void *pKey, int kLen, const vo
return 0;
}
static int tdbBtreeDecodePayload(SPage *pPage, const SCell *pCell, int nHeader, SCellDecoder *pDecoder) {
static int tdbBtreeDecodePayload(SPage *pPage, const SCell *pCell, int nHeader, SCellDecoder *pDecoder, TXN *pTxn, SBTree *pBt) {
int ret = 0;
int nPayload;
int maxLocal = pPage->maxLocal;
int kLen = pDecoder->kLen;
int vLen = pDecoder->vLen;
if (pDecoder->pVal) {
ASSERT(!TDB_BTREE_PAGE_IS_LEAF(pPage));
@ -932,24 +1162,171 @@ static int tdbBtreeDecodePayload(SPage *pPage, const SCell *pCell, int nHeader,
nPayload = pDecoder->kLen + pDecoder->vLen;
}
if (nHeader + nPayload <= pPage->maxLocal) {
if (nHeader + nPayload <= maxLocal) {
// no over flow case
pDecoder->pKey = pCell + nHeader;
pDecoder->pKey = (SCell *)pCell + nHeader;
if (pDecoder->pVal == NULL && pDecoder->vLen > 0) {
pDecoder->pVal = pCell + nHeader + pDecoder->kLen;
pDecoder->pVal = (SCell *)pCell + nHeader + pDecoder->kLen;
}
return 0;
} else {
// handle overflow case
// calc local storage size
int minLocal = pPage->minLocal;
int surplus = minLocal + (nPayload + nHeader - minLocal) % (maxLocal - sizeof(SPgno));
int nLocal = surplus <= maxLocal ? surplus : minLocal;
int nLeft = nPayload;
SPgno pgno = 0;
SPage *ofp;
SCell *ofpCell;
int bytes;
int lastPage = 0;
if (nLocal >= pDecoder->kLen + 4) {
pDecoder->pKey = (SCell *)pCell + nHeader;
nLeft -= kLen;
if (nLocal > kLen + 4) {
// read partial val to local
pDecoder->pVal = tdbRealloc(pDecoder->pVal, vLen);
if (pDecoder->pVal == NULL) {
return -1;
}
TDB_CELLDECODER_SET_FREE_VAL(pDecoder);
memcpy(pDecoder->pVal, pCell + nHeader + kLen, nLocal - kLen - sizeof(SPgno));
nLeft -= nLocal - kLen - sizeof(SPgno);
}
{
// TODO: handle overflow case
ASSERT(0);
memcpy(&pgno, pCell + nHeader + nPayload - nLeft, sizeof(pgno));
// unpack left val data from ovpages
while (pgno != 0) {
ret = tdbLoadOvflPage(pPage->pPager, &pgno, &ofp, pTxn, pBt);
if (ret < 0) {
return -1;
}
ofpCell = tdbPageGetCell(ofp, 0);
if (nLeft <= ofp->maxLocal - sizeof(SPgno)) {
bytes = nLeft;
lastPage = 1;
} else {
bytes = ofp->maxLocal - sizeof(SPgno);
}
memcpy(pDecoder->pVal + vLen - nLeft, ofpCell, bytes);
nLeft -= bytes;
memcpy(&pgno, ofpCell + bytes, sizeof(pgno));
}
} else {
int nLeftKey = kLen;
// load partial key and nextPgno
pDecoder->pKey = tdbRealloc(pDecoder->pKey, kLen);
if (pDecoder->pKey == NULL) {
return -1;
}
TDB_CELLDECODER_SET_FREE_KEY(pDecoder);
memcpy(pDecoder->pKey, pCell + nHeader, nLocal - 4);
nLeft -= nLocal - 4;
nLeftKey -= nLocal -4;
memcpy(&pgno, pCell + nHeader + nLocal - 4, sizeof(pgno));
int lastKeyPageSpace = 0;
// load left key & val to ovpages
while (pgno != 0) {
ret = tdbLoadOvflPage(pPage->pPager, &pgno, &ofp, pTxn, pBt);
if (ret < 0) {
return -1;
}
ofpCell = tdbPageGetCell(ofp, 0);
int lastKeyPage = 0;
if (nLeftKey <= maxLocal - sizeof(SPgno)) {
bytes = nLeftKey;
lastKeyPage = 1;
lastKeyPageSpace = ofp->maxLocal - sizeof(SPgno) - nLeftKey;
} else {
bytes = ofp->maxLocal - sizeof(SPgno);
}
// cpy key
memcpy(pDecoder->pKey + kLen - nLeftKey, ofpCell, bytes);
if (lastKeyPage) {
if (lastKeyPageSpace >= vLen) {
pDecoder->pVal = ofpCell + kLen -nLeftKey;
nLeft -= vLen;
pgno = 0;
} else {
// read partial val to local
pDecoder->pVal = tdbRealloc(pDecoder->pVal, vLen);
if (pDecoder->pVal == NULL) {
return -1;
}
TDB_CELLDECODER_SET_FREE_VAL(pDecoder);
memcpy(pDecoder->pVal, ofpCell + kLen -nLeftKey, lastKeyPageSpace);
nLeft -= lastKeyPageSpace;
}
}
memcpy(&pgno, ofpCell + bytes, sizeof(pgno));
nLeftKey -= bytes;
nLeft -= bytes;
}
while (nLeft > 0) {
ret = tdbLoadOvflPage(pPage->pPager, &pgno, &ofp, pTxn, pBt);
if (ret < 0) {
return -1;
}
ofpCell = tdbPageGetCell(ofp, 0);
// load left val data to ovpages
lastPage = 0;
if (nLeft <= ofp->maxLocal - sizeof(SPgno)) {
bytes = nLeft;
lastPage = 1;
} else {
bytes = ofp->maxLocal - sizeof(SPgno);
}
if (lastPage) {
pgno = 0;
}
if (!pDecoder->pVal) {
pDecoder->pVal = tdbRealloc(pDecoder->pVal, vLen);
if (pDecoder->pVal == NULL) {
return -1;
}
TDB_CELLDECODER_SET_FREE_VAL(pDecoder);
}
memcpy(pDecoder->pVal, ofpCell + vLen - nLeft, bytes);
nLeft -= bytes;
memcpy(&pgno, ofpCell + vLen - nLeft + bytes, sizeof(pgno));
nLeft -= bytes;
}
}
}
return 0;
}
static int tdbBtreeDecodeCell(SPage *pPage, const SCell *pCell, SCellDecoder *pDecoder) {
static int tdbBtreeDecodeCell(SPage *pPage, const SCell *pCell, SCellDecoder *pDecoder, TXN *pTxn, SBTree *pBt) {
u8 leaf;
int nHeader;
int ret;
@ -963,6 +1340,7 @@ static int tdbBtreeDecodeCell(SPage *pPage, const SCell *pCell, SCellDecoder *pD
pDecoder->vLen = -1;
pDecoder->pVal = NULL;
pDecoder->pgno = 0;
TDB_CELLDECODER_SET_FREE_NIL(pDecoder);
// 1. Decode header part
if (!leaf) {
@ -987,7 +1365,7 @@ static int tdbBtreeDecodeCell(SPage *pPage, const SCell *pCell, SCellDecoder *pD
}
// 2. Decode payload part
ret = tdbBtreeDecodePayload(pPage, pCell, nHeader, pDecoder);
ret = tdbBtreeDecodePayload(pPage, pCell, nHeader, pDecoder, pTxn, pBt);
if (ret < 0) {
return -1;
}
@ -995,41 +1373,71 @@ static int tdbBtreeDecodeCell(SPage *pPage, const SCell *pCell, SCellDecoder *pD
return 0;
}
static int tdbBtreeCellSize(const SPage *pPage, SCell *pCell) {
static int tdbBtreeCellSize(const SPage *pPage, SCell *pCell, int dropOfp, TXN *pTxn, SBTree *pBt) {
u8 leaf;
int szCell;
int kLen = 0, vLen = 0;
int kLen = 0, vLen = 0, nHeader = 0;
leaf = TDB_BTREE_PAGE_IS_LEAF(pPage);
szCell = 0;
if (!leaf) {
szCell += sizeof(SPgno);
nHeader += sizeof(SPgno);
}
if (pPage->kLen == TDB_VARIANT_LEN) {
szCell += tdbGetVarInt(pCell + szCell, &kLen);
nHeader += tdbGetVarInt(pCell + nHeader, &kLen);
} else {
kLen = pPage->kLen;
}
if (pPage->vLen == TDB_VARIANT_LEN) {
ASSERT(leaf);
szCell += tdbGetVarInt(pCell + szCell, &vLen);
nHeader += tdbGetVarInt(pCell + nHeader, &vLen);
} else if (leaf) {
vLen = pPage->vLen;
}
szCell = szCell + kLen + vLen;
int nPayload = kLen + vLen;
if (nHeader + nPayload <= pPage->maxLocal) {
return nHeader + kLen + vLen;
} else {
int maxLocal = pPage->maxLocal;
if (szCell <= pPage->maxLocal) {
return szCell;
// calc local storage size
int minLocal = pPage->minLocal;
int surplus = minLocal + (nPayload + nHeader - minLocal) % (maxLocal - sizeof(SPgno));
int nLocal = surplus <= maxLocal ? surplus : minLocal;
// free ofp pages' cells
if (dropOfp) {
int ret = 0;
SPgno pgno = *(SPgno *) (pCell + nHeader + nLocal - sizeof(SPgno));
int nLeft = nPayload - nLocal + sizeof(SPgno);
SPage *ofp;
int bytes;
while (pgno != 0) {
ret = tdbLoadOvflPage(pPage->pPager, &pgno, &ofp, pTxn, pBt);
if (ret < 0) {
return -1;
}
{
// TODO
ASSERT(0);
return 0;
SCell *ofpCell = tdbPageGetCell(ofp, 0);
if (nLeft <= ofp->maxLocal - sizeof(SPgno)) {
bytes = nLeft;
} else {
bytes = ofp->maxLocal - sizeof(SPgno);
}
memcpy(&pgno, ofpCell + bytes, sizeof(pgno));
tdbPagerReturnPage(pPage->pPager, ofp, pTxn);
nLeft -= bytes;
}
}
return nHeader + nLocal;
}
}
// TDB_BTREE_CELL
@ -1212,7 +1620,7 @@ int tdbBtreeNext(SBTC *pBtc, void **ppKey, int *kLen, void **ppVal, int *vLen) {
pCell = tdbPageGetCell(pBtc->pPage, pBtc->idx);
tdbBtreeDecodeCell(pBtc->pPage, pCell, &cd);
tdbBtreeDecodeCell(pBtc->pPage, pCell, &cd, pBtc->pTxn, pBtc->pBt);
pKey = tdbRealloc(*ppKey, cd.kLen);
if (pKey == NULL) {
@ -1258,7 +1666,7 @@ int tdbBtreePrev(SBTC *pBtc, void **ppKey, int *kLen, void **ppVal, int *vLen) {
pCell = tdbPageGetCell(pBtc->pPage, pBtc->idx);
tdbBtreeDecodeCell(pBtc->pPage, pCell, &cd);
tdbBtreeDecodeCell(pBtc->pPage, pCell, &cd, pBtc->pTxn, pBtc->pBt);
pKey = tdbRealloc(*ppKey, cd.kLen);
if (pKey == NULL) {
@ -1427,7 +1835,7 @@ int tdbBtcGet(SBTC *pBtc, const void **ppKey, int *kLen, const void **ppVal, int
}
pCell = tdbPageGetCell(pBtc->pPage, pBtc->idx);
tdbBtreeDecodeCell(pBtc->pPage, pCell, &pBtc->coder);
tdbBtreeDecodeCell(pBtc->pPage, pCell, &pBtc->coder, pBtc->pTxn, pBtc->pBt);
if (ppKey) {
*ppKey = (void *)pBtc->coder.pKey;
@ -1464,7 +1872,7 @@ int tdbBtcDelete(SBTC *pBtc) {
return -1;
}
tdbPageDropCell(pBtc->pPage, idx);
tdbPageDropCell(pBtc->pPage, idx, pBtc->pTxn, pBtc->pBt);
// update interior page or do balance
if (idx == nCells - 1) {
@ -1488,9 +1896,9 @@ int tdbBtcDelete(SBTC *pBtc) {
// update the cell with new key
pCell = tdbOsMalloc(nKey + 9);
tdbBtreeEncodeCell(pPage, pKey, nKey, &pgno, sizeof(pgno), pCell, &szCell);
tdbBtreeEncodeCell(pPage, pKey, nKey, &pgno, sizeof(pgno), pCell, &szCell, pBtc->pTxn, pBtc->pBt);
ret = tdbPageUpdateCell(pPage, idx, pCell, szCell);
ret = tdbPageUpdateCell(pPage, idx, pCell, szCell, pBtc->pTxn, pBtc->pBt);
if (ret < 0) {
tdbOsFree(pCell);
ASSERT(0);
@ -1538,7 +1946,7 @@ int tdbBtcUpsert(SBTC *pBtc, const void *pKey, int kLen, const void *pData, int
pCell = (SCell *)pBtc->pBt->pBuf;
// encode cell
ret = tdbBtreeEncodeCell(pBtc->pPage, pKey, kLen, pData, nData, pCell, &szCell);
ret = tdbBtreeEncodeCell(pBtc->pPage, pKey, kLen, pData, nData, pCell, &szCell, pBtc->pTxn, pBtc->pBt);
if (ret < 0) {
ASSERT(0);
return -1;
@ -1559,7 +1967,7 @@ int tdbBtcUpsert(SBTC *pBtc, const void *pKey, int kLen, const void *pData, int
} else {
ASSERT(pBtc->idx < nCells);
ret = tdbPageUpdateCell(pBtc->pPage, pBtc->idx, pCell, szCell);
ret = tdbPageUpdateCell(pBtc->pPage, pBtc->idx, pCell, szCell, pBtc->pTxn, pBtc->pBt);
}
if (ret < 0) {
ASSERT(0);
@ -1620,7 +2028,7 @@ int tdbBtcMoveTo(SBTC *pBtc, const void *pKey, int kLen, int *pCRst) {
// check if key <= current position
if (idx < nCells) {
pCell = tdbPageGetCell(pPage, idx);
tdbBtreeDecodeCell(pPage, pCell, &cd);
tdbBtreeDecodeCell(pPage, pCell, &cd, pBtc->pTxn, pBtc->pBt);
c = pBt->kcmpr(pKey, kLen, cd.pKey, cd.kLen);
if (c > 0) break;
}
@ -1629,7 +2037,7 @@ int tdbBtcMoveTo(SBTC *pBtc, const void *pKey, int kLen, int *pCRst) {
if (idx > 0) {
pCell = tdbPageGetCell(pPage, idx - 1);
tdbBtreeDecodeCell(pPage, pCell, &cd);
c = pBt->kcmpr(pKey, kLen, cd.pKey, cd.kLen);
c = pBt->kcmpr(pKey, kLen, cd.pKey, cd.kLen, pBtc->pTxn, pBtc->pBt);
if (c <= 0) break;
}
}

