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

Merge branch '3.0' into feature/TD-14481-3.0

This commit is contained in:
Cary Xu 2022-06-11 15:16:20 +08:00
parent ed829455a9 3097bf2af8
commit 25911e12a4
108 changed files with 3547 additions and 1558 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

8
docs-en/14-reference/03-connector/csharp.mdx
View File

@ -48,7 +48,7 @@ Please refer to [version support list](/reference/connector#version-support)
* Install the [.NET SDK](https://dotnet.microsoft.com/download)
* [Nuget Client](https://docs.microsoft.com/en-us/nuget/install-nuget-client-tools) (optional installation)
* Install TDengine client driver, please refer to [Install client driver](/reference/connector#Install client driver) for details
* Install TDengine client driver, please refer to [Install client driver](/reference/connector/#install-client-driver) for details
### Install via dotnet CLI
@ -57,7 +57,7 @@ Please refer to [version support list](/reference/connector#version-support)
You can reference the `TDengine.Connector` published in Nuget to the current project via the `dotnet` command under the path of the existing .NET project.
``` bash
```
dotnet add package TDengine.Connector
```
@ -66,7 +66,7 @@ dotnet add package TDengine.Connector
You can download TDengine's source code and directly reference the latest version of the TDengine.Connector library
```bash
```
git clone https://github.com/taosdata/TDengine.git
cd TDengine/src/connector/C#/src/
cp -r TDengineDriver/ myProject
@ -79,7 +79,7 @@ dotnet add TDengineDriver/TDengineDriver.csproj
## Create a connection
``` C#
```csharp
using TDengineDriver;
namespace TDengineExample

19
docs-en/14-reference/03-connector/go.mdx
View File

@ -55,25 +55,27 @@ A "REST connection" is a connection between the application and the TDengine ins
### Pre-installation
* Install Go development environment (Go 1.14 and above, GCC 4.8.5 and above)
* If you use the native connector, please install the TDengine client driver. Please refer to [Install Client Driver](/reference/connector#Install Client Driver) for specific steps
- Install Go development environment (Go 1.14 and above, GCC 4.8.5 and above)
- If you use the native connector, please install the TDengine client driver. Please refer to [Install Client Driver](/reference/connector/#install-client-driver) for specific steps
Configure the environment variables and check the command.
* ```go env``
* ```gcc -v``
* `go env`
* `gcc -v`
### Use go get to install
``go get -u github.com/taosdata/driver-go/v2@develop``
```
go get -u github.com/taosdata/driver-go/v2@develop
```
### Manage with go mod
1. Initialize the project with the `go mod` command.
``text
go mod init taos-demo
```text
go mod init taos-demo
```
2. Introduce taosSql
@ -88,7 +90,7 @@ Configure the environment variables and check the command.
```text
go mod tidy
``` 4.
```
4. Run the program with `go run taos-demo` or compile the binary with the `go build` command.
@ -309,6 +311,7 @@ func main() {
:::info
This API is created successfully without checking permissions, but only when you execute a Query or Exec, and check if user/password/host/port is legal.
:::
* `func (db *DB) Exec(query string, args . .interface{}) (Result, error)`

11
docs-en/14-reference/03-connector/java.mdx
View File

@ -69,7 +69,7 @@ Before using Java Connector to connect to the database, the following conditions
### Install the connectors
<Tabs defaultValue="maven">
<TabItem value="maven" label="install via Maven">
<TabItem value="maven" label="Install via Maven">
- [sonatype](https://search.maven.org/artifact/com.taosdata.jdbc/taos-jdbcdriver)
- [mvnrepository](https://mvnrepository.com/artifact/com.taosdata.jdbc/taos-jdbcdriver)
@ -77,7 +77,7 @@ Before using Java Connector to connect to the database, the following conditions
Add following dependency in the `pom.xml` file of your Maven project:
```xml-dtd
```xml
<dependency>
<groupId>com.taosdata.jdbc</groupId>
<artifactId>taos-jdbcdriver</artifactId>
@ -90,7 +90,7 @@ Add following dependency in the `pom.xml` file of your Maven project:
You can build Java connector from source code after cloning the TDengine project:
```shell
```
git clone https://github.com/taosdata/taos-connector-jdbc.git
cd taos-connector-jdbc
mvn clean install -Dmaven.test.skip=true
@ -173,7 +173,10 @@ In the above example, JDBC uses the client's configuration file to establish a c
In TDengine, as long as one node in firstEp and secondEp is valid, the connection to the cluster can be established normally.
> **Note**: The configuration file here refers to the configuration file on the machine where the application that calls the JDBC Connector is located, the default path is `/etc/taos/taos.cfg` on Linux, and the default path is `C://TDengine/cfg/taos.cfg` on Windows.
:::note
The configuration file here refers to the configuration file on the machine where the application that calls the JDBC Connector is located, the default path is `/etc/taos/taos.cfg` on Linux, and the default path is `C://TDengine/cfg/taos.cfg` on Windows.
:::
</TabItem>
<TabItem value="rest" label="REST connection">

6
docs-en/14-reference/03-connector/rust.mdx
View File

@ -45,15 +45,15 @@ Add the [libtaos][libtaos] dependency to the [Rust](https://rust-lang.org) proje
<Tabs defaultValue="native">
<TabItem value="native" label="native connection">
Add [libtaos][libtaos] to the ``Cargo.toml`'' file.
Add [libtaos][libtaos] to the `Cargo.toml` file.
``toml
```toml
[dependencies]
# use default feature
libtaos = "*"
```
</TabItem
</TabItem>
<TabItem value="rest" label="REST connection">
Add [libtaos][libtaos] to the `Cargo.toml` file and enable the `rest` feature.

3
include/client/taos.h
View File

@ -240,7 +240,10 @@ DLL_EXPORT const char *tmq_err2str(tmq_resp_err_t);
DLL_EXPORT tmq_resp_err_t tmq_subscribe(tmq_t *tmq, const tmq_list_t *topic_list);
DLL_EXPORT tmq_resp_err_t tmq_unsubscribe(tmq_t *tmq);
DLL_EXPORT tmq_resp_err_t tmq_subscription(tmq_t *tmq, tmq_list_t **topics);
// timeout: -1 means infinitely waiting
DLL_EXPORT TAOS_RES *tmq_consumer_poll(tmq_t *tmq, int64_t timeout);
DLL_EXPORT tmq_resp_err_t tmq_consumer_close(tmq_t *tmq);
DLL_EXPORT tmq_resp_err_t tmq_commit_sync(tmq_t *tmq, const tmq_topic_vgroup_list_t *offsets);
DLL_EXPORT void tmq_commit_async(tmq_t *tmq, const tmq_topic_vgroup_list_t *offsets, tmq_commit_cb *cb, void *param);

3
include/common/tcommon.h
View File

@ -50,6 +50,7 @@ typedef enum EStreamType {
STREAM_INVERT,
STREAM_REPROCESS,
STREAM_INVALID,
STREAM_GET_ALL,
} EStreamType;
typedef struct {
@ -116,6 +117,8 @@ typedef struct SQueryTableDataCond {
int32_t type; // data block load type:
int32_t numOfTWindows;
STimeWindow* twindows;
int32_t startVersion;
int32_t endVersion;
} SQueryTableDataCond;
void* blockDataDestroy(SSDataBlock* pBlock);

5
include/common/tmsg.h
View File

@ -1494,10 +1494,9 @@ typedef struct {
int32_t code;
} STaskDropRsp;
#define STREAM_TRIGGER_AT_ONCE_SMA 0
#define STREAM_TRIGGER_AT_ONCE 1
#define STREAM_TRIGGER_WINDOW_CLOSE 2
#define STREAM_TRIGGER_WINDOW_CLOSE_SMA 3
#define STREAM_TRIGGER_MAX_DELAY 3
typedef struct {
char name[TSDB_TABLE_FNAME_LEN];
@ -2488,7 +2487,7 @@ typedef struct {
int64_t offset;
int64_t sliding;
int64_t dstTbUid;
int32_t dstVgId; // for stream
int32_t dstVgId;
SEpSet epSet;
char* expr;
} STableIndexInfo;

2
include/common/tmsgdef.h
View File

@ -220,7 +220,7 @@ enum {
TD_DEF_MSG_TYPE(TDMT_MON_QM_LOAD, "monitor-qload", NULL, NULL)
TD_NEW_MSG_SEG(TDMT_SYNC_MSG)
TD_DEF_MSG_TYPE(TDMT_SYNC_TIMEOUT, "sync-timeout", NULL, NULL)
TD_DEF_MSG_TYPE(TDMT_SYNC_TIMEOUT, "sync-timer", NULL, NULL)
TD_DEF_MSG_TYPE(TDMT_SYNC_PING, "sync-ping", NULL, NULL)
TD_DEF_MSG_TYPE(TDMT_SYNC_PING_REPLY, "sync-ping-reply", NULL, NULL)
TD_DEF_MSG_TYPE(TDMT_SYNC_CLIENT_REQUEST, "sync-client-request", NULL, NULL)

47
include/libs/catalog/catalog.h
View File

@ -21,13 +21,13 @@ extern "C" {
#endif
#include "os.h"
#include "taosdef.h"
#include "query.h"
#include "tname.h"
#include "tcommon.h"
#include "taosdef.h"
#include "tarray.h"
#include "tcommon.h"
#include "thash.h"
#include "tmsg.h"
#include "tname.h"
#include "transport.h"
typedef struct SCatalog SCatalog;
@ -159,7 +159,8 @@ int32_t catalogGetDBVgVersion(SCatalog* pCtg, const char* dbFName, int32_t* vers
* @param pVgroupList (output, vgroup info list, element is SVgroupInfo, NEED to simply free the array by caller)
* @return error code
*/
int32_t catalogGetDBVgInfo(SCatalog* pCatalog, void *pTransporter, const SEpSet* pMgmtEps, const char* pDBName, SArray** pVgroupList);
int32_t catalogGetDBVgInfo(SCatalog* pCatalog, void* pTransporter, const SEpSet* pMgmtEps, const char* pDBName,
SArray** pVgroupList);
int32_t catalogUpdateDBVgInfo(SCatalog* pCatalog, const char* dbName, uint64_t dbId, SDBVgInfo* dbInfo);
@ -178,7 +179,8 @@ int32_t catalogRemoveStbMeta(SCatalog* pCtg, const char* dbFName, uint64_t dbId,
* @param pTableMeta(output, table meta data, NEED to free it by calller)
* @return error code
*/
int32_t catalogGetTableMeta(SCatalog* pCatalog, void * pTransporter, const SEpSet* pMgmtEps, const SName* pTableName, STableMeta** pTableMeta);
int32_t catalogGetTableMeta(SCatalog* pCatalog, void* pTransporter, const SEpSet* pMgmtEps, const SName* pTableName,
STableMeta** pTableMeta);
/**
* Get a super table's meta data.
@ -189,11 +191,11 @@ int32_t catalogGetTableMeta(SCatalog* pCatalog, void * pTransporter, const SEpSe
* @param pTableMeta(output, table meta data, NEED to free it by calller)
* @return error code
*/
int32_t catalogGetSTableMeta(SCatalog* pCatalog, void * pTransporter, const SEpSet* pMgmtEps, const SName* pTableName, STableMeta** pTableMeta);
int32_t catalogGetSTableMeta(SCatalog* pCatalog, void* pTransporter, const SEpSet* pMgmtEps, const SName* pTableName,
STableMeta** pTableMeta);
int32_t catalogUpdateTableMeta(SCatalog* pCatalog, STableMetaRsp* rspMsg);
/**
* Force refresh DB's local cached vgroup info.
* @param pCtg (input, got with catalogGetHandle)
@ -215,7 +217,8 @@ int32_t catalogChkTbMetaVersion(SCatalog* pCtg, void *pTrans, const SEpSet* pMgm
* @param isSTable (input, is super table or not, 1:supposed to be stable, 0: supposed not to be stable, -1:not sure)
* @return error code
*/
int32_t catalogRefreshTableMeta(SCatalog* pCatalog, void *pTransporter, const SEpSet* pMgmtEps, const SName* pTableName, int32_t isSTable);
int32_t catalogRefreshTableMeta(SCatalog* pCatalog, void* pTransporter, const SEpSet* pMgmtEps, const SName* pTableName,
int32_t isSTable);
/**
* Force refresh a table's local cached meta data and get the new one.
@ -227,9 +230,8 @@ int32_t catalogRefreshTableMeta(SCatalog* pCatalog, void *pTransporter, const SE
* @param isSTable (input, is super table or not, 1:supposed to be stable, 0: supposed not to be stable, -1:not sure)
* @return error code
*/
int32_t catalogRefreshGetTableMeta(SCatalog* pCatalog, void *pTransporter, const SEpSet* pMgmtEps, const SName* pTableName, STableMeta** pTableMeta, int32_t isSTable);
int32_t catalogRefreshGetTableMeta(SCatalog* pCatalog, void* pTransporter, const SEpSet* pMgmtEps,
const SName* pTableName, STableMeta** pTableMeta, int32_t isSTable);
/**
* Get a table's actual vgroup, for stable it's all possible vgroup list.
@ -240,7 +242,8 @@ int32_t catalogRefreshGetTableMeta(SCatalog* pCatalog, void *pTransporter, const
* @param pVgroupList (output, vgroup info list, element is SVgroupInfo, NEED to simply free the array by caller)
* @return error code
*/
int32_t catalogGetTableDistVgInfo(SCatalog* pCatalog, void *pTransporter, const SEpSet* pMgmtEps, const SName* pTableName, SArray** pVgroupList);
int32_t catalogGetTableDistVgInfo(SCatalog* pCatalog, void* pTransporter, const SEpSet* pMgmtEps,
const SName* pTableName, SArray** pVgroupList);
/**
* Get a table's vgroup from its name's hash value.
@ -251,8 +254,8 @@ int32_t catalogGetTableDistVgInfo(SCatalog* pCatalog, void *pTransporter, const
* @param vgInfo (output, vgroup info)
* @return error code
*/
int32_t catalogGetTableHashVgroup(SCatalog* pCatalog, void * pTransporter, const SEpSet* pMgmtEps, const SName* pName, SVgroupInfo* vgInfo);
int32_t catalogGetTableHashVgroup(SCatalog* pCatalog, void* pTransporter, const SEpSet* pMgmtEps, const SName* pName,
SVgroupInfo* vgInfo);
/**
* Get all meta data required in pReq.
@ -263,9 +266,11 @@ int32_t catalogGetTableHashVgroup(SCatalog* pCatalog, void * pTransporter, const
* @param pRsp (output, response data)
* @return error code
*/
int32_t catalogGetAllMeta(SCatalog* pCatalog, void *pTransporter, const SEpSet* pMgmtEps, const SCatalogReq* pReq, SMetaData* pRsp);
int32_t catalogGetAllMeta(SCatalog* pCatalog, void* pTransporter, const SEpSet* pMgmtEps, const SCatalogReq* pReq,
SMetaData* pRsp);
int32_t catalogAsyncGetAllMeta(SCatalog* pCtg, void *pTrans, const SEpSet* pMgmtEps, uint64_t reqId, const SCatalogReq* pReq, catalogCallback fp, void* param, int64_t* jobId);
int32_t catalogAsyncGetAllMeta(SCatalog* pCtg, void* pTrans, const SEpSet* pMgmtEps, uint64_t reqId,
const SCatalogReq* pReq, catalogCallback fp, void* param, int64_t* jobId);
int32_t catalogGetQnodeList(SCatalog* pCatalog, void* pTransporter, const SEpSet* pMgmtEps, SArray* pQnodeList);
@ -277,13 +282,16 @@ int32_t catalogGetExpiredUsers(SCatalog* pCtg, SUserAuthVersion **users, uint32_
int32_t catalogGetDBCfg(SCatalog* pCtg, void* pRpc, const SEpSet* pMgmtEps, const char* dbFName, SDbCfgInfo* pDbCfg);
int32_t catalogGetIndexMeta(SCatalog* pCtg, void *pRpc, const SEpSet* pMgmtEps, const char* indexName, SIndexInfo* pInfo);
int32_t catalogGetIndexMeta(SCatalog* pCtg, void* pRpc, const SEpSet* pMgmtEps, const char* indexName,
SIndexInfo* pInfo);
int32_t catalogGetTableIndex(SCatalog* pCtg, void *pTrans, const SEpSet* pMgmtEps, const SName* pTableName, SArray** pRes);
int32_t catalogGetTableIndex(SCatalog* pCtg, void* pTrans, const SEpSet* pMgmtEps, const SName* pTableName,
SArray** pRes);
int32_t catalogGetUdfInfo(SCatalog* pCtg, void* pRpc, const SEpSet* pMgmtEps, const char* funcName, SFuncInfo* pInfo);
int32_t catalogChkAuth(SCatalog* pCtg, void *pRpc, const SEpSet* pMgmtEps, const char* user, const char* dbFName, AUTH_TYPE type, bool *pass);
int32_t catalogChkAuth(SCatalog* pCtg, void* pRpc, const SEpSet* pMgmtEps, const char* user, const char* dbFName,
AUTH_TYPE type, bool* pass);
int32_t catalogUpdateUserAuthInfo(SCatalog* pCtg, SGetUserAuthRsp* pAuth);
@ -291,7 +299,6 @@ int32_t catalogUpdateVgEpSet(SCatalog* pCtg, const char* dbFName, int32_t vgId,
int32_t ctgdLaunchAsyncCall(SCatalog* pCtg, void* pTrans, const SEpSet* pMgmtEps, uint64_t reqId, bool forceUpdate);
/**
* Destroy catalog and relase all resources
*/

2
include/libs/nodes/plannodes.h
View File

@ -62,6 +62,7 @@ typedef struct SScanLogicNode {
int64_t watermark;
int16_t tsColId;
double filesFactor;
SArray* pSmaIndexes;
} SScanLogicNode;
typedef struct SJoinLogicNode {
@ -437,7 +438,6 @@ typedef struct SQueryPlan {
int32_t numOfSubplans;
SNodeList* pSubplans; // Element is SNodeListNode. The execution level of subplan, starting from 0.
SExplainInfo explainInfo;
SArray* pPlaceholderValues;
} SQueryPlan;
void nodesWalkPhysiPlan(SNode* pNode, FNodeWalker walker, void* pContext);

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

@ -144,6 +144,7 @@ typedef struct SRealTableNode {
SVgroupsInfo* pVgroupList;
char qualDbName[TSDB_DB_NAME_LEN]; // SHOW qualDbName.TABLES
double ratio;
SArray* pSmaIndexes;
} SRealTableNode;
typedef struct STempTableNode {

2
include/libs/parser/parser.h
View File

@ -63,7 +63,7 @@ int32_t qAnalyseSqlSemantic(SParseContext* pCxt, const struct SCatalogReq* pCata
void qDestroyQuery(SQuery* pQueryNode);
int32_t qExtractResultSchema(const SNode* pRoot, int32_t* numOfCols, SSchema** pSchema);
int32_t qSetSTableIdForRSma(SNode* pStmt, int64_t uid);
int32_t qSetSTableIdForRsma(SNode* pStmt, int64_t uid);
int32_t qBuildStmtOutput(SQuery* pQuery, SHashObj* pVgHash, SHashObj* pBlockHash);
int32_t qResetStmtDataBlock(void* block, bool keepBuf);