16
source/libs/tdb/src/db/tdbPage.c
View File

@ -82,7 +82,8 @@ int tdbPageDestroy(SPage *pPage, void (*xFree)(void *arg, void *ptr), void *arg)
return 0;
}
void tdbPageZero(SPage *pPage, u8 szAmHdr, int (*xCellSize)(const SPage *, SCell *)) {
void tdbPageZero(SPage *pPage, u8 szAmHdr, int (*xCellSize)(const SPage *, SCell *, int,
TXN *, SBTree *pBt)) {
pPage->pPageHdr = pPage->pData + szAmHdr;
TDB_PAGE_NCELLS_SET(pPage, 0);
TDB_PAGE_CCELLS_SET(pPage, pPage->pageSize - sizeof(SPageFtr));
@ -98,7 +99,8 @@ void tdbPageZero(SPage *pPage, u8 szAmHdr, int (*xCellSize)(const SPage *, SCell
ASSERT((u8 *)pPage->pPageFtr == pPage->pFreeEnd);
}
void tdbPageInit(SPage *pPage, u8 szAmHdr, int (*xCellSize)(const SPage *, SCell *)) {
void tdbPageInit(SPage *pPage, u8 szAmHdr, int (*xCellSize)(const SPage *, SCell *, int,
TXN *, SBTree *pBt)) {
pPage->pPageHdr = pPage->pData + szAmHdr;
pPage->pCellIdx = pPage->pPageHdr + TDB_PAGE_HDR_SIZE(pPage);
pPage->pFreeStart = pPage->pCellIdx + TDB_PAGE_OFFSET_SIZE(pPage) * TDB_PAGE_NCELLS(pPage);
@ -171,12 +173,12 @@ int tdbPageInsertCell(SPage *pPage, int idx, SCell *pCell, int szCell, u8 asOvfl
return 0;
}
int tdbPageUpdateCell(SPage *pPage, int idx, SCell *pCell, int szCell) {
tdbPageDropCell(pPage, idx);
int tdbPageUpdateCell(SPage *pPage, int idx, SCell *pCell, int szCell, TXN *pTxn, SBTree *pBt) {
tdbPageDropCell(pPage, idx, pTxn, pBt);
return tdbPageInsertCell(pPage, idx, pCell, szCell, 0);
}
int tdbPageDropCell(SPage *pPage, int idx) {
int tdbPageDropCell(SPage *pPage, int idx, TXN *pTxn, SBTree *pBt) {
int lidx;
SCell *pCell;
int szCell;
@ -205,7 +207,7 @@ int tdbPageDropCell(SPage *pPage, int idx) {
lidx = idx - iOvfl;
pCell = TDB_PAGE_CELL_AT(pPage, lidx);
szCell = (*pPage->xCellSize)(pPage, pCell);
szCell = (*pPage->xCellSize)(pPage, pCell, 1, pTxn, pBt);
tdbPageFree(pPage, lidx, pCell, szCell);
TDB_PAGE_NCELLS_SET(pPage, nCells - 1);
@ -420,7 +422,7 @@ static int tdbPageDefragment(SPage *pPage) {
ASSERT(pCell != NULL);
szCell = (*pPage->xCellSize)(pPage, pCell);
szCell = (*pPage->xCellSize)(pPage, pCell, 0, NULL, NULL);
ASSERT(pCell + szCell <= pNextCell);
if (pCell + szCell < pNextCell) {

30
source/libs/tdb/src/inc/tdbInt.h
View File

@ -116,13 +116,25 @@ typedef struct SBtInfo {
int nData;
} SBtInfo;
#define TDB_CELLD_F_NIL 0x0
#define TDB_CELLD_F_KEY 0x1
#define TDB_CELLD_F_VAL 0x2
#define TDB_CELLDECODER_SET_FREE_NIL(pCellDecoder) ((pCellDecoder)->freeKV = TDB_CELLD_F_NIL)
#define TDB_CELLDECODER_SET_FREE_KEY(pCellDecoder) ((pCellDecoder)->freeKV |= TDB_CELLD_F_KEY)
#define TDB_CELLDECODER_SET_FREE_VAL(pCellDecoder) ((pCellDecoder)->freeKV |= TDB_CELLD_F_VAL)
#define TDB_CELLDECODER_FREE_KEY(pCellDecoder) ((pCellDecoder)->freeKV & TDB_CELLD_F_KEY)
#define TDB_CELLDECODER_FREE_VAL(pCellDecoder) ((pCellDecoder)->freeKV & TDB_CELLD_F_VAL)
typedef struct {
int kLen;
const u8 *pKey;
u8 *pKey;
int vLen;
const u8 *pVal;
u8 *pVal;
SPgno pgno;
u8 *pBuf;
u8 freeKV;
} SCellDecoder;
struct SBTC {
@ -251,7 +263,7 @@ struct SPage {
int vLen; // value length of the page, -1 for unknown
int maxLocal;
int minLocal;
int (*xCellSize)(const SPage *, SCell *);
int (*xCellSize)(const SPage *, SCell *, int, TXN *pTxn, SBTree *pBt);
// Fields used by SPCache
TDB_PCACHE_PAGE
};
@ -298,16 +310,18 @@ static inline int tdbTryLockPage(tdb_spinlock_t *pLock) {
#define TDB_PAGE_USABLE_SIZE(pPage) ((u8 *)(pPage)->pPageFtr - (pPage)->pCellIdx)
#define TDB_PAGE_FREE_SIZE(pPage) (*(pPage)->pPageMethods->getFreeBytes)(pPage)
#define TDB_PAGE_PGNO(pPage) ((pPage)->pgid.pgno)
#define TDB_BYTES_CELL_TAKEN(pPage, pCell) ((*(pPage)->xCellSize)(pPage, pCell) + (pPage)->pPageMethods->szOffset)
#define TDB_BYTES_CELL_TAKEN(pPage, pCell) ((*(pPage)->xCellSize)(pPage, pCell, 0, NULL, NULL) + (pPage)->pPageMethods->szOffset)
#define TDB_PAGE_OFFSET_SIZE(pPage) ((pPage)->pPageMethods->szOffset)
int tdbPageCreate(int pageSize, SPage **ppPage, void *(*xMalloc)(void *, size_t), void *arg);
int tdbPageDestroy(SPage *pPage, void (*xFree)(void *arg, void *ptr), void *arg);
void tdbPageZero(SPage *pPage, u8 szAmHdr, int (*xCellSize)(const SPage *, SCell *));
void tdbPageInit(SPage *pPage, u8 szAmHdr, int (*xCellSize)(const SPage *, SCell *));
void tdbPageZero(SPage *pPage, u8 szAmHdr, int (*xCellSize)(const SPage *, SCell *, int,
TXN *, SBTree *pBt));
void tdbPageInit(SPage *pPage, u8 szAmHdr, int (*xCellSize)(const SPage *, SCell *, int,
TXN *, SBTree *pBt));
int tdbPageInsertCell(SPage *pPage, int idx, SCell *pCell, int szCell, u8 asOvfl);
int tdbPageDropCell(SPage *pPage, int idx);
int tdbPageUpdateCell(SPage *pPage, int idx, SCell *pCell, int szCell);
int tdbPageDropCell(SPage *pPage, int idx, TXN *pTxn, SBTree *pBt);
int tdbPageUpdateCell(SPage *pPage, int idx, SCell *pCell, int szCell, TXN *pTxn, SBTree *pBt);
void tdbPageCopy(SPage *pFromPage, SPage *pToPage);
int tdbPageCapacity(int pageSize, int amHdrSize);