9
include/libs/sync/sync.h
View File

@ -83,8 +83,10 @@ typedef struct SReConfigCbMeta {
SyncTerm term;
SyncTerm currentTerm;
SSyncCfg oldCfg;
SSyncCfg newCfg;
bool isDrop;
uint64_t flag;
uint64_t seqNum;
} SReConfigCbMeta;
typedef struct SSnapshot {
@ -106,7 +108,7 @@ typedef struct SSyncFSM {
void (*FpRollBackCb)(struct SSyncFSM* pFsm, const SRpcMsg* pMsg, SFsmCbMeta cbMeta);
void (*FpRestoreFinishCb)(struct SSyncFSM* pFsm);
void (*FpReConfigCb)(struct SSyncFSM* pFsm, SSyncCfg newCfg, SReConfigCbMeta cbMeta);
void (*FpReConfigCb)(struct SSyncFSM* pFsm, const SRpcMsg* pMsg, SReConfigCbMeta cbMeta);
int32_t (*FpGetSnapshot)(struct SSyncFSM* pFsm, SSnapshot* pSnapshot);
@ -184,7 +186,6 @@ int64_t syncOpen(const SSyncInfo* pSyncInfo);
void syncStart(int64_t rid);
void syncStop(int64_t rid);
int32_t syncSetStandby(int64_t rid);
int32_t syncReconfig(int64_t rid, const SSyncCfg* pSyncCfg);
ESyncState syncGetMyRole(int64_t rid);
const char* syncGetMyRoleStr(int64_t rid);
SyncTerm syncGetMyTerm(int64_t rid);
@ -194,9 +195,11 @@ 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 syncGetSnapshotMeta(int64_t rid, struct SSnapshotMeta* sMeta);
int32_t syncReconfig(int64_t rid, const SSyncCfg* pNewCfg);
int32_t syncReconfigRaw(int64_t rid, const SSyncCfg* pNewCfg, SRpcMsg* pRpcMsg);
// to be moved to static
void syncStartNormal(int64_t rid);
void syncStartStandBy(int64_t rid);

3
include/util/taoserror.h
View File

@ -85,7 +85,6 @@ int32_t* taosGetErrno();
#define TSDB_CODE_RPC_NETWORK_UNAVAIL TAOS_DEF_ERROR_CODE(0, 0x0102)
#define TSDB_CODE_RPC_FQDN_ERROR TAOS_DEF_ERROR_CODE(0, 0x0103)
#define TSDB_CODE_RPC_PORT_EADDRINUSE TAOS_DEF_ERROR_CODE(0, 0x0104)
#define TSDB_CODE_RPC_INDIRECT_NETWORK_UNAVAIL TAOS_DEF_ERROR_CODE(0, 0x0105)
//client
#define TSDB_CODE_TSC_INVALID_OPERATION TAOS_DEF_ERROR_CODE(0, 0x0200)
@ -220,6 +219,7 @@ int32_t* taosGetErrno();
#define TSDB_CODE_MND_VGROUP_ALREADY_IN_DNODE TAOS_DEF_ERROR_CODE(0, 0x0392)
#define TSDB_CODE_MND_VGROUP_UN_CHANGED TAOS_DEF_ERROR_CODE(0, 0x0393)
#define TSDB_CODE_MND_HAS_OFFLINE_DNODE TAOS_DEF_ERROR_CODE(0, 0x0394)
#define TSDB_CODE_MND_INVALID_REPLICA TAOS_DEF_ERROR_CODE(0, 0x0395)
// mnode-stable
#define TSDB_CODE_MND_STB_ALREADY_EXIST TAOS_DEF_ERROR_CODE(0, 0x03A0)
@ -260,6 +260,7 @@ int32_t* taosGetErrno();
#define TSDB_CODE_MND_TRANS_CONFLICT TAOS_DEF_ERROR_CODE(0, 0x03D3)
#define TSDB_CODE_MND_TRANS_UNKNOW_ERROR TAOS_DEF_ERROR_CODE(0, 0x03D4)
#define TSDB_CODE_MND_TRANS_CLOG_IS_NULL TAOS_DEF_ERROR_CODE(0, 0x03D5)
#define TSDB_CODE_MND_TRANS_NETWORK_UNAVAILL TAOS_DEF_ERROR_CODE(0, 0x03D6)
// mnode-mq
#define TSDB_CODE_MND_TOPIC_ALREADY_EXIST TAOS_DEF_ERROR_CODE(0, 0x03E0)

3
include/util/tjson.h
View File

@ -17,6 +17,7 @@
#define _TD_UTIL_JSON_H_
#include "os.h"
#include "tarray.h"
#ifdef __cplusplus
extern "C" {
@ -66,12 +67,14 @@ typedef int32_t (*FToJson)(const void* pObj, SJson* pJson);
int32_t tjsonAddObject(SJson* pJson, const char* pName, FToJson func, const void* pObj);
int32_t tjsonAddItem(SJson* pJson, FToJson func, const void* pObj);
int32_t tjsonAddArray(SJson* pJson, const char* pName, FToJson func, const void* pArray, int32_t itemSize, int32_t num);
int32_t tjsonAddTArray(SJson* pJson, const char* pName, FToJson func, const SArray* pArray);
typedef int32_t (*FToObject)(const SJson* pJson, void* pObj);
int32_t tjsonToObject(const SJson* pJson, const char* pName, FToObject func, void* pObj);
int32_t tjsonMakeObject(const SJson* pJson, const char* pName, FToObject func, void** pObj, int32_t objSize);
int32_t tjsonToArray(const SJson* pJson, const char* pName, FToObject func, void* pArray, int32_t itemSize);
int32_t tjsonToTArray(const SJson* pJson, const char* pName, FToObject func, SArray** pArray, int32_t itemSize);
char* tjsonToString(const SJson* pJson);
char* tjsonToUnformattedString(const SJson* pJson);

1
source/client/src/clientImpl.c
View File

@ -692,6 +692,7 @@ void launchAsyncQuery(SRequestObj* pRequest, SQuery* pQuery) {
if (TSDB_CODE_SUCCESS == code) {
code = qCreateQueryPlan(&cxt, &pRequest->body.pDag, pNodeList);
tscError("0x%"PRIx64" failed to create query plan, code:%s 0x%"PRIx64, pRequest->self, tstrerror(code), pRequest->requestId);
}
if (TSDB_CODE_SUCCESS == code) {

207
source/client/src/clientSml.c
View File

@ -17,6 +17,9 @@
#include "tname.h"
#include "cJSON.h"
#include "tglobal.h"
#include "osSemaphore.h"
#include "osThread.h"
//=================================================================================================
#define SPACE ' '
@ -67,6 +70,9 @@ for (int i = 1; i < keyLen; ++i) { \
#define BINARY_ADD_LEN 2 // "binary" 2 means " "
#define NCHAR_ADD_LEN 3 // L"nchar" 3 means L" "
#define MAX_RETRY_TIMES 5
#define LINE_BATCH 20
//=================================================================================================
typedef TSDB_SML_PROTOCOL_TYPE SMLProtocolType;
@ -153,8 +159,17 @@ typedef struct {
int64_t endTime;
} SSmlCostInfo;
typedef struct{
SRequestObj* request;
SCatalog* catalog;
tsem_t sem;
TdThreadSpinlock lock;
} Params;
typedef struct {
int64_t id;
Params *params;
bool isLast;
SMLProtocolType protocol;
int8_t precision;
@ -303,7 +318,7 @@ static int32_t smlApplySchemaAction(SSmlHandle* info, SSchemaAction* action) {
uError("SML:0x%" PRIx64 " apply schema action. reset query cache. error: %s", info->id, taos_errstr(res2));
}
taos_free_result(res2);
taosMsleep(10);
taosMsleep(500);
}
break;
}
@ -327,7 +342,7 @@ static int32_t smlApplySchemaAction(SSmlHandle* info, SSchemaAction* action) {
uError("SML:0x%" PRIx64 " apply schema action. reset query cache. error: %s", info->id, taos_errstr(res2));
}
taos_free_result(res2);
taosMsleep(10);
taosMsleep(500);
}
break;
}
@ -350,7 +365,7 @@ static int32_t smlApplySchemaAction(SSmlHandle* info, SSchemaAction* action) {
uError("SML:0x%" PRIx64 " apply schema action. reset query cache. error: %s", info->id, taos_errstr(res2));
}
taos_free_result(res2);
taosMsleep(10);
taosMsleep(500);
}
break;
}
@ -373,7 +388,7 @@ static int32_t smlApplySchemaAction(SSmlHandle* info, SSchemaAction* action) {
uError("SML:0x%" PRIx64 " apply schema action. reset query cache. error: %s", info->id, taos_errstr(res2));
}
taos_free_result(res2);
taosMsleep(10);
taosMsleep(500);
}
break;
}
@ -424,7 +439,7 @@ static int32_t smlApplySchemaAction(SSmlHandle* info, SSchemaAction* action) {
uError("SML:0x%" PRIx64 " apply schema action. reset query cache. error: %s", info->id, taos_errstr(res2));
}
taos_free_result(res2);
taosMsleep(10);
taosMsleep(500);
}
break;
}
@ -541,56 +556,6 @@ end:
return code;
}
//=========================================================================
/* Field Escape charaters
1: measurement Comma,Space
2: tag_key, tag_value, field_key Comma,Equal Sign,Space
3: field_value Double quote,Backslash
*/
//static void escapeSpecialCharacter(uint8_t field, const char **pos) {
// const char *cur = *pos;
// if (*cur != '\\') {
// return;
// }
// switch (field) {
// case 1:
// switch (*(cur + 1)) {
// case ',':
// case ' ':
// cur++;
// break;
// default:
// break;
// }
// break;
// case 2:
// switch (*(cur + 1)) {
// case ',':
// case ' ':
// case '=':
// cur++;
// break;
// default:
// break;
// }
// break;
// case 3:
// switch (*(cur + 1)) {
// case '"':
// case '\\':
// cur++;
// break;
// default:
// break;
// }
// break;
// default:
// break;
// }
// *pos = cur;
//}
static bool smlParseNumber(SSmlKv *kvVal, SSmlMsgBuf *msg){
const char *pVal = kvVal->value;
int32_t len = kvVal->length;
@ -1426,6 +1391,7 @@ static void smlDestroyInfo(SSmlHandle* info){
if(!info->dataFormat){
taosArrayDestroy(info->colsContainer);
}
destroyRequest(info->pRequest);
taosMemoryFreeClear(info);
}
@ -1453,11 +1419,6 @@ static SSmlHandle* smlBuildSmlInfo(TAOS* taos, SRequestObj* request, SMLProtocol
((SVnodeModifOpStmt*)(info->pQuery->pRoot))->payloadType = PAYLOAD_TYPE_KV;
info->taos = (STscObj *)taos;
code = catalogGetHandle(info->taos->pAppInfo->clusterId, &info->pCatalog);
if(code != TSDB_CODE_SUCCESS){
uError("SML:0x%"PRIx64" get catalog error %d", info->id, code);
goto cleanup;
}
info->precision = precision;
info->protocol = protocol;
@ -2206,7 +2167,6 @@ end:
return ret;
}
static int32_t smlInsertData(SSmlHandle* info) {
int32_t code = TSDB_CODE_SUCCESS;
@ -2248,10 +2208,12 @@ static int32_t smlInsertData(SSmlHandle* info) {
}
info->cost.insertRpcTime = taosGetTimestampUs();
launchQueryImpl(info->pRequest, info->pQuery, true, NULL);
//launchQueryImpl(info->pRequest, info->pQuery, false, NULL);
// info->affectedRows = taos_affected_rows(info->pRequest);
// return info->pRequest->code;
info->affectedRows = taos_affected_rows(info->pRequest);
return info->pRequest->code;
launchAsyncQuery(info->pRequest, info->pQuery);
return TSDB_CODE_SUCCESS;
}
static void smlPrintStatisticInfo(SSmlHandle *info){
@ -2311,7 +2273,7 @@ static int smlProcess(SSmlHandle *info, char* lines[], int numLines) {
do{
code = smlModifyDBSchemas(info);
if (code == 0) break;
} while (retryNum++ < taosHashGetSize(info->superTables));
} while (retryNum++ < taosHashGetSize(info->superTables) * MAX_RETRY_TIMES);
if (code != 0) {
uError("SML:0x%"PRIx64" smlModifyDBSchemas error : %s", info->id, tstrerror(code));
@ -2332,30 +2294,48 @@ cleanup:
return code;
}
static int32_t isSchemalessDb(SSmlHandle* info){
static int32_t isSchemalessDb(STscObj *taos, SCatalog *catalog){
SName name;
tNameSetDbName(&name, info->taos->acctId, info->taos->db, strlen(info->taos->db));
tNameSetDbName(&name, taos->acctId, taos->db, strlen(taos->db));
char dbFname[TSDB_DB_FNAME_LEN] = {0};
tNameGetFullDbName(&name, dbFname);
SDbCfgInfo pInfo = {0};
SEpSet ep = getEpSet_s(&info->taos->pAppInfo->mgmtEp);
SEpSet ep = getEpSet_s(&taos->pAppInfo->mgmtEp);
int32_t code = catalogGetDBCfg(info->pCatalog, info->taos->pAppInfo->pTransporter, &ep, dbFname, &pInfo);
int32_t code = catalogGetDBCfg(catalog, taos->pAppInfo->pTransporter, &ep, dbFname, &pInfo);
if (code != TSDB_CODE_SUCCESS) {
info->pRequest->code = code;
smlBuildInvalidDataMsg(&info->msgBuf, "catalogGetDBCfg error, code:", tstrerror(code));
return code;
}
taosArrayDestroy(pInfo.pRetensions);
if (!pInfo.schemaless){
info->pRequest->code = TSDB_CODE_SML_INVALID_DB_CONF;
smlBuildInvalidDataMsg(&info->msgBuf, "can not insert into schemaless db:", dbFname);
return TSDB_CODE_SML_INVALID_DB_CONF;
}
return TSDB_CODE_SUCCESS;
}
static void smlInsertCallback(void* param, void* res, int32_t code) {
SRequestObj *pRequest = (SRequestObj *)res;
SSmlHandle* info = (SSmlHandle *)param;
// lock
if(code != TSDB_CODE_SUCCESS){
taosThreadSpinLock(&info->params->lock);
info->params->request->code = code;
taosThreadSpinUnlock(&info->params->lock);
}
// unlock
printf("SML:0x%" PRIx64 " insert finished, code: %d, total: %d\n", info->id, code, info->affectedRows);
Params *pParam = info->params;
bool isLast = info->isLast;
smlDestroyInfo(info);
if(isLast){
tsem_post(&pParam->sem);
}
}
/**
* taos_schemaless_insert() parse and insert data points into database according to
* different protocol.
@ -2384,48 +2364,95 @@ TAOS_RES* taos_schemaless_insert(TAOS* taos, char* lines[], int numLines, int pr
return NULL;
}
SSmlHandle* info = smlBuildSmlInfo(taos, request, (SMLProtocolType)protocol, precision);
if(!info){
return (TAOS_RES*)request;
}
((STscObj *)taos)->schemalessType = 1;
SSmlMsgBuf msg = {ERROR_MSG_BUF_DEFAULT_SIZE, request->msgBuf};
info->taos->schemalessType = 1;
if(request->pDb == NULL){
request->code = TSDB_CODE_PAR_DB_NOT_SPECIFIED;
smlBuildInvalidDataMsg(&info->msgBuf, "Database not specified", NULL);
int cnt = ceil(((double)numLines)/LINE_BATCH);
Params params;
params.request = request;
tsem_init(&params.sem, 0, 0);
taosThreadSpinInit(&(params.lock), 0);
int32_t code = catalogGetHandle(((STscObj *)taos)->pAppInfo->clusterId, &params.catalog);
if(code != TSDB_CODE_SUCCESS){
uError("SML get catalog error %d", code);
request->code = code;
goto end;
}
if(isSchemalessDb(info) != TSDB_CODE_SUCCESS){
if(request->pDb == NULL){
request->code = TSDB_CODE_PAR_DB_NOT_SPECIFIED;
smlBuildInvalidDataMsg(&msg, "Database not specified", NULL);
goto end;
}
if(isSchemalessDb(((STscObj *)taos), params.catalog) != TSDB_CODE_SUCCESS){
request->code = TSDB_CODE_SML_INVALID_DB_CONF;
smlBuildInvalidDataMsg(&info->msgBuf, "Cannot write data to a non schemaless database", NULL);
smlBuildInvalidDataMsg(&msg, "Cannot write data to a non schemaless database", NULL);
goto end;
}
if (!lines) {
request->code = TSDB_CODE_SML_INVALID_DATA;
smlBuildInvalidDataMsg(&info->msgBuf, "lines is null", NULL);
smlBuildInvalidDataMsg(&msg, "lines is null", NULL);
goto end;
}
if(protocol < TSDB_SML_LINE_PROTOCOL || protocol > TSDB_SML_JSON_PROTOCOL){
request->code = TSDB_CODE_SML_INVALID_PROTOCOL_TYPE;
smlBuildInvalidDataMsg(&info->msgBuf, "protocol invalidate", NULL);
smlBuildInvalidDataMsg(&msg, "protocol invalidate", NULL);
goto end;
}
if(protocol == TSDB_SML_LINE_PROTOCOL && (precision < TSDB_SML_TIMESTAMP_NOT_CONFIGURED || precision > TSDB_SML_TIMESTAMP_NANO_SECONDS)){
request->code = TSDB_CODE_SML_INVALID_PRECISION_TYPE;
smlBuildInvalidDataMsg(&info->msgBuf, "precision invalidate for line protocol", NULL);
smlBuildInvalidDataMsg(&msg, "precision invalidate for line protocol", NULL);
goto end;
}
info->pRequest->code = smlProcess(info, lines, numLines);
for (int i = 0; i < cnt; ++i) {
SRequestObj* req = (SRequestObj*)createRequest((STscObj *)taos, TSDB_SQL_INSERT);
if(!req){
request->code = TSDB_CODE_OUT_OF_MEMORY;
uError("SML:taos_schemaless_insert error request is null");
goto end;
}
SSmlHandle* info = smlBuildSmlInfo(taos, req, (SMLProtocolType)protocol, precision);
if(!info){
request->code = TSDB_CODE_OUT_OF_MEMORY;
uError("SML:taos_schemaless_insert error SSmlHandle is null");
goto end;
}
int32_t perBatch = LINE_BATCH;
if(numLines > perBatch){
numLines -= perBatch;
info->isLast = false;
}else{
perBatch = numLines;
numLines = 0;
info->isLast = true;
}
info->params = &params;
info->pCatalog = params.catalog;
info->affectedRows = perBatch;
info->pRequest->body.queryFp = smlInsertCallback;
info->pRequest->body.param = info;
code = smlProcess(info, lines, perBatch);
lines += perBatch;
if (code != TSDB_CODE_SUCCESS){
info->pRequest->body.queryFp(info, req, code);
}
}
tsem_wait(&params.sem);
end:
info->taos->schemalessType = 0;
uDebug("result:%s", info->msgBuf.buf);
smlDestroyInfo(info);
taosThreadSpinDestroy(&params.lock);
tsem_destroy(&params.sem);
((STscObj *)taos)->schemalessType = 0;
uDebug("result:%s", request->msgBuf);
return (TAOS_RES*)request;
}

2
source/client/src/tmq.c
View File

@ -1281,7 +1281,7 @@ TAOS_RES* tmq_consumer_poll(tmq_t* tmq, int64_t timeout) {
if (rspObj) {
return (TAOS_RES*)rspObj;
}
if (timeout != 0) {
if (timeout != -1) {
int64_t endTime = taosGetTimestampMs();
int64_t leftTime = endTime - startTime;
if (leftTime > timeout) {

2
source/client/test/smlTest.cpp
View File

@ -1325,7 +1325,7 @@ TEST(testCase, sml_oom_Test) {
pRes = taos_query(taos, "use oom");
taos_free_result(pRes);
TAOS_RES* res = taos_schemaless_insert(taos, (char**)sql, 100, TSDB_SML_LINE_PROTOCOL, 0);
TAOS_RES* res = taos_schemaless_insert(taos, (char**)sql, sizeof(sql)/sizeof(sql[0]), TSDB_SML_LINE_PROTOCOL, 0);
ASSERT_EQ(taos_errno(res), 0);
taos_free_result(pRes);
}