5
source/libs/tdb/test/CMakeLists.txt
View File

@ -5,3 +5,8 @@ target_link_libraries(tdbTest tdb gtest gtest_main)
# tdbUtilTest
add_executable(tdbUtilTest "tdbUtilTest.cpp")
target_link_libraries(tdbUtilTest tdb gtest gtest_main)
# tdbUtilTest
add_executable(tdbExOVFLTest "tdbExOVFLTest.cpp")
target_link_libraries(tdbExOVFLTest tdb gtest gtest_main)

469
source/libs/tdb/test/tdbExOVFLTest.cpp Normal file
View File

@ -0,0 +1,469 @@
#include <gtest/gtest.h>
#define ALLOW_FORBID_FUNC
#include "os.h"
#include "tdb.h"
#include <shared_mutex>
#include <string>
#include <thread>
#include <vector>
typedef struct SPoolMem {
int64_t size;
struct SPoolMem *prev;
struct SPoolMem *next;
} SPoolMem;
static SPoolMem *openPool() {
SPoolMem *pPool = (SPoolMem *)taosMemoryMalloc(sizeof(*pPool));
pPool->prev = pPool->next = pPool;
pPool->size = 0;
return pPool;
}
static void clearPool(SPoolMem *pPool) {
SPoolMem *pMem;
do {
pMem = pPool->next;
if (pMem == pPool) break;
pMem->next->prev = pMem->prev;
pMem->prev->next = pMem->next;
pPool->size -= pMem->size;
taosMemoryFree(pMem);
} while (1);
assert(pPool->size == 0);
}
static void closePool(SPoolMem *pPool) {
clearPool(pPool);
taosMemoryFree(pPool);
}
static void *poolMalloc(void *arg, size_t size) {
void *ptr = NULL;
SPoolMem *pPool = (SPoolMem *)arg;
SPoolMem *pMem;
pMem = (SPoolMem *)taosMemoryMalloc(sizeof(*pMem) + size);
if (pMem == NULL) {
assert(0);
}
pMem->size = sizeof(*pMem) + size;
pMem->next = pPool->next;
pMem->prev = pPool;
pPool->next->prev = pMem;
pPool->next = pMem;
pPool->size += pMem->size;
ptr = (void *)(&pMem[1]);
return ptr;
}
static void poolFree(void *arg, void *ptr) {
SPoolMem *pPool = (SPoolMem *)arg;
SPoolMem *pMem;
pMem = &(((SPoolMem *)ptr)[-1]);
pMem->next->prev = pMem->prev;
pMem->prev->next = pMem->next;
pPool->size -= pMem->size;
taosMemoryFree(pMem);
}
static int tKeyCmpr(const void *pKey1, int kLen1, const void *pKey2, int kLen2) {
int k1, k2;
std::string s1((char *)pKey1 + 3, kLen1 - 3);
std::string s2((char *)pKey2 + 3, kLen2 - 3);
k1 = stoi(s1);
k2 = stoi(s2);
if (k1 < k2) {
return -1;
} else if (k1 > k2) {
return 1;
} else {
return 0;
}
}
static int tDefaultKeyCmpr(const void *pKey1, int keyLen1, const void *pKey2, int keyLen2) {
int mlen;
int cret;
ASSERT(keyLen1 > 0 && keyLen2 > 0 && pKey1 != NULL && pKey2 != NULL);
mlen = keyLen1 < keyLen2 ? keyLen1 : keyLen2;
cret = memcmp(pKey1, pKey2, mlen);
if (cret == 0) {
if (keyLen1 < keyLen2) {
cret = -1;
} else if (keyLen1 > keyLen2) {
cret = 1;
} else {
cret = 0;
}
}
return cret;
}
TEST(TdbOVFLPagesTest, TbUpsertTest) {
}
TEST(TdbOVFLPagesTest, TbPGetTest) {
}
static void generateBigVal(char *val, int valLen) {
for (int i = 0; i < valLen; ++i) {
char c = char(i & 0xff);
if (c == 0) {
c = 1;
}
val[i] = c;
}
}
static TDB *openEnv(char const *envName, int const pageSize, int const pageNum) {
TDB *pEnv = NULL;
int ret = tdbOpen(envName, pageSize, pageNum, &pEnv);
if (ret) {
pEnv = NULL;
}
return pEnv;
}
static void insertOfp(void) {
int ret = 0;
taosRemoveDir("tdb");
// open Env
int const pageSize = 4096;
int const pageNum = 64;
TDB *pEnv = openEnv("tdb", pageSize, pageNum);
GTEST_ASSERT_NE(pEnv, nullptr);
// open db
TTB *pDb = NULL;
tdb_cmpr_fn_t compFunc = tKeyCmpr;
ret = tdbTbOpen("ofp_insert.db", -1, -1, compFunc, pEnv, &pDb);
GTEST_ASSERT_EQ(ret, 0);
// open the pool
SPoolMem *pPool = openPool();
// start a transaction
TXN txn;
int64_t txnid = 0;
++txnid;
tdbTxnOpen(&txn, txnid, poolMalloc, poolFree, pPool, TDB_TXN_WRITE | TDB_TXN_READ_UNCOMMITTED);
tdbBegin(pEnv, &txn);
// generate value payload
char val[((4083 - 4 - 3 - 2)+1)*100]; // pSize(4096) - amSize(1) - pageHdr(8) - footerSize(4)
int valLen = sizeof(val) / sizeof(val[0]);
generateBigVal(val, valLen);
// insert the generated big data
ret = tdbTbInsert(pDb, "key1", strlen("key1"), val, valLen, &txn);
GTEST_ASSERT_EQ(ret, 0);
// commit current transaction
tdbCommit(pEnv, &txn);
tdbTxnClose(&txn);
}
//TEST(TdbOVFLPagesTest, DISABLED_TbInsertTest) {
TEST(TdbOVFLPagesTest, TbInsertTest) {
insertOfp();
}
//TEST(TdbOVFLPagesTest, DISABLED_TbGetTest) {
TEST(TdbOVFLPagesTest, TbGetTest) {
insertOfp();
// open Env
int const pageSize = 4096;
int const pageNum = 64;
TDB *pEnv = openEnv("tdb", pageSize, pageNum);
GTEST_ASSERT_NE(pEnv, nullptr);
// open db
TTB *pDb = NULL;
tdb_cmpr_fn_t compFunc = tKeyCmpr;
int ret = tdbTbOpen("ofp_insert.db", -1, -1, compFunc, pEnv, &pDb);
GTEST_ASSERT_EQ(ret, 0);
// generate value payload
char val[((4083 - 4 - 3 - 2)+1)*100]; // pSize(4096) - amSize(1) - pageHdr(8) - footerSize(4)
int valLen = sizeof(val) / sizeof(val[0]);
generateBigVal(val, valLen);
{ // Query the data
void *pVal = NULL;
int vLen;
ret = tdbTbGet(pDb, "key1", strlen("key1"), &pVal, &vLen);
ASSERT(ret == 0);
GTEST_ASSERT_EQ(ret, 0);
GTEST_ASSERT_EQ(vLen, valLen);
GTEST_ASSERT_EQ(memcmp(val, pVal, vLen), 0);
tdbFree(pVal);
}
}
TEST(TdbOVFLPagesTest, TbDeleteTest) {
int ret = 0;
taosRemoveDir("tdb");
// open Env
int const pageSize = 4096;
int const pageNum = 64;
TDB *pEnv = openEnv("tdb", pageSize, pageNum);
GTEST_ASSERT_NE(pEnv, nullptr);
// open db
TTB *pDb = NULL;
tdb_cmpr_fn_t compFunc = tKeyCmpr;
ret = tdbTbOpen("ofp_insert.db", -1, -1, compFunc, pEnv, &pDb);
GTEST_ASSERT_EQ(ret, 0);
// open the pool
SPoolMem *pPool = openPool();
// start a transaction
TXN txn;
int64_t txnid = 0;
++txnid;
tdbTxnOpen(&txn, txnid, poolMalloc, poolFree, pPool, TDB_TXN_WRITE | TDB_TXN_READ_UNCOMMITTED);
tdbBegin(pEnv, &txn);
// generate value payload
char val[((4083 - 4 - 3 - 2)+1)*100]; // pSize(4096) - amSize(1) - pageHdr(8) - footerSize(4)
int valLen = sizeof(val) / sizeof(val[0]);
generateBigVal(val, valLen);
{ // insert the generated big data
ret = tdbTbInsert(pDb, "key1", strlen("key1"), val, valLen, &txn);
GTEST_ASSERT_EQ(ret, 0);
}
{ // query the data
void *pVal = NULL;
int vLen;
ret = tdbTbGet(pDb, "key1", strlen("key1"), &pVal, &vLen);
ASSERT(ret == 0);
GTEST_ASSERT_EQ(ret, 0);
GTEST_ASSERT_EQ(vLen, valLen);
GTEST_ASSERT_EQ(memcmp(val, pVal, vLen), 0);
tdbFree(pVal);
}
/* open to debug committed file
tdbCommit(pEnv, &txn);
tdbTxnClose(&txn);
++txnid;
tdbTxnOpen(&txn, txnid, poolMalloc, poolFree, pPool, TDB_TXN_WRITE | TDB_TXN_READ_UNCOMMITTED);
tdbBegin(pEnv, &txn);
*/
{ // upsert the data
ret = tdbTbUpsert(pDb, "key1", strlen("key1"), "value1", strlen("value1"), &txn);
GTEST_ASSERT_EQ(ret, 0);
}
{ // query the upserted data
void *pVal = NULL;
int vLen;
ret = tdbTbGet(pDb, "key1", strlen("key1"), &pVal, &vLen);
ASSERT(ret == 0);
GTEST_ASSERT_EQ(ret, 0);
GTEST_ASSERT_EQ(vLen, strlen("value1"));
GTEST_ASSERT_EQ(memcmp("value1", pVal, vLen), 0);
tdbFree(pVal);
}
{ // delete the data
ret = tdbTbDelete(pDb, "key1", strlen("key1"), &txn);
GTEST_ASSERT_EQ(ret, 0);
}
{ // query the deleted data
void *pVal = NULL;
int vLen = -1;
ret = tdbTbGet(pDb, "key1", strlen("key1"), &pVal, &vLen);
ASSERT(ret == -1);
GTEST_ASSERT_EQ(ret, -1);
GTEST_ASSERT_EQ(vLen, -1);
GTEST_ASSERT_EQ(pVal, nullptr);
tdbFree(pVal);
}
// commit current transaction
tdbCommit(pEnv, &txn);
tdbTxnClose(&txn);
}
TEST(tdb_test, DISABLED_simple_insert1) {
//TEST(tdb_test, simple_insert1) {
int ret;
TDB *pEnv;
TTB *pDb;
tdb_cmpr_fn_t compFunc;
int nData = 1;
TXN txn;
int const pageSize = 4096;
taosRemoveDir("tdb");
// Open Env
ret = tdbOpen("tdb", pageSize, 64, &pEnv);
GTEST_ASSERT_EQ(ret, 0);
// Create a database
compFunc = tKeyCmpr;
ret = tdbTbOpen("db.db", -1, -1, compFunc, pEnv, &pDb);
GTEST_ASSERT_EQ(ret, 0);
{
char key[64];
//char val[(4083 - 4 - 3 - 2)]; // pSize(4096) - amSize(1) - pageHdr(8) - footerSize(4)
char val[(4083 - 4 - 3 - 2)+1]; // pSize(4096) - amSize(1) - pageHdr(8) - footerSize(4)
int64_t poolLimit = 4096; // 1M pool limit
int64_t txnid = 0;
SPoolMem *pPool;
// open the pool
pPool = openPool();
// start a transaction
txnid++;
tdbTxnOpen(&txn, txnid, poolMalloc, poolFree, pPool, TDB_TXN_WRITE | TDB_TXN_READ_UNCOMMITTED);
tdbBegin(pEnv, &txn);
for (int iData = 1; iData <= nData; iData++) {
sprintf(key, "key0");
sprintf(val, "value%d", iData);
//ret = tdbTbInsert(pDb, key, strlen(key), val, strlen(val), &txn);
//GTEST_ASSERT_EQ(ret, 0);
// generate value payload
int valLen = sizeof(val) / sizeof(val[0]);
for (int i = 6; i < valLen; ++i) {
char c = char(i & 0xff);
if (c == 0) {
c = 1;
}
val[i] = c;
}
ret = tdbTbInsert(pDb, "key1", strlen("key1"), val, valLen, &txn);
GTEST_ASSERT_EQ(ret, 0);
// if pool is full, commit the transaction and start a new one
if (pPool->size >= poolLimit) {
// commit current transaction
tdbCommit(pEnv, &txn);
tdbTxnClose(&txn);
// start a new transaction
clearPool(pPool);
txnid++;
tdbTxnOpen(&txn, txnid, poolMalloc, poolFree, pPool, TDB_TXN_WRITE | TDB_TXN_READ_UNCOMMITTED);
tdbBegin(pEnv, &txn);
}
}
// commit the transaction
tdbCommit(pEnv, &txn);
tdbTxnClose(&txn);
{ // Query the data
void *pVal = NULL;
int vLen;
for (int i = 1; i <= nData; i++) {
sprintf(key, "key%d", i);
sprintf(val, "value%d", i);
ret = tdbTbGet(pDb, key, strlen(key), &pVal, &vLen);
ASSERT(ret == 0);
GTEST_ASSERT_EQ(ret, 0);
GTEST_ASSERT_EQ(vLen, strlen(val));
GTEST_ASSERT_EQ(memcmp(val, pVal, vLen), 0);
}
tdbFree(pVal);
}
{ // Iterate to query the DB data
TBC *pDBC;
void *pKey = NULL;
void *pVal = NULL;
int vLen, kLen;
int count = 0;
ret = tdbTbcOpen(pDb, &pDBC, NULL);
GTEST_ASSERT_EQ(ret, 0);
tdbTbcMoveToFirst(pDBC);
for (;;) {
ret = tdbTbcNext(pDBC, &pKey, &kLen, &pVal, &vLen);
if (ret < 0) break;
// std::cout.write((char *)pKey, kLen) /* << " " << kLen */ << " ";
// std::cout.write((char *)pVal, vLen) /* << " " << vLen */;
// std::cout << std::endl;
count++;
}
GTEST_ASSERT_EQ(count, nData);
tdbTbcClose(pDBC);
tdbFree(pKey);
tdbFree(pVal);
}
}
ret = tdbTbDrop(pDb);
GTEST_ASSERT_EQ(ret, 0);
// Close a database
tdbTbClose(pDb);
// Close Env
ret = tdbClose(pEnv);
GTEST_ASSERT_EQ(ret, 0);
}