2
source/common/src/systable.c
View File

@ -221,7 +221,7 @@ static const SSysDbTableSchema transSchema[] = {
{.name = "db", .bytes = SYSTABLE_SCH_DB_NAME_LEN, .type = TSDB_DATA_TYPE_VARCHAR},
{.name = "failed_times", .bytes = 4, .type = TSDB_DATA_TYPE_INT},
{.name = "last_exec_time", .bytes = 8, .type = TSDB_DATA_TYPE_TIMESTAMP},
{.name = "last_error", .bytes = (TSDB_TRANS_ERROR_LEN - 1) + VARSTR_HEADER_SIZE, .type = TSDB_DATA_TYPE_VARCHAR},
{.name = "last_action_info", .bytes = (TSDB_TRANS_ERROR_LEN - 1) + VARSTR_HEADER_SIZE, .type = TSDB_DATA_TYPE_VARCHAR},
};
static const SSysDbTableSchema configSchema[] = {

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

@ -294,7 +294,7 @@ int32_t colDataMergeCol(SColumnInfoData* pColumnInfoData, uint32_t numOfRow1, in
int32_t colDataAssign(SColumnInfoData* pColumnInfoData, const SColumnInfoData* pSource, int32_t numOfRows) {
ASSERT(pColumnInfoData != NULL && pSource != NULL && pColumnInfoData->info.type == pSource->info.type);
if (numOfRows == 0) {
if (numOfRows <= 0) {
return numOfRows;
}
@ -322,6 +322,7 @@ int32_t colDataAssign(SColumnInfoData* pColumnInfoData, const SColumnInfoData* p
pColumnInfoData->varmeta.length = pSource->varmeta.length;
} else {
char* tmp = taosMemoryRealloc(pColumnInfoData->nullbitmap, BitmapLen(numOfRows));
printf("----------------%d\n", BitmapLen(numOfRows));
if (tmp == NULL) {
return TSDB_CODE_OUT_OF_MEMORY;
}
@ -1219,6 +1220,8 @@ SSDataBlock* createOneDataBlock(const SSDataBlock* pDataBlock, bool copyData) {
pBlock->info.hasVarCol = pDataBlock->info.hasVarCol;
pBlock->info.rowSize = pDataBlock->info.rowSize;
pBlock->info.groupId = pDataBlock->info.groupId;
pBlock->info.childId = pDataBlock->info.childId;
pBlock->info.type = pDataBlock->info.type;
for (int32_t i = 0; i < numOfCols; ++i) {
SColumnInfoData colInfo = {0};
@ -1237,6 +1240,9 @@ SSDataBlock* createOneDataBlock(const SSDataBlock* pDataBlock, bool copyData) {
return NULL;
}
if (pSrc->pData == NULL) {
continue;
}
colDataAssign(pDst, pSrc, pDataBlock->info.rows);
}
@ -1499,6 +1505,7 @@ void blockDebugShowData(const SArray* dataBlocks, const char* flag) {
SSDataBlock* pDataBlock = taosArrayGet(dataBlocks, i);
int32_t colNum = pDataBlock->info.numOfCols;
int32_t rows = pDataBlock->info.rows;
printf("%s |block type %d |child id %d|\n", flag, (int32_t)pDataBlock->info.type, pDataBlock->info.childId);
for (int32_t j = 0; j < rows; j++) {
printf("%s |", flag);
for (int32_t k = 0; k < colNum; k++) {

6
source/dnode/mgmt/mgmt_dnode/src/dmHandle.c
View File

@ -38,10 +38,14 @@ static void dmProcessStatusRsp(SDnodeMgmt *pMgmt, SRpcMsg *pRsp) {
SStatusRsp statusRsp = {0};
if (pRsp->pCont != NULL && pRsp->contLen > 0 &&
tDeserializeSStatusRsp(pRsp->pCont, pRsp->contLen, &statusRsp) == 0) {
dTrace("status msg received from mnode, dnodeVer:%" PRId64 " saved:%" PRId64, statusRsp.dnodeVer,
pMgmt->pData->dnodeVer);
if (pMgmt->pData->dnodeVer != statusRsp.dnodeVer) {
pMgmt->pData->dnodeVer = statusRsp.dnodeVer;
dmUpdateDnodeCfg(pMgmt, &statusRsp.dnodeCfg);
dmUpdateEps(pMgmt->pData, statusRsp.pDnodeEps);
}
}
rpcFreeCont(pRsp->pCont);
tFreeSStatusRsp(&statusRsp);
}
@ -89,7 +93,7 @@ void dmSendStatusReq(SDnodeMgmt *pMgmt) {
SRpcMsg rpcMsg = {.pCont = pHead, .contLen = contLen, .msgType = TDMT_MND_STATUS, .info.ahandle = (void *)0x9527};
SRpcMsg rpcRsp = {0};
dTrace("send status msg to mnode");
dTrace("send status msg to mnode, dnodeVer:%" PRId64, req.dnodeVer);
SEpSet epSet = {0};
dmGetMnodeEpSet(pMgmt->pData, &epSet);

3
source/dnode/mgmt/mgmt_mnode/src/mmHandle.c
View File

@ -170,6 +170,9 @@ SArray *mmGetMsgHandles() {
if (dmSetMgmtHandle(pArray, TDMT_MND_COMPACT_DB, mmPutNodeMsgToWriteQueue, 0) == NULL) goto _OVER;
if (dmSetMgmtHandle(pArray, TDMT_MND_GET_DB_CFG, mmPutNodeMsgToReadQueue, 0) == NULL) goto _OVER;
if (dmSetMgmtHandle(pArray, TDMT_MND_VGROUP_LIST, mmPutNodeMsgToReadQueue, 0) == NULL) goto _OVER;
if (dmSetMgmtHandle(pArray, TDMT_MND_REDISTRIBUTE_VGROUP, mmPutNodeMsgToWriteQueue, 0) == NULL) goto _OVER;
if (dmSetMgmtHandle(pArray, TDMT_MND_MERGE_VGROUP, mmPutNodeMsgToWriteQueue, 0) == NULL) goto _OVER;
if (dmSetMgmtHandle(pArray, TDMT_MND_BALANCE_VGROUP, mmPutNodeMsgToWriteQueue, 0) == NULL) goto _OVER;
if (dmSetMgmtHandle(pArray, TDMT_MND_CREATE_FUNC, mmPutNodeMsgToWriteQueue, 0) == NULL) goto _OVER;
if (dmSetMgmtHandle(pArray, TDMT_MND_RETRIEVE_FUNC, mmPutNodeMsgToWriteQueue, 0) == NULL) goto _OVER;
if (dmSetMgmtHandle(pArray, TDMT_MND_DROP_FUNC, mmPutNodeMsgToWriteQueue, 0) == NULL) goto _OVER;

27
source/dnode/mgmt/node_util/src/dmEps.c
View File

@ -81,6 +81,13 @@ int32_t dmReadEps(SDnodeData *pData) {
}
pData->dnodeId = dnodeId->valueint;
cJSON *dnodeVer = cJSON_GetObjectItem(root, "dnodeVer");
if (!dnodeVer || dnodeVer->type != cJSON_String) {
dError("failed to read %s since dnodeVer not found", file);
goto _OVER;
}
pData->dnodeVer = atoll(dnodeVer->valuestring);
cJSON *clusterId = cJSON_GetObjectItem(root, "clusterId");
if (!clusterId || clusterId->type != cJSON_String) {
dError("failed to read %s since clusterId not found", file);
@ -193,6 +200,7 @@ int32_t dmWriteEps(SDnodeData *pData) {
len += snprintf(content + len, maxLen - len, "{\n");
len += snprintf(content + len, maxLen - len, " \"dnodeId\": %d,\n", pData->dnodeId);
len += snprintf(content + len, maxLen - len, " \"dnodeVer\": \"%" PRId64 "\",\n", pData->dnodeVer);
len += snprintf(content + len, maxLen - len, " \"clusterId\": \"%" PRId64 "\",\n", pData->clusterId);
len += snprintf(content + len, maxLen - len, " \"dropped\": %d,\n", pData->dropped);
len += snprintf(content + len, maxLen - len, " \"dnodes\": [{\n");
@ -224,30 +232,15 @@ int32_t dmWriteEps(SDnodeData *pData) {
}
pData->updateTime = taosGetTimestampMs();
dDebug("successed to write %s", realfile);
dDebug("successed to write %s, dnodeVer:%" PRId64, realfile, pData->dnodeVer);
return 0;
}
void dmUpdateEps(SDnodeData *pData, SArray *eps) {
int32_t numOfEps = taosArrayGetSize(eps);
if (numOfEps <= 0) return;
taosThreadRwlockWrlock(&pData->lock);
int32_t numOfEpsOld = (int32_t)taosArrayGetSize(pData->dnodeEps);
if (numOfEps != numOfEpsOld) {
dDebug("new dnode list get from mnode");
dDebug("new dnode list get from mnode, dnodeVer:%" PRId64, pData->dnodeVer);
dmResetEps(pData, eps);
dmWriteEps(pData);
} else {
int32_t size = numOfEps * sizeof(SDnodeEp);
if (memcmp(pData->dnodeEps->pData, eps->pData, size) != 0) {
dDebug("new dnode list get from mnode");
dmResetEps(pData, eps);
dmWriteEps(pData);
}
}
taosThreadRwlockUnlock(&pData->lock);
}

7
source/dnode/mnode/impl/inc/mndDef.h
View File

@ -120,10 +120,10 @@ typedef struct {
SArray* commitActions;
int64_t createdTime;
int64_t lastExecTime;
int32_t lastErrorAction;
int32_t lastAction;
int32_t lastErrorNo;
tmsg_t lastErrorMsgType;
SEpSet lastErrorEpset;
tmsg_t lastMsgType;
SEpSet lastEpset;
char dbname[TSDB_DB_FNAME_LEN];
int32_t startFunc;
int32_t stopFunc;
@ -485,6 +485,7 @@ typedef struct {
int64_t stbUid;
SHashObj* consumerHash; // consumerId -> SMqConsumerEp
SArray* unassignedVgs; // SArray<SMqVgEp*>
char dbName[TSDB_DB_FNAME_LEN];
} SMqSubscribeObj;
SMqSubscribeObj* tNewSubscribeObj(const char key[TSDB_SUBSCRIBE_KEY_LEN]);

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

@ -29,8 +29,8 @@ int32_t mndSchedInitSubEp(SMnode* pMnode, const SMqTopicObj* pTopic, SMqSubscrib
int32_t mndScheduleStream(SMnode* pMnode, STrans* pTrans, SStreamObj* pStream);
int32_t mndConvertRSmaTask(const char* ast, int64_t uid, int8_t triggerType, int64_t watermark, char** pStr,
int32_t* pLen, double filesFactor);
int32_t mndConvertRsmaTask(char** pDst, int32_t* pDstLen, const char* ast, int64_t uid, int8_t triggerType,
int64_t watermark, double filesFactor);
#ifdef __cplusplus
}

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

@ -29,6 +29,7 @@ void mndReleaseVgroup(SMnode *pMnode, SVgObj *pVgroup);
SSdbRaw *mndVgroupActionEncode(SVgObj *pVgroup);
SEpSet mndGetVgroupEpset(SMnode *pMnode, const SVgObj *pVgroup);
int32_t mndGetVnodesNum(SMnode *pMnode, int32_t dnodeId);
void mndSortVnodeGid(SVgObj *pVgroup);
SArray *mndBuildDnodesArray(SMnode *, int32_t exceptDnodeId);
int32_t mndAllocSmaVgroup(SMnode *, SDbObj *pDb, SVgObj *pVgroup);

31
source/dnode/mnode/impl/src/mndConsumer.c
View File

@ -35,7 +35,7 @@
#define MND_CONSUMER_LOST_HB_CNT 3
static int8_t mqRebLock = 0;
static int8_t mqRebInExecCnt = 0;
static const char *mndConsumerStatusName(int status);
@ -76,15 +76,15 @@ int32_t mndInitConsumer(SMnode *pMnode) {
void mndCleanupConsumer(SMnode *pMnode) {}
bool mndRebTryStart() {
int8_t old = atomic_val_compare_exchange_8(&mqRebLock, 0, 1);
int8_t old = atomic_val_compare_exchange_8(&mqRebInExecCnt, 0, 1);
return old == 0;
}
void mndRebEnd() { atomic_sub_fetch_8(&mqRebLock, 1); }
void mndRebEnd() { atomic_sub_fetch_8(&mqRebInExecCnt, 1); }
void mndRebCntInc() { atomic_add_fetch_8(&mqRebLock, 1); }
void mndRebCntInc() { atomic_add_fetch_8(&mqRebInExecCnt, 1); }
void mndRebCntDec() { atomic_sub_fetch_8(&mqRebLock, 1); }
void mndRebCntDec() { atomic_sub_fetch_8(&mqRebInExecCnt, 1); }
static int32_t mndProcessConsumerLostMsg(SRpcMsg *pMsg) {
SMnode *pMnode = pMsg->info.node;
@ -92,7 +92,6 @@ static int32_t mndProcessConsumerLostMsg(SRpcMsg *pMsg) {
SMqConsumerObj *pConsumer = mndAcquireConsumer(pMnode, pLostMsg->consumerId);
ASSERT(pConsumer);
mInfo("receive consumer lost msg, consumer id %ld, status %s", pLostMsg->consumerId,
mndConsumerStatusName(pConsumer->status));
@ -106,7 +105,7 @@ static int32_t mndProcessConsumerLostMsg(SRpcMsg *pMsg) {
mndReleaseConsumer(pMnode, pConsumer);
STrans *pTrans = mndTransCreate(pMnode, TRN_POLICY_RETRY, TRN_CONFLICT_NOTHING, pMsg);
STrans *pTrans = mndTransCreate(pMnode, TRN_POLICY_ROLLBACK, TRN_CONFLICT_NOTHING, pMsg);
if (pTrans == NULL) goto FAIL;
if (mndSetConsumerCommitLogs(pMnode, pTrans, pConsumerNew) != 0) goto FAIL;
if (mndTransPrepare(pMnode, pTrans) != 0) goto FAIL;
@ -125,6 +124,14 @@ static int32_t mndProcessConsumerRecoverMsg(SRpcMsg *pMsg) {
SMqConsumerObj *pConsumer = mndAcquireConsumer(pMnode, pRecoverMsg->consumerId);
ASSERT(pConsumer);
mInfo("receive consumer recover msg, consumer id %ld, status %s", pRecoverMsg->consumerId,
mndConsumerStatusName(pConsumer->status));
if (pConsumer->status != MQ_CONSUMER_STATUS__READY) {
mndReleaseConsumer(pMnode, pConsumer);
return -1;
}
SMqConsumerObj *pConsumerNew = tNewSMqConsumerObj(pConsumer->consumerId, pConsumer->cgroup);
pConsumerNew->updateType = CONSUMER_UPDATE__RECOVER;
@ -440,6 +447,8 @@ static int32_t mndProcessSubscribeReq(SRpcMsg *pMsg) {
pConsumerOld = mndAcquireConsumer(pMnode, consumerId);
if (pConsumerOld == NULL) {
mInfo("receive subscribe request from new consumer: %ld", consumerId);
pConsumerNew = tNewSMqConsumerObj(consumerId, cgroup);
tstrncpy(pConsumerNew->clientId, subscribe.clientId, 256);
pConsumerNew->updateType = CONSUMER_UPDATE__MODIFY;
@ -456,7 +465,12 @@ static int32_t mndProcessSubscribeReq(SRpcMsg *pMsg) {
} else {
/*taosRLockLatch(&pConsumerOld->lock);*/
int32_t status = atomic_load_32(&pConsumerOld->status);
mInfo("receive subscribe request from old consumer: %ld, current status: %s", consumerId,
mndConsumerStatusName(status));
if (status != MQ_CONSUMER_STATUS__READY) {
terrno = TSDB_CODE_MND_CONSUMER_NOT_READY;
goto SUBSCRIBE_OVER;
@ -844,10 +858,11 @@ static int32_t mndRetrieveConsumer(SRpcMsg *pReq, SShowObj *pShow, SSDataBlock *
topicSz = 1;
}
for (int32_t i = 0; i < topicSz; i++) {
if (numOfRows + topicSz > rowsCapacity) {
blockDataEnsureCapacity(pBlock, numOfRows + topicSz);
}
for (int32_t i = 0; i < topicSz; i++) {
SColumnInfoData *pColInfo;
int32_t cols = 0;

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

@ -415,6 +415,7 @@ SMqSubscribeObj *tCloneSubscribeObj(const SMqSubscribeObj *pSub) {
taosHashPut(pSubNew->consumerHash, &newEp.consumerId, sizeof(int64_t), &newEp, sizeof(SMqConsumerEp));
}
pSubNew->unassignedVgs = taosArrayDeepCopy(pSub->unassignedVgs, (FCopy)tCloneSMqVgEp);
memcpy(pSubNew->dbName, pSub->dbName, TSDB_DB_FNAME_LEN);
return pSubNew;
}
@ -445,6 +446,7 @@ int32_t tEncodeSubscribeObj(void **buf, const SMqSubscribeObj *pSub) {
}
ASSERT(cnt == sz);
tlen += taosEncodeArray(buf, pSub->unassignedVgs, (FEncode)tEncodeSMqVgEp);
tlen += taosEncodeString(buf, pSub->dbName);
return tlen;
}
@ -467,6 +469,7 @@ void *tDecodeSubscribeObj(const void *buf, SMqSubscribeObj *pSub) {
}
buf = taosDecodeArray(buf, &pSub->unassignedVgs, (FDecode)tDecodeSMqVgEp, sizeof(SMqVgEp));
buf = taosDecodeStringTo(buf, pSub->dbName);
return (void *)buf;
}

4
source/dnode/mnode/impl/src/mndDnode.c
View File

@ -385,7 +385,7 @@ static int32_t mndProcessStatusReq(SRpcMsg *pReq) {
int64_t dnodeVer = sdbGetTableVer(pMnode->pSdb, SDB_DNODE) + sdbGetTableVer(pMnode->pSdb, SDB_MNODE);
int64_t curMs = taosGetTimestampMs();
bool online = mndIsDnodeOnline(pDnode, curMs);
bool dnodeChanged = (statusReq.dnodeVer != dnodeVer);
bool dnodeChanged = (statusReq.dnodeVer == 0) || (statusReq.dnodeVer != dnodeVer);
bool reboot = (pDnode->rebootTime != statusReq.rebootTime);
bool needCheck = !online || dnodeChanged || reboot;
@ -427,7 +427,7 @@ static int32_t mndProcessStatusReq(SRpcMsg *pReq) {
if (!online) {
mInfo("dnode:%d, from offline to online", pDnode->id);
} else {
mDebug("dnode:%d, send dnode epset, online:%d dnode_ver:%" PRId64 ":%" PRId64 " reboot:%d", pDnode->id, online,
mDebug("dnode:%d, send dnode epset, online:%d dnodeVer:%" PRId64 ":%" PRId64 " reboot:%d", pDnode->id, online,
statusReq.dnodeVer, dnodeVer, reboot);
}