3
source/util/src/terror.c
View File

@ -423,6 +423,9 @@ TAOS_DEFINE_ERROR(TSDB_CODE_SYN_IS_LEADER, "Sync is leader")
TAOS_DEFINE_ERROR(TSDB_CODE_SYN_NOT_LEADER, "Sync not leader")
TAOS_DEFINE_ERROR(TSDB_CODE_SYN_ONE_REPLICA, "Sync one replica")
TAOS_DEFINE_ERROR(TSDB_CODE_SYN_NOT_IN_NEW_CONFIG, "Sync not in new config")
TAOS_DEFINE_ERROR(TSDB_CODE_SYN_NEW_CONFIG_ERROR, "Sync new config error")
TAOS_DEFINE_ERROR(TSDB_CODE_SYN_RECONFIG_NOT_READY, "Sync not ready for reconfig")
TAOS_DEFINE_ERROR(TSDB_CODE_SYN_PROPOSE_NOT_READY, "Sync not ready for propose")
TAOS_DEFINE_ERROR(TSDB_CODE_SYN_INTERNAL_ERROR, "Sync internal error")
// wal

7
tests/script/jenkins/basic.txt
View File

@ -72,14 +72,15 @@
./test.sh -f tsim/stream/basic0.sim
./test.sh -f tsim/stream/basic1.sim
./test.sh -f tsim/stream/basic2.sim
# ./test.sh -f tsim/stream/distributeInterval0.sim
./test.sh -f tsim/stream/distributeInterval0.sim
# ./test.sh -f tsim/stream/distributesession0.sim
# ./test.sh -f tsim/stream/session0.sim
# ./test.sh -f tsim/stream/session1.sim
# ./test.sh -f tsim/stream/state0.sim
# ./test.sh -f tsim/stream/triggerInterval0.sim
./test.sh -f tsim/stream/triggerInterval0.sim
# ./test.sh -f tsim/stream/triggerSession0.sim
# ./test.sh -f tsim/stream/partitionby.sim
./test.sh -f tsim/stream/partitionby.sim
./test.sh -f tsim/stream/schedSnode.sim
# ---- transaction

21
tests/script/tsim/dnode/redistribute_vgroup_replica3_v1_leader.sim
View File

@ -53,31 +53,10 @@ endi
if $data(4)[4] != ready then
goto step1
endi
#if $data(5)[4] != ready then
# goto step1
#endi
print =============== step2: create db
sql create database d1 vgroups 1 replica 3
# dnode not exist
sql_error redistribute vgroup 3 dnode 6 dnode 3 dnode 4
# vgroup not exist
sql_error redistribute vgroup 3 dnode 5 dnode 3 dnode 4
# un changed
sql_error redistribute vgroup 2 dnode 2 dnode 3 dnode 4
# no enought vnodes
sql_error redistribute vgroup 2 dnode 1 dnode 3 dnode 4
# offline vnodes
sql_error redistribute vgroup 2 dnode 5 dnode 3 dnode 4
# Invalid replica
sql_error redistribute vgroup 2 dnode 5
sql_error redistribute vgroup 2 dnode 5 dnode 3
sql_error redistribute vgroup 2 dnode 2 dnode 3
sql_error redistribute vgroup 2 dnode 2 dnode 2
sql_error redistribute vgroup 3 dnode 2 dnode 2
sql_error redistribute vgroup 2 dnode 2 dnode 2 dnode 3
system sh/exec.sh -n dnode5 -s start
$x = 0
step2:

4
tests/script/tsim/sma/rsmaCreateInsertQuery.sim
View File

@ -5,7 +5,7 @@ sleep 50
sql connect
print =============== create database with retentions
sql create database d0 retentions 15s:7d,1m:21d,15m:365d;
sql create database d0 retentions 5s:7d,10s:21d,15s:365d;
sql use d0
print =============== create super table and register rsma
@ -29,6 +29,8 @@ sql insert into ct1 values(now, 10);
sql insert into ct1 values(now+1s, 1);
sql insert into ct1 values(now+2s, 100);
print =============== wait maxdelay 15+1 seconds for results
sleep 16000
print =============== select * from retention level 2 from memory
sql select * from ct1;

173
tests/script/tsim/stream/schedSnode.sim Normal file
View File

@ -0,0 +1,173 @@
system sh/stop_dnodes.sh
system sh/deploy.sh -n dnode1 -i 1
system sh/deploy.sh -n dnode2 -i 2
system sh/exec.sh -n dnode1 -s start
sleep 50
sql connect
sql create snode on dnode 1
sql create database test vgroups 1;
sql create database target vgroups 1;
sql use test;
sql create stable st(ts timestamp, a int, b int , c int, d double) tags(ta int,tb int,tc int);
sql create table ts1 using st tags(1,1,1);
sql create table ts2 using st tags(2,2,2);
sql create table ts3 using st tags(3,2,2);
sql create table ts4 using st tags(4,2,2);
sql create stream stream_t1 trigger at_once into target.streamtST1 as select _wstartts, count(*) c1, count(d) c2 , sum(a) c3 , max(b) c4, min(c) c5 from st interval(10s);
sleep 1000
sql insert into ts1 values(1648791213001,1,12,3,1.0);
sql insert into ts2 values(1648791213001,1,12,3,1.0);
sql insert into ts3 values(1648791213001,1,12,3,1.0);
sql insert into ts4 values(1648791213001,1,12,3,1.0);
sql insert into ts1 values(1648791213002,NULL,NULL,NULL,NULL);
sql insert into ts2 values(1648791213002,NULL,NULL,NULL,NULL);
sql insert into ts3 values(1648791213002,NULL,NULL,NULL,NULL);
sql insert into ts4 values(1648791213002,NULL,NULL,NULL,NULL);
sql insert into ts1 values(1648791223002,2,2,3,1.1);
sql insert into ts1 values(1648791233003,3,2,3,2.1);
sql insert into ts2 values(1648791243004,4,2,43,73.1);
sql insert into ts1 values(1648791213002,24,22,23,4.1);
sql insert into ts1 values(1648791243005,4,20,3,3.1);
sql insert into ts2 values(1648791243006,4,2,3,3.1) (1648791243007,4,2,3,3.1) ;
sql insert into ts1 values(1648791243008,4,2,30,3.1) (1648791243009,4,2,3,3.1) (1648791243010,4,2,3,3.1) ;
sql insert into ts2 values(1648791243011,4,2,3,3.1) (1648791243012,34,32,33,3.1) (1648791243013,4,2,3,3.1) (1648791243014,4,2,13,3.1);
sql insert into ts1 values(1648791243005,4,42,3,3.1) (1648791243003,4,2,33,3.1) (1648791243006,4,2,3,3.1) (1648791213001,1,52,13,1.0) (1648791223001,22,22,83,1.1) ;
sql insert into ts2 values(1648791243005,4,42,3,3.1) (1648791243003,4,2,33,3.1) (1648791243006,4,2,3,3.1) (1648791213001,1,52,13,1.0) (1648791223001,22,22,83,1.1) (1648791233004,13,12,13,2.1) ;
sql insert into ts1 values(1648791243006,4,2,3,3.1) (1648791213001,1,52,13,1.0) (1648791223001,22,22,83,1.1) ;
sql insert into ts3 values(1648791223002,2,2,3,1.1);
sql insert into ts4 values(1648791233003,3,2,3,2.1);
sql insert into ts3 values(1648791243004,4,2,43,73.1);
sql insert into ts4 values(1648791213002,24,22,23,4.1);
sql insert into ts3 values(1648791243005,4,20,3,3.1);
sql insert into ts4 values(1648791243006,4,2,3,3.1) (1648791243007,4,2,3,3.1) ;
sql insert into ts3 values(1648791243008,4,2,30,3.1) (1648791243009,4,2,3,3.1) (1648791243010,4,2,3,3.1) ;
sql insert into ts4 values(1648791243011,4,2,3,3.1) (1648791243012,34,32,33,3.1) (1648791243013,4,2,3,3.1) (1648791243014,4,2,13,3.1);
sql insert into ts3 values(1648791243005,4,42,3,3.1) (1648791243003,4,2,33,3.1) (1648791243006,4,2,3,3.1) (1648791213001,1,52,13,1.0) (1648791223001,22,22,83,1.1) ;
sql insert into ts4 values(1648791243005,4,42,3,3.1) (1648791243003,4,2,33,3.1) (1648791243006,4,2,3,3.1) (1648791213001,1,52,13,1.0) (1648791223001,22,22,83,1.1) (1648791233004,13,12,13,2.1) ;
sql insert into ts3 values(1648791243006,4,2,3,3.1) (1648791213001,1,52,13,1.0) (1648791223001,22,22,83,1.1) ;
$loop_count = 0
loop1:
sql select * from target.streamtST1;
sleep 300
$loop_count = $loop_count + 1
if $loop_count == 10 then
return -1
endi
# row 0
if $data01 != 8 then
print =====data01=$data01
goto loop1
endi
if $data02 != 4 then
print =====data02=$data02
goto loop1
endi
if $data03 != 4 then
print ======$data03
return -1
endi
if $data04 != 52 then
print ======$data04
return -1
endi
if $data05 != 13 then
print ======$data05
return -1
endi
# row 1
if $data11 != 6 then
print =====data11=$data11
goto loop1
endi
if $data12 != 6 then
print =====data12=$data12
goto loop1
endi
if $data13 != 92 then
print ======$data13
return -1
endi
if $data14 != 22 then
print ======$data14
return -1
endi
if $data15 != 3 then
print ======$data15
return -1
endi
# row 2
if $data21 != 4 then
print =====data21=$data21
goto loop1
endi
if $data22 != 4 then
print =====data22=$data22
goto loop1
endi
if $data23 != 32 then
print ======$data23
return -1
endi
if $data24 != 12 then
print ======$data24
return -1
endi
if $data25 != 3 then
print ======$data25
return -1
endi
# row 3
if $data31 != 30 then
print =====data31=$data31
goto loop1
endi
if $data32 != 30 then
print =====data32=$data32
goto loop1
endi
if $data33 != 180 then
print ======$data33
return -1
endi
if $data34 != 42 then
print ======$data34
return -1
endi
if $data35 != 3 then
print ======$data35
return -1
endi
sql select _wstartts, count(*) c1, count(d) c2 , sum(a) c3 , max(b) c4, min(c) c5, avg(d) from st interval(10s);

32
tests/script/tsim/stream/windowClose.sim Normal file
View File

@ -0,0 +1,32 @@
system sh/stop_dnodes.sh
system sh/deploy.sh -n dnode1 -i 1
system sh/exec.sh -n dnode1 -s start
sleep 50
sql connect
print =============== create database
sql create database test vgroups 1
sql show databases
if $rows != 3 then
return -1
endi
print $data00 $data01 $data02
sql use test
sql create stable st(ts timestamp, a int) tags(t int);
sql create table tu1 using st tags(1);
sql create table tu2 using st tags(2);
sql create stream stream1 trigger window_close into streamt as select _wstartts, sum(a) from st interval(10s);
sql insert into tu1 values(now, 1);
sleep 300
sql select * from streamt;
if $rows != 0 then
print ======$rows
return -1
endi
system sh/exec.sh -n dnode1 -s stop -x SIGINT

113
tests/system-test/1-insert/create_table_comment.py
View File

@ -1,113 +0,0 @@
###################################################################
# Copyright (c) 2016 by TAOS Technologies, Inc.
# All rights reserved.
#
# This file is proprietary and confidential to TAOS Technologies.
# No part of this file may be reproduced, stored, transmitted,
# disclosed or used in any form or by any means other than as
# expressly provided by the written permission from Jianhui Tao
#
###################################################################
# -*- coding: utf-8 -*-
import random
import string
from util.log import *
from util.cases import *
from util.sql import *
class TDTestCase:
def init(self, conn, logSql):
tdLog.debug("start to execute %s" % __file__)
tdSql.init(conn.cursor())
def get_long_name(self, length, mode="mixed"):
"""
generate long name
mode could be numbers/letters/letters_mixed/mixed
"""
if mode == "numbers":
population = string.digits
elif mode == "letters":
population = string.ascii_letters.lower()
elif mode == "letters_mixed":
population = string.ascii_letters.upper() + string.ascii_letters.lower()
else:
population = string.ascii_letters.lower() + string.digits
return "".join(random.choices(population, k=length))
def __create_tb(self,dbname,stbname,tbname,comment):
tdSql.execute(f'create database if not exists {dbname}')
tdSql.execute(f'use {dbname}')
tdSql.execute(
f'create table {stbname} (ts timestamp,c0 int) tags(t0 int) ')
tdSql.execute(
f'create table {tbname} using {stbname} tags(1) comment "{comment}"')
def __create_normaltb(self,dbname,tbname,comment):
tdSql.execute(f'create database if not exists {dbname}')
tdSql.execute(f'use {dbname}')
tdSql.execute(
f'create table {tbname} (ts timestamp,c0 int) comment "{comment}"')
def check_comment(self):
dbname = self.get_long_name(length=10, mode="letters")
ntbname = self.get_long_name(length=5, mode="letters")
# create normal table with comment
comment = self.get_long_name(length=10, mode="letters")
self.__create_normaltb(dbname,ntbname,comment)
ntb_kv_list = tdSql.getResult("show tables")
print(ntb_kv_list)
tdSql.checkEqual(ntb_kv_list[0][8], comment)
tdSql.error('alter table {ntbname} comment "test1"')
tdSql.execute(f'drop database {dbname}')
# max length(1024)
comment = self.get_long_name(length=1024, mode="letters")
self.__create_normaltb(dbname,ntbname,comment)
ntb_kv_list = tdSql.getResult("show tables")
tdSql.checkEqual(ntb_kv_list[0][8], comment)
tdSql.execute(f'drop database {dbname}')
# error overlength
comment = self.get_long_name(length=1025, mode="letters")
tdSql.execute(f'create database if not exists {dbname}')
tdSql.execute(f'use {dbname}')
tdSql.error(f"create table ntb (ts timestamp,c0 int) comment '{comment}'")
tdSql.execute(f'drop database {dbname}')
# create child table with comment
comment = self.get_long_name(length=10, mode="letters")
stbname = self.get_long_name(length=5, mode="letters")
tbname = self.get_long_name(length=3, mode="letters")
self.__create_tb(dbname,stbname,tbname,comment)
ntb_kv_list = tdSql.getResult("show tables")
tdSql.checkEqual(ntb_kv_list[0][8], comment)
tdSql.error(f'alter table {tbname} comment "test1"')
tdSql.execute(f'drop database {dbname}')
# max length 1024
comment = self.get_long_name(length=1024, mode="letters")
self.__create_tb(dbname,ntbname,comment)
ntb_kv_list = tdSql.getResult("show tables")
tdSql.checkEqual(ntb_kv_list[0][8], comment)
tdSql.execute(f'drop database {dbname}')
# error overlength
comment = self.get_long_name(length=1025, mode="letters")
tdSql.execute(f'create database if not exists {dbname}')
tdSql.execute(f'use {dbname}')
tdSql.execute(f"create table stb (ts timestamp,c0 int) tags(t0 int)")
tdSql.error(f'create table stb_1 us stb tags(1) comment "{comment}"')
tdSql.execute(f'drop database {dbname}')
def run(self):
self.check_comment()
def stop(self):
tdSql.close()
tdLog.success("%s successfully executed" % __file__)
tdCases.addWindows(__file__, TDTestCase())
tdCases.addLinux(__file__, TDTestCase())