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

@ -217,8 +217,8 @@ static int32_t mndInitSteps(SMnode *pMnode) {
if (mndAllocStep(pMnode, "mnode-cluster", mndInitCluster, mndCleanupCluster) != 0) return -1;
if (mndAllocStep(pMnode, "mnode-mnode", mndInitMnode, mndCleanupMnode) != 0) return -1;
if (mndAllocStep(pMnode, "mnode-qnode", mndInitQnode, mndCleanupQnode) != 0) return -1;
if (mndAllocStep(pMnode, "mnode-qnode", mndInitSnode, mndCleanupSnode) != 0) return -1;
if (mndAllocStep(pMnode, "mnode-qnode", mndInitBnode, mndCleanupBnode) != 0) return -1;
if (mndAllocStep(pMnode, "mnode-snode", mndInitSnode, mndCleanupSnode) != 0) return -1;
if (mndAllocStep(pMnode, "mnode-bnode", mndInitBnode, mndCleanupBnode) != 0) return -1;
if (mndAllocStep(pMnode, "mnode-dnode", mndInitDnode, mndCleanupDnode) != 0) return -1;
if (mndAllocStep(pMnode, "mnode-user", mndInitUser, mndCleanupUser) != 0) return -1;
if (mndAllocStep(pMnode, "mnode-grant", mndInitGrant, mndCleanupGrant) != 0) return -1;
@ -403,6 +403,10 @@ int32_t mndProcessSyncMsg(SRpcMsg *pMsg) {
char logBuf[512] = {0};
char *syncNodeStr = sync2SimpleStr(pMgmt->sync);
snprintf(logBuf, sizeof(logBuf), "==vnodeProcessSyncReq== 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);
}
syncRpcMsgLog2(logBuf, pMsg);
taosMemoryFree(syncNodeStr);

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

@ -183,7 +183,10 @@ static int32_t mndProcessCommitOffsetReq(SRpcMsg *pMsg) {
for (int32_t i = 0; i < commitOffsetReq.num; i++) {
SMqOffset *pOffset = &commitOffsetReq.offsets[i];
mInfo("commit offset %ld to vg %d of consumer group %s on topic %s", pOffset->offset, pOffset->vgId,
pOffset->cgroup, pOffset->topicName);
if (mndMakePartitionKey(key, pOffset->cgroup, pOffset->topicName, pOffset->vgId) < 0) {
mError("submit offset to topic %s failed", pOffset->topicName);
return -1;
}
bool create = false;