134
tests/system-test/1-insert/table_comment.py Normal file
View File

@ -0,0 +1,134 @@
###################################################################
# Copyright (c) 2016 by TAOS Technologies, Inc.
# All rights reserved.
#
# This file is proprietary and confidential to TAOS Technologies.
# No part of this file may be reproduced, stored, transmitted,
# disclosed or used in any form or by any means other than as
# expressly provided by the written permission from Jianhui Tao
#
###################################################################
# -*- coding: utf-8 -*-
import random
import string
from util.log import *
from util.cases import *
from util.sql import *
from util.common import *
class TDTestCase:
def init(self, conn, logSql):
tdLog.debug("start to execute %s" % __file__)
tdSql.init(conn.cursor())
# prepare data
self.ntbname = 'ntb'
self.stbname = 'stb'
self.column_dict = {
'ts':'timestamp',
'c1':'int',
'c2':'float',
'c3':'double',
'c4':'timestamp'
}
self.tag_dict = {
't0':'int'
}
self.comment_length = [0,1024]
self.error_comment_length = [1025]
self.table_type_list = ['normal_table','stable','child_table']
self.comment_flag_list = [True,False]
def __set_and_alter_comment(self,tb_type='',comment_flag= False):
column_sql = ''
tag_sql = ''
for k,v in self.column_dict.items():
column_sql += f"{k} {v},"
for k,v in self.tag_dict.items():
tag_sql += f"{k} {v},"
if tb_type == 'normal_table' or tb_type == '':
if comment_flag == False:
tdSql.execute(f'create table {self.ntbname} ({column_sql[:-1]})')
self.check_comment_info()
self.alter_comment(self.ntbname)
tdSql.execute(f'drop table {self.ntbname}')
elif comment_flag == True:
for i in self.comment_length:
comment_info = tdCom.getLongName(i)
tdSql.execute(f'create table {self.ntbname} ({column_sql[:-1]}) comment "{comment_info}"')
self.check_comment_info(comment_info)
self.alter_comment(self.ntbname)
tdSql.execute(f'drop table {self.ntbname}')
for i in self.error_comment_length:
comment_info = tdCom.getLongName(i)
tdSql.error(f'create table {self.ntbname} ({column_sql[:-1]}) comment "{comment_info}"')
elif tb_type == 'stable':
for operation in ['table','stable']:
if comment_flag == False:
tdSql.execute(f'create {operation} {self.stbname} ({column_sql[:-1]}) tags({tag_sql[:-1]})')
self.check_comment_info(None,'stable')
self.alter_comment(self.stbname,'stable')
tdSql.execute(f'drop table {self.stbname}')
elif comment_flag == True:
for i in self.comment_length:
comment_info = tdCom.getLongName(i)
tdSql.execute(f'create {operation} {self.stbname} ({column_sql[:-1]}) tags({tag_sql[:-1]}) comment "{comment_info}"')
self.check_comment_info(comment_info,'stable')
self.alter_comment(self.stbname,'stable')
tdSql.execute(f'drop table {self.stbname}')
elif tb_type == 'child_table':
tdSql.execute(f'create table if not exists {self.stbname} ({column_sql[:-1]}) tags({tag_sql[:-1]})')
if comment_flag == False:
tdSql.execute(f'create table if not exists {self.stbname}_ctb using {self.stbname} tags(1)')
self.check_comment_info()
self.alter_comment(f'{self.stbname}_ctb')
tdSql.execute(f'drop table {self.stbname}_ctb')
elif comment_flag == True:
for j in self.comment_length:
comment_info = tdCom.getLongName(j)
tdSql.execute(f'create table if not exists {self.stbname}_ctb using {self.stbname} tags(1) comment "{comment_info}"')
self.check_comment_info(comment_info)
self.alter_comment(f'{self.stbname}_ctb')
tdSql.execute(f'drop table {self.stbname}_ctb')
tdSql.execute(f'drop table {self.stbname}')
def alter_comment(self,tbname,tb_type=''):
for i in self.comment_length:
comment_info = tdCom.getLongName(i)
print(comment_info)
tdSql.execute(f'alter table {tbname} comment "{comment_info}"')
self.check_comment_info(comment_info,tb_type)
for i in self.error_comment_length:
comment_info = tdCom.getLongName(i)
tdSql.error(f'alter table {tbname} comment "{comment_info}"')
def check_comment_info(self,comment_info=None,tb_type=''):
if tb_type == '' or tb_type == 'normal_table' or tb_type == 'child_table':
tdSql.query('show tables')
if comment_info == None:
tdSql.checkEqual(tdSql.queryResult[0][8],None)
else :
tdSql.checkEqual(tdSql.queryResult[0][8],comment_info)
elif tb_type == 'stable':
tdSql.query('show stables')
if comment_info == None:
tdSql.checkEqual(tdSql.queryResult[0][6],None)
else :
tdSql.checkEqual(tdSql.queryResult[0][6],comment_info)
def comment_check_case(self,table_type,comment_flag):
tdSql.prepare()
for tb in table_type:
for flag in comment_flag:
self.__set_and_alter_comment(tb,flag)
tdSql.execute('drop database db')
def run(self):
self.comment_check_case(self.table_type_list,self.comment_flag_list)
def stop(self):
tdSql.close()
tdLog.success("%s successfully executed" % __file__)
tdCases.addWindows(__file__, TDTestCase())
tdCases.addLinux(__file__, TDTestCase())

46
tests/system-test/6-cluster/5dnode3mnodeDrop.py
View File

@ -75,7 +75,7 @@ class TDTestCase:
testCluster = False
valgrind = 0
hostname = socket.gethostname()
print(hostname)
tdLog.debug(hostname)
dnodes = []
start_port = 6030
start_port_sec = 6130
@ -102,12 +102,12 @@ class TDTestCase:
# create cluster
for dnode in self.TDDnodes.dnodes[1:]:
# print(dnode.cfgDict)
# tdLog.debug(dnode.cfgDict)
dnode_id = dnode.cfgDict["fqdn"] + ":" +dnode.cfgDict["serverPort"]
dnode_first_host = dnode.cfgDict["firstEp"].split(":")[0]
dnode_first_port = dnode.cfgDict["firstEp"].split(":")[-1]
cmd = f" taos -h {dnode_first_host} -P {dnode_first_port} -s ' create dnode \"{dnode_id} \" ' ;"
print(cmd)
tdLog.debug(cmd)
os.system(cmd)
time.sleep(2)
@ -119,26 +119,26 @@ class TDTestCase:
time.sleep(1)
tdSql.query("show mnodes;")
if tdSql.checkRows(3) :
print("mnode is three nodes")
tdLog.debug("mnode is three nodes")
if tdSql.queryResult[0][2]=='leader' :
if tdSql.queryResult[1][2]=='follower':
if tdSql.queryResult[2][2]=='follower':
print("three mnodes is ready in 10s")
tdLog.debug("three mnodes is ready in 10s")
break
elif tdSql.queryResult[0][2]=='follower' :
if tdSql.queryResult[1][2]=='leader':
if tdSql.queryResult[2][2]=='follower':
print("three mnodes is ready in 10s")
tdLog.debug("three mnodes is ready in 10s")
break
elif tdSql.queryResult[0][2]=='follower' :
if tdSql.queryResult[1][2]=='follower':
if tdSql.queryResult[2][2]=='leader':
print("three mnodes is ready in 10s")
tdLog.debug("three mnodes is ready in 10s")
break
count+=1
else:
print(tdSql.queryResult)
print("three mnodes is not ready in 10s ")
tdLog.debug(tdSql.queryResult)
tdLog.debug("three mnodes is not ready in 10s ")
return -1
tdSql.query("show mnodes;")
@ -156,19 +156,19 @@ class TDTestCase:
time.sleep(1)
tdSql.query("show mnodes;")
if tdSql.checkRows(3) :
print("mnode is three nodes")
tdLog.debug("mnode is three nodes")
if tdSql.queryResult[0][2]=='offline' :
if tdSql.queryResult[1][2]=='leader':
if tdSql.queryResult[2][2]=='follower':
print("stop mnodes on dnode 2 successfully in 10s")
tdLog.debug("stop mnodes on dnode 2 successfully in 10s")
break
elif tdSql.queryResult[1][2]=='follower':
if tdSql.queryResult[2][2]=='leader':
print("stop mnodes on dnode 2 successfully in 10s")
tdLog.debug("stop mnodes on dnode 2 successfully in 10s")
break
count+=1
else:
print("stop mnodes on dnode 2 failed in 10s ")
tdLog.debug("stop mnodes on dnode 2 failed in 10s ")
return -1
tdSql.error("drop mnode on dnode 1;")
@ -188,15 +188,15 @@ class TDTestCase:
time.sleep(1)
tdSql.query("show mnodes;")
if tdSql.checkRows(3) :
print("mnode is three nodes")
tdLog.debug("mnode is three nodes")
if tdSql.queryResult[0][2]=='leader' :
if tdSql.queryResult[1][2]=='offline':
if tdSql.queryResult[2][2]=='follower':
print("stop mnodes on dnode 2 successfully in 10s")
tdLog.debug("stop mnodes on dnode 2 successfully in 10s")
break
count+=1
else:
print("stop mnodes on dnode 2 failed in 10s ")
tdLog.debug("stop mnodes on dnode 2 failed in 10s ")
return -1
tdSql.error("drop mnode on dnode 2;")
@ -218,15 +218,15 @@ class TDTestCase:
time.sleep(1)
tdSql.query("show mnodes;")
if tdSql.checkRows(3) :
print("mnode is three nodes")
tdLog.debug("mnode is three nodes")
if tdSql.queryResult[0][2]=='leader' :
if tdSql.queryResult[2][2]=='offline':
if tdSql.queryResult[1][2]=='follower':
print("stop mnodes on dnode 3 successfully in 10s")
tdLog.debug("stop mnodes on dnode 3 successfully in 10s")
break
count+=1
else:
print("stop mnodes on dnode 3 failed in 10s")
tdLog.debug("stop mnodes on dnode 3 failed in 10s")
return -1
tdSql.error("drop mnode on dnode 3;")
tdSql.query("show mnodes;")
@ -268,7 +268,7 @@ class TDTestCase:
tdSql.error("drop mnode on dnode 1")
tdSql.query("show dnodes;")
print(tdSql.queryResult)
tdLog.debug(tdSql.queryResult)
# drop follower of mnode
dropcount =0
@ -282,7 +282,7 @@ class TDTestCase:
time.sleep(1)
tdSql.query("show mnodes;")
if tdSql.checkRows(2):
print("drop mnode %d successfully"%(i+1))
tdLog.debug("drop mnode %d successfully"%(i+1))
break
count+=1
tdLog.debug("create mnode on dnode %d"%(i+1))
@ -292,7 +292,7 @@ class TDTestCase:
time.sleep(1)
tdSql.query("show mnodes;")
if tdSql.checkRows(3):
print("drop mnode %d successfully"%(i+1))
tdLog.debug("drop mnode %d successfully"%(i+1))
break
count+=1
dropcount+=1
@ -307,7 +307,7 @@ class TDTestCase:
def run(self):
# print(self.master_dnode.cfgDict)
# tdLog.debug(self.master_dnode.cfgDict)
self.buildcluster(5)
self.five_dnode_three_mnode()

38
tests/system-test/6-cluster/5dnode3mnodeStop.py
View File

@ -99,12 +99,12 @@ class TDTestCase:
# create cluster
for dnode in self.TDDnodes.dnodes[1:]:
# print(dnode.cfgDict)
# tdLog.debug(dnode.cfgDict)
dnode_id = dnode.cfgDict["fqdn"] + ":" +dnode.cfgDict["serverPort"]
dnode_first_host = dnode.cfgDict["firstEp"].split(":")[0]
dnode_first_port = dnode.cfgDict["firstEp"].split(":")[-1]
cmd = f" taos -h {dnode_first_host} -P {dnode_first_port} -s ' create dnode \"{dnode_id} \" ' ;"
print(cmd)
tdLog.debug(cmd)
os.system(cmd)
time.sleep(2)
@ -116,25 +116,25 @@ class TDTestCase:
time.sleep(1)
tdSql.query("show mnodes;")
if tdSql.checkRows(3) :
print("mnode is three nodes")
tdLog.debug("mnode is three nodes")
if tdSql.queryResult[0][2]=='leader' :
if tdSql.queryResult[1][2]=='follower':
if tdSql.queryResult[2][2]=='follower':
print("three mnodes is ready in 10s")
tdLog.debug("three mnodes is ready in 10s")
break
elif tdSql.queryResult[0][2]=='follower' :
if tdSql.queryResult[1][2]=='leader':
if tdSql.queryResult[2][2]=='follower':
print("three mnodes is ready in 10s")
tdLog.debug("three mnodes is ready in 10s")
break
elif tdSql.queryResult[0][2]=='follower' :
if tdSql.queryResult[1][2]=='follower':
if tdSql.queryResult[2][2]=='leader':
print("three mnodes is ready in 10s")
tdLog.debug("three mnodes is ready in 10s")
break
count+=1
else:
print("three mnodes is not ready in 10s ")
tdLog.debug("three mnodes is not ready in 10s ")
return -1
tdSql.query("show mnodes;")
@ -152,19 +152,19 @@ class TDTestCase:
time.sleep(1)
tdSql.query("show mnodes;")
if tdSql.checkRows(3) :
print("mnode is three nodes")
tdLog.debug("mnode is three nodes")
if tdSql.queryResult[0][2]=='offline' :
if tdSql.queryResult[1][2]=='leader':
if tdSql.queryResult[2][2]=='follower':
print("stop mnodes on dnode 2 successfully in 10s")
tdLog.debug("stop mnodes on dnode 2 successfully in 10s")
break
elif tdSql.queryResult[1][2]=='follower':
if tdSql.queryResult[2][2]=='leader':
print("stop mnodes on dnode 2 successfully in 10s")
tdLog.debug("stop mnodes on dnode 2 successfully in 10s")
break
count+=1
else:
print("stop mnodes on dnode 2 failed in 10s ")
tdLog.debug("stop mnodes on dnode 2 failed in 10s ")
return -1
tdSql.error("drop mnode on dnode 1;")
@ -184,15 +184,15 @@ class TDTestCase:
time.sleep(1)
tdSql.query("show mnodes;")
if tdSql.checkRows(3) :
print("mnode is three nodes")
tdLog.debug("mnode is three nodes")
if tdSql.queryResult[0][2]=='leader' :
if tdSql.queryResult[1][2]=='offline':
if tdSql.queryResult[2][2]=='follower':
print("stop mnodes on dnode 2 successfully in 10s")
tdLog.debug("stop mnodes on dnode 2 successfully in 10s")
break
count+=1
else:
print("stop mnodes on dnode 2 failed in 10s ")
tdLog.debug("stop mnodes on dnode 2 failed in 10s ")
return -1
tdSql.error("drop mnode on dnode 2;")
@ -214,15 +214,15 @@ class TDTestCase:
time.sleep(1)
tdSql.query("show mnodes;")
if tdSql.checkRows(3) :
print("mnode is three nodes")
tdLog.debug("mnode is three nodes")
if tdSql.queryResult[0][2]=='leader' :
if tdSql.queryResult[2][2]=='offline':
if tdSql.queryResult[1][2]=='follower':
print("stop mnodes on dnode 3 successfully in 10s")
tdLog.debug("stop mnodes on dnode 3 successfully in 10s")
break
count+=1
else:
print("stop mnodes on dnode 3 failed in 10s")
tdLog.debug("stop mnodes on dnode 3 failed in 10s")
return -1
tdSql.error("drop mnode on dnode 3;")
tdSql.query("show mnodes;")
@ -262,7 +262,7 @@ class TDTestCase:
tdSql.error("create mnode on dnode 2")
tdSql.query("show dnodes;")
print(tdSql.queryResult)
tdLog.debug(tdSql.queryResult)
tdLog.debug("stop and follower of mnode")
self.TDDnodes.stoptaosd(2)
@ -303,7 +303,7 @@ class TDTestCase:
def run(self):
# print(self.master_dnode.cfgDict)
# tdLog.debug(self.master_dnode.cfgDict)
self.buildcluster(5)
self.five_dnode_three_mnode(5)

54
tests/system-test/6-cluster/5dnode3mnodeStopInsert.py
View File

@ -77,7 +77,7 @@ class TDTestCase:
for couti in range(countstart,countstop):
tdLog.debug("drop database if exists db%d" %couti)
tdSql.execute("drop database if exists db%d" %couti)
print("create database if not exists db%d replica 1 duration 300" %couti)
tdLog.debug("create database if not exists db%d replica 1 duration 300" %couti)
tdSql.execute("create database if not exists db%d replica 1 duration 300" %couti)
tdSql.execute("use db%d" %couti)
tdSql.execute(
@ -126,12 +126,12 @@ class TDTestCase:
# create cluster
for dnode in self.TDDnodes.dnodes[1:]:
# print(dnode.cfgDict)
# tdLog.debug(dnode.cfgDict)
dnode_id = dnode.cfgDict["fqdn"] + ":" +dnode.cfgDict["serverPort"]
dnode_first_host = dnode.cfgDict["firstEp"].split(":")[0]
dnode_first_port = dnode.cfgDict["firstEp"].split(":")[-1]
cmd = f" taos -h {dnode_first_host} -P {dnode_first_port} -s ' create dnode \"{dnode_id} \" ' ;"
print(cmd)
tdLog.debug(cmd)
os.system(cmd)
time.sleep(2)
@ -144,22 +144,22 @@ class TDTestCase:
statusReadyBumber=0
tdSql.query("show dnodes;")
if tdSql.checkRows(dnodenumber) :
print("dnode is %d nodes"%dnodenumber)
tdLog.debug("dnode is %d nodes"%dnodenumber)
for i in range(dnodenumber):
if tdSql.queryResult[i][4] !='ready' :
status=tdSql.queryResult[i][4]
print("dnode:%d status is %s "%(i,status))
tdLog.debug("dnode:%d status is %s "%(i,status))
break
else:
statusReadyBumber+=1
print(statusReadyBumber)
tdLog.debug(statusReadyBumber)
if statusReadyBumber == dnodenumber :
print("all of %d mnodes is ready in 10s "%dnodenumber)
tdLog.debug("all of %d mnodes is ready in 10s "%dnodenumber)
return True
break
count+=1
else:
print("%d mnodes is not ready in 10s "%dnodenumber)
tdLog.debug("%d mnodes is not ready in 10s "%dnodenumber)
return False
@ -169,25 +169,25 @@ class TDTestCase:
time.sleep(1)
tdSql.query("show mnodes;")
if tdSql.checkRows(3) :
print("mnode is three nodes")
tdLog.debug("mnode is three nodes")
if tdSql.queryResult[0][2]=='leader' :
if tdSql.queryResult[1][2]=='follower':
if tdSql.queryResult[2][2]=='follower':
print("three mnodes is ready in 10s")
tdLog.debug("three mnodes is ready in 10s")
break
elif tdSql.queryResult[0][2]=='follower' :
if tdSql.queryResult[1][2]=='leader':
if tdSql.queryResult[2][2]=='follower':
print("three mnodes is ready in 10s")
tdLog.debug("three mnodes is ready in 10s")
break
elif tdSql.queryResult[0][2]=='follower' :
if tdSql.queryResult[1][2]=='follower':
if tdSql.queryResult[2][2]=='leader':
print("three mnodes is ready in 10s")
tdLog.debug("three mnodes is ready in 10s")
break
count+=1
else:
print("three mnodes is not ready in 10s ")
tdLog.debug("three mnodes is not ready in 10s ")
return -1
tdSql.query("show mnodes;")
@ -205,19 +205,19 @@ class TDTestCase:
time.sleep(1)
tdSql.query("show mnodes;")
if tdSql.checkRows(3) :
print("mnode is three nodes")
tdLog.debug("mnode is three nodes")
if tdSql.queryResult[0][2]=='offline' :
if tdSql.queryResult[1][2]=='leader':
if tdSql.queryResult[2][2]=='follower':
print("stop mnodes on dnode 2 successfully in 10s")
tdLog.debug("stop mnodes on dnode 2 successfully in 10s")
break
elif tdSql.queryResult[1][2]=='follower':
if tdSql.queryResult[2][2]=='leader':
print("stop mnodes on dnode 2 successfully in 10s")
tdLog.debug("stop mnodes on dnode 2 successfully in 10s")
break
count+=1
else:
print("stop mnodes on dnode 2 failed in 10s ")
tdLog.debug("stop mnodes on dnode 2 failed in 10s ")
return -1
tdSql.error("drop mnode on dnode 1;")
@ -237,15 +237,15 @@ class TDTestCase:
time.sleep(1)
tdSql.query("show mnodes;")
if tdSql.checkRows(3) :
print("mnode is three nodes")
tdLog.debug("mnode is three nodes")
if tdSql.queryResult[0][2]=='leader' :
if tdSql.queryResult[1][2]=='offline':
if tdSql.queryResult[2][2]=='follower':
print("stop mnodes on dnode 2 successfully in 10s")
tdLog.debug("stop mnodes on dnode 2 successfully in 10s")
break
count+=1
else:
print("stop mnodes on dnode 2 failed in 10s ")
tdLog.debug("stop mnodes on dnode 2 failed in 10s ")
return -1
tdSql.error("drop mnode on dnode 2;")
@ -267,15 +267,15 @@ class TDTestCase:
time.sleep(1)
tdSql.query("show mnodes;")
if tdSql.checkRows(3) :
print("mnode is three nodes")
tdLog.debug("mnode is three nodes")
if tdSql.queryResult[0][2]=='leader' :
if tdSql.queryResult[2][2]=='offline':
if tdSql.queryResult[1][2]=='follower':
print("stop mnodes on dnode 3 successfully in 10s")
tdLog.debug("stop mnodes on dnode 3 successfully in 10s")
break
count+=1
else:
print("stop mnodes on dnode 3 failed in 10s")
tdLog.debug("stop mnodes on dnode 3 failed in 10s")
return -1
tdSql.error("drop mnode on dnode 3;")
tdSql.query("show mnodes;")
@ -311,7 +311,7 @@ class TDTestCase:
tdSql.error("create mnode on dnode 2")
tdSql.query("show dnodes;")
print(tdSql.queryResult)
tdLog.debug(tdSql.queryResult)
tdLog.debug("stop all of mnode ")
stopcount =0
@ -325,10 +325,10 @@ class TDTestCase:
self.TDDnodes.starttaosd(i+1)
if self.checkdnodes(5):
print("123")
tdLog.debug("123")
threads.join()
else:
print("456")
tdLog.debug("456")
self.stop_thread(threads)
assert 1 == 2 ,"some dnode started failed"
return False
@ -345,7 +345,7 @@ class TDTestCase:
def run(self):
# print(self.master_dnode.cfgDict)
# tdLog.debug(self.master_dnode.cfgDict)
self.buildcluster(5)
self.five_dnode_three_mnode(5)

38
tests/system-test/7-tmq/subscribeDb3.py
View File

@ -54,9 +54,11 @@ class TDTestCase:
tdSql.query("create database if not exists %s vgroups 1"%(cdbName))
tdSql.query("drop table if exists %s.consumeinfo "%(cdbName))
tdSql.query("drop table if exists %s.consumeresult "%(cdbName))
tdSql.query("drop table if exists %s.notifyinfo "%(cdbName))
tdSql.query("create table %s.consumeinfo (ts timestamp, consumerid int, topiclist binary(1024), keylist binary(1024), expectmsgcnt bigint, ifcheckdata int, ifmanualcommit int)"%cdbName)
tdSql.query("create table %s.consumeresult (ts timestamp, consumerid int, consummsgcnt bigint, consumrowcnt bigint, checkresult int)"%cdbName)
tdSql.query("create table %s.notifyinfo (ts timestamp, cmdid int, consumerid int)"%cdbName)
def insertConsumerInfo(self,consumerId, expectrowcnt,topicList,keyList,ifcheckdata,ifmanualcommit,cdbName='cdb'):
sql = "insert into %s.consumeinfo values "%cdbName
@ -64,6 +66,27 @@ class TDTestCase:
tdLog.info("consume info sql: %s"%sql)
tdSql.query(sql)
def getStartConsumeNotifyFromTmqsim(self,cdbName='cdb'):
while 1:
tdSql.query("select * from %s.notifyinfo"%cdbName)
#tdLog.info("row: %d, %l64d, %l64d"%(tdSql.getData(0, 1),tdSql.getData(0, 2),tdSql.getData(0, 3))
if (tdSql.getRows() == 1) and (tdSql.getData(0, 1) == 0):
break
else:
time.sleep(0.1)
return
def getStartCommitNotifyFromTmqsim(self,cdbName='cdb'):
while 1:
tdSql.query("select * from %s.notifyinfo"%cdbName)
#tdLog.info("row: %d, %l64d, %l64d"%(tdSql.getData(0, 1),tdSql.getData(0, 2),tdSql.getData(0, 3))
if tdSql.getRows() == 2 :
print(tdSql.getData(0, 1), tdSql.getData(1, 1))
if tdSql.getData(1, 1) == 1:
break
time.sleep(0.1)
return
def selectConsumeResult(self,expectRows,cdbName='cdb'):
resultList=[]
while 1:
@ -72,7 +95,7 @@ class TDTestCase:
if tdSql.getRows() == expectRows:
break
else:
time.sleep(5)
time.sleep(1)
for i in range(expectRows):
tdLog.info ("ts: %s, consume id: %d, consume msgs: %d, consume rows: %d"%(tdSql.getData(i , 0), tdSql.getData(i , 1), tdSql.getData(i , 2), tdSql.getData(i , 3)))
@ -207,7 +230,9 @@ class TDTestCase:
showRow = 1
self.startTmqSimProcess(buildPath,cfgPath,pollDelay,parameterDict["dbName"],showMsg, showRow)
time.sleep(2)
tdLog.info("wait the notify info of start consume")
self.getStartConsumeNotifyFromTmqsim()
tdLog.info("pkill consume processor")
if (platform.system().lower() == 'windows'):
os.system("TASKKILL /F /IM tmq_sim.exe")
@ -282,14 +307,17 @@ class TDTestCase:
showRow = 1
self.startTmqSimProcess(buildPath,cfgPath,pollDelay,parameterDict["dbName"],showMsg, showRow)
time.sleep(6)
# time.sleep(6)
tdLog.info("start to wait commit notify")
self.getStartCommitNotifyFromTmqsim()
tdLog.info("pkill consume processor")
if (platform.system().lower() == 'windows'):
os.system("TASKKILL /F /IM tmq_sim.exe")
else:
os.system('pkill tmq_sim')
expectRows = 0
resultList = self.selectConsumeResult(expectRows)
# expectRows = 0
# resultList = self.selectConsumeResult(expectRows)
# wait for data ready
prepareEnvThread.join()

2
tests/system-test/fulltest.sh
View File

@ -22,7 +22,7 @@ python3 ./test.py -f 1-insert/opentsdb_json_taosc_insert.py
python3 ./test.py -f 1-insert/alter_stable.py
python3 ./test.py -f 1-insert/alter_table.py
python3 ./test.py -f 1-insert/insertWithMoreVgroup.py
# python3 ./test.py -f 1-inerst/create_table_comment.py
python3 ./test.py -f 1-insert/table_comment.py
python3 ./test.py -f 2-query/between.py
python3 ./test.py -f 2-query/distinct.py
python3 ./test.py -f 2-query/varchar.py

6
tests/test/c/sdbDump.c
View File

@ -114,8 +114,10 @@ void dumpStb(SSdb *pSdb, SJson *json) {
tjsonAddIntegerToObject(item, "tagVer", pObj->tagVer);
tjsonAddIntegerToObject(item, "colVer", pObj->colVer);
tjsonAddIntegerToObject(item, "nextColId", pObj->nextColId);
tjsonAddIntegerToObject(item, "xFilesFactor", pObj->xFilesFactor * 10000);
tjsonAddIntegerToObject(item, "delay", pObj->delay);
tjsonAddIntegerToObject(item, "watermark1", pObj->watermark[0]);
tjsonAddIntegerToObject(item, "watermark2", pObj->watermark[1]);
tjsonAddIntegerToObject(item, "maxdelay1", pObj->maxdelay[0]);
tjsonAddIntegerToObject(item, "maxdelay2", pObj->maxdelay[1]);
tjsonAddIntegerToObject(item, "ttl", pObj->ttl);
tjsonAddIntegerToObject(item, "numOfColumns", pObj->numOfColumns);
tjsonAddIntegerToObject(item, "numOfTags", pObj->numOfTags);

87
tests/test/c/tmqSim.c
View File

@ -34,6 +34,12 @@
#define MAX_CONSUMER_THREAD_CNT (16)
#define MAX_VGROUP_CNT (32)
typedef enum {
NOTIFY_CMD_START_CONSUM,
NOTIFY_CMD_START_COMMIT,
NOTIFY_CMD_ID_BUTT
}NOTIFY_CMD_ID;
typedef struct {
TdThread thread;
int32_t consumerId;
@ -67,6 +73,8 @@ typedef struct {
int32_t rowsOfPerVgroups[MAX_VGROUP_CNT][2]; // [i][0]: vgroup id, [i][1]: rows of consume
int64_t ts;
TAOS* taos;
} SThreadInfo;
typedef struct {
@ -339,8 +347,37 @@ int queryDB(TAOS* taos, char* command) {
return 0;
}
static void appNothing(void* param, TAOS_RES* res, int32_t numOfRows) {
}
int32_t notifyMainScript(SThreadInfo* pInfo, int32_t cmdId) {
char sqlStr[1024] = {0};
int64_t now = taosGetTimestampMs();
// schema: ts timestamp, consumerid int, consummsgcnt bigint, checkresult int
sprintf(sqlStr, "insert into %s.notifyinfo values (%"PRId64", %d, %d)",
g_stConfInfo.cdbName,
now,
cmdId,
pInfo->consumerId);
taos_query_a(pInfo->taos, sqlStr, appNothing, NULL);
taosFprintfFile(g_fp, "notifyMainScript success, sql: %s\n", sqlStr);
return 0;
}
static int32_t g_once_commit_flag = 0;
static void tmq_commit_cb_print(tmq_t* tmq, int32_t code, void* param) {
pError("tmq_commit_cb_print() commit %d\n", code);
if (0 == g_once_commit_flag) {
g_once_commit_flag = 1;
notifyMainScript((SThreadInfo*)param, (int32_t)NOTIFY_CMD_START_COMMIT);
}
taosFprintfFile(g_fp, "tmq_commit_cb_print() be called\n");
}
void build_consumer(SThreadInfo* pInfo) {
@ -353,7 +390,7 @@ void build_consumer(SThreadInfo* pInfo) {
// tmq_conf_set(conf, "td.connect.db", g_stConfInfo.dbName);
tmq_conf_set_auto_commit_cb(conf, tmq_commit_cb_print, NULL);
tmq_conf_set_auto_commit_cb(conf, tmq_commit_cb_print, pInfo);
// tmq_conf_set(conf, "group.id", "cgrp1");
for (int32_t i = 0; i < pInfo->numOfKey; i++) {
@ -392,9 +429,6 @@ void build_topic_list(SThreadInfo* pInfo) {
int32_t saveConsumeResult(SThreadInfo* pInfo) {
char sqlStr[1024] = {0};
TAOS* pConn = taos_connect(NULL, "root", "taosdata", NULL, 0);
assert(pConn != NULL);
int64_t now = taosGetTimestampMs();
// schema: ts timestamp, consumerid int, consummsgcnt bigint, checkresult int
@ -404,7 +438,7 @@ int32_t saveConsumeResult(SThreadInfo* pInfo) {
char tmpString[128];
taosFprintfFile(g_fp, "%s, consume id %d result: %s\n", getCurrentTimeString(tmpString), pInfo->consumerId, sqlStr);
TAOS_RES* pRes = taos_query(pConn, sqlStr);
TAOS_RES* pRes = taos_query(pInfo->taos, sqlStr);
if (taos_errno(pRes) != 0) {
pError("error in save consumeinfo, reason:%s\n", taos_errstr(pRes));
taos_free_result(pRes);
@ -413,38 +447,14 @@ int32_t saveConsumeResult(SThreadInfo* pInfo) {
taos_free_result(pRes);
#if 0
// vgroups
for (i = 0; i < pInfo->numOfVgroups; i++) {
// schema: ts timestamp, consumerid int, consummsgcnt bigint, checkresult int
sprintf(sqlStr, "insert into %s.vgroup_%d values (%"PRId64", %d, %" PRId64 ", %" PRId64 ", %d)",
g_stConfInfo.cdbName,
now,
pInfo->consumerId,
pInfo->consumeMsgCnt,
pInfo->consumeRowCnt,
pInfo->checkresult);
char tmpString[128];
taosFprintfFile(g_fp, "%s, consume id %d result: %s\n", getCurrentTimeString(tmpString), pInfo->consumerId ,sqlStr);
TAOS_RES* pRes = taos_query(pConn, sqlStr);
if (taos_errno(pRes) != 0) {
pError("error in save consumeinfo, reason:%s\n", taos_errstr(pRes));
taos_free_result(pRes);
exit(-1);
}
taos_free_result(pRes);
}
#endif
return 0;
}
void loop_consume(SThreadInfo* pInfo) {
int32_t code;
int32_t once_flag = 0;
int64_t totalMsgs = 0;
int64_t totalRows = 0;
@ -465,6 +475,11 @@ void loop_consume(SThreadInfo* pInfo) {
totalMsgs++;
if (0 == once_flag) {
once_flag = 1;
notifyMainScript(pInfo, NOTIFY_CMD_START_CONSUM);
}
if (totalRows >= pInfo->expectMsgCnt) {
char tmpString[128];
taosFprintfFile(g_fp, "%s over than expect rows, so break consume\n", getCurrentTimeString(tmpString));
@ -489,6 +504,12 @@ void* consumeThreadFunc(void* param) {
SThreadInfo* pInfo = (SThreadInfo*)param;
pInfo->taos = taos_connect(NULL, "root", "taosdata", NULL, 0);
if (pInfo->taos == NULL) {
taosFprintfFile(g_fp, "taos_connect() fail, can not notify and save consume result to main scripte\n");
exit(-1);
}
build_consumer(pInfo);
build_topic_list(pInfo);
if ((NULL == pInfo->tmq) || (NULL == pInfo->topicList)) {
@ -508,7 +529,6 @@ void* consumeThreadFunc(void* param) {
loop_consume(pInfo);
if (pInfo->ifManualCommit) {
taosFprintfFile(g_fp, "tmq_commit() manual commit when consume end.\n");
pPrint("tmq_commit() manual commit when consume end.\n");
/*tmq_commit(pInfo->tmq, NULL, 0);*/
tmq_commit_sync(pInfo->tmq, NULL);
@ -539,6 +559,9 @@ void* consumeThreadFunc(void* param) {
taosFprintfFile(g_fp, "vgroups: %04d, rows: %d\n", pInfo->rowsOfPerVgroups[i][0], pInfo->rowsOfPerVgroups[i][1]);
}
taos_close(pInfo->taos);
pInfo->taos = NULL;
return NULL;
}