283
source/dnode/mnode/impl/src/mndScheduler.c
View File

@ -42,8 +42,8 @@ static int32_t mndAddTaskToTaskSet(SArray* pArray, SStreamTask* pTask) {
return 0;
}
int32_t mndConvertRSmaTask(const char* ast, int64_t uid, int8_t triggerType, int64_t watermark, char** pStr,
int32_t* pLen, double filesFactor) {
int32_t mndConvertRsmaTask(char** pDst, int32_t* pDstLen, const char* ast, int64_t uid, int8_t triggerType,
int64_t watermark, double filesFactor) {
SNode* pAst = NULL;
SQueryPlan* pPlan = NULL;
terrno = TSDB_CODE_SUCCESS;
@ -53,7 +53,7 @@ int32_t mndConvertRSmaTask(const char* ast, int64_t uid, int8_t triggerType, int
goto END;
}
if (qSetSTableIdForRSma(pAst, uid) < 0) {
if (qSetSTableIdForRsma(pAst, uid) < 0) {
terrno = TSDB_CODE_QRY_INVALID_INPUT;
goto END;
}
@ -86,7 +86,7 @@ int32_t mndConvertRSmaTask(const char* ast, int64_t uid, int8_t triggerType, int
}
SSubplan* plan = nodesListGetNode(inner->pNodeList, 0);
if (qSubPlanToString(plan, pStr, pLen) < 0) {
if (qSubPlanToString(plan, pDst, pDstLen) < 0) {
terrno = TSDB_CODE_QRY_INVALID_INPUT;
goto END;
}
@ -127,6 +127,61 @@ int32_t mndPersistTaskDeployReq(STrans* pTrans, SStreamTask* pTask, const SEpSet
return 0;
}
int32_t mndAddSinkToTask(SMnode* pMnode, STrans* pTrans, SStreamObj* pStream, SStreamTask* pTask) {
pTask->dispatchType = TASK_DISPATCH__NONE;
// sink
if (pStream->createdBy == STREAM_CREATED_BY__SMA) {
pTask->sinkType = TASK_SINK__SMA;
pTask->smaSink.smaId = pStream->smaId;
} else {
pTask->sinkType = TASK_SINK__TABLE;
pTask->tbSink.stbUid = pStream->targetStbUid;
memcpy(pTask->tbSink.stbFullName, pStream->targetSTbName, TSDB_TABLE_FNAME_LEN);
pTask->tbSink.pSchemaWrapper = tCloneSSchemaWrapper(&pStream->outputSchema);
}
return 0;
}
int32_t mndAddDispatcherToInnerTask(SMnode* pMnode, STrans* pTrans, SStreamObj* pStream, SStreamTask* pTask) {
pTask->sinkType = TASK_SINK__NONE;
if (pStream->fixedSinkVgId == 0) {
pTask->dispatchType = TASK_DISPATCH__SHUFFLE;
pTask->dispatchMsgType = TDMT_STREAM_TASK_DISPATCH;
SDbObj* pDb = mndAcquireDb(pMnode, pStream->targetDb);
ASSERT(pDb);
if (mndExtractDbInfo(pMnode, pDb, &pTask->shuffleDispatcher.dbInfo, NULL) < 0) {
sdbRelease(pMnode->pSdb, pDb);
SArray* pVgs = pTask->shuffleDispatcher.dbInfo.pVgroupInfos;
int32_t sz = taosArrayGetSize(pVgs);
SArray* sinkLv = taosArrayGetP(pStream->tasks, 0);
int32_t sinkLvSize = taosArrayGetSize(sinkLv);
for (int32_t i = 0; i < sz; i++) {
SVgroupInfo* pVgInfo = taosArrayGet(pVgs, i);
for (int32_t j = 0; j < sinkLvSize; j++) {
SStreamTask* pLastLevelTask = taosArrayGetP(sinkLv, j);
if (pLastLevelTask->nodeId == pVgInfo->vgId) {
pVgInfo->taskId = pLastLevelTask->taskId;
break;
}
}
}
}
} else {
pTask->dispatchType = TASK_DISPATCH__FIXED;
pTask->dispatchMsgType = TDMT_STREAM_TASK_DISPATCH;
SArray* pArray = taosArrayGetP(pStream->tasks, 0);
// one sink only
ASSERT(taosArrayGetSize(pArray) == 1);
SStreamTask* lastLevelTask = taosArrayGetP(pArray, 0);
pTask->fixedEpDispatcher.taskId = lastLevelTask->taskId;
pTask->fixedEpDispatcher.nodeId = lastLevelTask->nodeId;
pTask->fixedEpDispatcher.epSet = lastLevelTask->epSet;
}
return 0;
}
int32_t mndAssignTaskToVg(SMnode* pMnode, STrans* pTrans, SStreamTask* pTask, SSubplan* plan, const SVgObj* pVgroup) {
int32_t msgLen;
pTask->nodeId = pVgroup->vgId;
@ -139,6 +194,7 @@ int32_t mndAssignTaskToVg(SMnode* pMnode, STrans* pTrans, SStreamTask* pTask, SS
terrno = TSDB_CODE_QRY_INVALID_INPUT;
return -1;
}
ASSERT(pTask->dispatchType != TASK_DISPATCH__NONE || pTask->sinkType != TASK_SINK__NONE);
mndPersistTaskDeployReq(pTrans, pTask, &plan->execNode.epSet, TDMT_STREAM_TASK_DEPLOY, pVgroup->vgId);
return 0;
}
@ -182,7 +238,7 @@ SVgObj* mndSchedFetchOneVg(SMnode* pMnode, int64_t dbUid) {
return pVgroup;
}
int32_t mndAddShuffledSinkToStream(SMnode* pMnode, STrans* pTrans, SStreamObj* pStream) {
int32_t mndAddShuffleSinkTasksToStream(SMnode* pMnode, STrans* pTrans, SStreamObj* pStream) {
SSdb* pSdb = pMnode->pSdb;
void* pIter = NULL;
SArray* tasks = taosArrayGetP(pStream->tasks, 0);
@ -234,7 +290,7 @@ int32_t mndAddShuffledSinkToStream(SMnode* pMnode, STrans* pTrans, SStreamObj* p
return 0;
}
int32_t mndAddFixedSinkToStream(SMnode* pMnode, STrans* pTrans, SStreamObj* pStream) {
int32_t mndAddFixedSinkTaskToStream(SMnode* pMnode, STrans* pTrans, SStreamObj* pStream) {
ASSERT(pStream->fixedSinkVgId != 0);
SArray* tasks = taosArrayGetP(pStream->tasks, 0);
SStreamTask* pTask = tNewSStreamTask(pStream->uid);
@ -300,26 +356,49 @@ int32_t mndScheduleStream(SMnode* pMnode, STrans* pTrans, SStreamObj* pStream) {
// add extra sink
hasExtraSink = true;
if (pStream->fixedSinkVgId == 0) {
mndAddShuffledSinkToStream(pMnode, pTrans, pStream);
mndAddShuffleSinkTasksToStream(pMnode, pTrans, pStream);
} else {
mndAddFixedSinkToStream(pMnode, pTrans, pStream);
mndAddFixedSinkTaskToStream(pMnode, pTrans, pStream);
}
}
for (int32_t level = 0; level < totLevel; level++) {
SArray* taskOneLevel = taosArrayInit(0, sizeof(void*));
SNodeListNode* inner = nodesListGetNode(pPlan->pSubplans, level);
ASSERT(LIST_LENGTH(inner->pNodeList) == 1);
if (totLevel > 1) {
SStreamTask* pFinalTask;
// inner plan
{
SArray* taskInnerLevel = taosArrayInit(0, sizeof(void*));
taosArrayPush(pStream->tasks, &taskInnerLevel);
SNodeListNode* inner = nodesListGetNode(pPlan->pSubplans, 0);
SSubplan* plan = nodesListGetNode(inner->pNodeList, 0);
ASSERT(plan->subplanType == SUBPLAN_TYPE_MERGE);
// if (level == totLevel - 1 /* or no snode */) {
if (level == totLevel - 1) {
// last level, source, must assign to vnode
// must be scan type
pFinalTask = tNewSStreamTask(pStream->uid);
mndAddTaskToTaskSet(taskInnerLevel, pFinalTask);
// input
pFinalTask->inputType = TASK_INPUT_TYPE__DATA_BLOCK;
// dispatch
mndAddDispatcherToInnerTask(pMnode, pTrans, pStream, pFinalTask);
// exec
pFinalTask->execType = TASK_EXEC__PIPE;
SVgObj* pVgroup = mndSchedFetchOneVg(pMnode, pStream->dbUid);
if (mndAssignTaskToVg(pMnode, pTrans, pFinalTask, plan, pVgroup) < 0) {
sdbRelease(pSdb, pVgroup);
qDestroyQueryPlan(pPlan);
return -1;
}
}
// source plan
SArray* taskSourceLevel = taosArrayInit(0, sizeof(void*));
taosArrayPush(pStream->tasks, &taskSourceLevel);
SNodeListNode* inner = nodesListGetNode(pPlan->pSubplans, 1);
SSubplan* plan = nodesListGetNode(inner->pNodeList, 0);
ASSERT(plan->subplanType == SUBPLAN_TYPE_SCAN);
// replicate task to each vnode
void* pIter = NULL;
while (1) {
SVgObj* pVgroup;
@ -330,147 +409,39 @@ int32_t mndScheduleStream(SMnode* pMnode, STrans* pTrans, SStreamObj* pStream) {
continue;
}
SStreamTask* pTask = tNewSStreamTask(pStream->uid);
mndAddTaskToTaskSet(taskOneLevel, pTask);
// source part
pTask->sourceType = TASK_SOURCE__SCAN;
mndAddTaskToTaskSet(taskSourceLevel, pTask);
// input
pTask->inputType = TASK_INPUT_TYPE__SUMBIT_BLOCK;
// sink part
if (level == 0) {
// only for inplace
// add fixed vg dispatch
pTask->sinkType = TASK_SINK__NONE;
if (!hasExtraSink) {
#if 1
if (pStream->createdBy == STREAM_CREATED_BY__SMA) {
pTask->sinkType = TASK_SINK__SMA;
pTask->smaSink.smaId = pStream->smaId;
} else {
pTask->sinkType = TASK_SINK__TABLE;
pTask->tbSink.stbUid = pStream->targetStbUid;
memcpy(pTask->tbSink.stbFullName, pStream->targetSTbName, TSDB_TABLE_FNAME_LEN);
pTask->tbSink.pSchemaWrapper = tCloneSSchemaWrapper(&pStream->outputSchema);
}
#endif
}
} else {
pTask->sinkType = TASK_SINK__NONE;
}
// dispatch part
if (level == 0) {
pTask->dispatchType = TASK_DISPATCH__NONE;
} else {
// add fixed ep dispatcher
int32_t lastLevel = level - 1;
ASSERT(lastLevel == 0);
if (hasExtraSink) lastLevel++;
SArray* pArray = taosArrayGetP(pStream->tasks, lastLevel);
// one merge only
ASSERT(taosArrayGetSize(pArray) == 1);
SStreamTask* lastLevelTask = taosArrayGetP(pArray, 0);
/*pTask->dispatchMsgType = TDMT_VND_TASK_MERGE_EXEC;*/
pTask->dispatchMsgType = TDMT_STREAM_TASK_DISPATCH;
pTask->dispatchType = TASK_DISPATCH__FIXED;
pTask->fixedEpDispatcher.taskId = lastLevelTask->taskId;
pTask->fixedEpDispatcher.nodeId = lastLevelTask->nodeId;
pTask->fixedEpDispatcher.epSet = lastLevelTask->epSet;
}
pTask->fixedEpDispatcher.taskId = pFinalTask->taskId;
pTask->fixedEpDispatcher.nodeId = pFinalTask->nodeId;
pTask->fixedEpDispatcher.epSet = pFinalTask->epSet;
// exec part
// exec
pTask->execType = TASK_EXEC__PIPE;
if (mndAssignTaskToVg(pMnode, pTrans, pTask, plan, pVgroup) < 0) {
sdbRelease(pSdb, pVgroup);
qDestroyQueryPlan(pPlan);
return -1;
}
sdbRelease(pSdb, pVgroup);
}
} else {
// merge plan
// TODO if has snode, assign to snode
// else, assign to vnode
ASSERT(plan->subplanType == SUBPLAN_TYPE_MERGE);
SStreamTask* pTask = tNewSStreamTask(pStream->uid);
mndAddTaskToTaskSet(taskOneLevel, pTask);
// source part, currently only support multi source
pTask->sourceType = TASK_SOURCE__PIPE;
pTask->inputType = TASK_INPUT_TYPE__DATA_BLOCK;
// sink part
pTask->sinkType = TASK_SINK__NONE;
// dispatch part
ASSERT(hasExtraSink);
/*pTask->dispatchType = TASK_DISPATCH__NONE;*/
#if 1
if (hasExtraSink) {
// add dispatcher
if (pStream->fixedSinkVgId == 0) {
pTask->dispatchType = TASK_DISPATCH__SHUFFLE;
pTask->dispatchMsgType = TDMT_STREAM_TASK_DISPATCH;
SDbObj* pDb = mndAcquireDb(pMnode, pStream->targetDb);
ASSERT(pDb);
if (mndExtractDbInfo(pMnode, pDb, &pTask->shuffleDispatcher.dbInfo, NULL) < 0) {
sdbRelease(pSdb, pDb);
qDestroyQueryPlan(pPlan);
return -1;
}
sdbRelease(pSdb, pDb);
// put taskId to useDbRsp
// TODO: optimize
SArray* pVgs = pTask->shuffleDispatcher.dbInfo.pVgroupInfos;
int32_t sz = taosArrayGetSize(pVgs);
SArray* sinkLv = taosArrayGetP(pStream->tasks, 0);
int32_t sinkLvSize = taosArrayGetSize(sinkLv);
for (int32_t i = 0; i < sz; i++) {
SVgroupInfo* pVgInfo = taosArrayGet(pVgs, i);
for (int32_t j = 0; j < sinkLvSize; j++) {
SStreamTask* pLastLevelTask = taosArrayGetP(sinkLv, j);
if (pLastLevelTask->nodeId == pVgInfo->vgId) {
pVgInfo->taskId = pLastLevelTask->taskId;
break;
}
}
}
} else {
pTask->dispatchType = TASK_DISPATCH__FIXED;
/*pTask->dispatchMsgType = TDMT_VND_TASK_WRITE_EXEC;*/
pTask->dispatchMsgType = TDMT_STREAM_TASK_DISPATCH;
SArray* pArray = taosArrayGetP(pStream->tasks, 0);
// one sink only
ASSERT(taosArrayGetSize(pArray) == 1);
SStreamTask* lastLevelTask = taosArrayGetP(pArray, 0);
pTask->fixedEpDispatcher.taskId = lastLevelTask->taskId;
pTask->fixedEpDispatcher.nodeId = lastLevelTask->nodeId;
pTask->fixedEpDispatcher.epSet = lastLevelTask->epSet;
}
}
#endif
// exec part
pTask->execType = TASK_EXEC__MERGE;
SVgObj* pVgroup = mndSchedFetchOneVg(pMnode, pStream->dbUid);
ASSERT(pVgroup);
if (mndAssignTaskToVg(pMnode, pTrans, pTask, plan, pVgroup) < 0) {
sdbRelease(pSdb, pVgroup);
qDestroyQueryPlan(pPlan);
return -1;
}
sdbRelease(pSdb, pVgroup);
}
if (totLevel == 1) {
SArray* taskOneLevel = taosArrayInit(0, sizeof(void*));
taosArrayPush(pStream->tasks, &taskOneLevel);
}
#if 0
if (totLevel == 2) {
SNodeListNode* inner = nodesListGetNode(pPlan->pSubplans, 0);
ASSERT(LIST_LENGTH(inner->pNodeList) == 1);
SSubplan* plan = nodesListGetNode(inner->pNodeList, 0);
ASSERT(plan->subplanType == SUBPLAN_TYPE_SCAN);
void* pIter = NULL;
while (1) {
SVgObj* pVgroup;
@ -481,26 +452,28 @@ int32_t mndScheduleStream(SMnode* pMnode, STrans* pTrans, SStreamObj* pStream) {
continue;
}
SStreamTask* pTask = tNewSStreamTask(pStream->uid);
mndAddTaskToTaskSet(taskOneLevel, pTask);
// source part
pTask->sourceType = TASK_SOURCE__MERGE;
pTask->inputType = TASK_INPUT_TYPE__DATA_BLOCK;
// input
pTask->inputType = TASK_INPUT_TYPE__SUMBIT_BLOCK;
// sink part
pTask->sinkType = TASK_SINK__NONE;
// dispatch part
pTask->dispatchType = TASK_DISPATCH__NONE;
// exec part
pTask->execType = TASK_EXEC__NONE;
// sink or dispatch
if (hasExtraSink) {
mndAddDispatcherToInnerTask(pMnode, pTrans, pStream, pTask);
} else {
mndAddSinkToTask(pMnode, pTrans, pStream, pTask);
}
}
#endif
// free memory
// exec
pTask->execType = TASK_EXEC__PIPE;
if (mndAssignTaskToVg(pMnode, pTrans, pTask, plan, pVgroup) < 0) {
sdbRelease(pSdb, pVgroup);
qDestroyQueryPlan(pPlan);
return -1;
}
}
}
qDestroyQueryPlan(pPlan);
return 0;
}

4
source/dnode/mnode/impl/src/mndStb.c
View File

@ -395,13 +395,13 @@ static void *mndBuildVCreateStbReq(SMnode *pMnode, SVgObj *pVgroup, SStbObj *pSt
req.pRSmaParam.xFilesFactor = pStb->xFilesFactor;
req.pRSmaParam.delay = pStb->delay;
if (pStb->ast1Len > 0) {
if (mndConvertRSmaTask(pStb->pAst1, pStb->uid, 0, 0, &req.pRSmaParam.qmsg1, &req.pRSmaParam.qmsg1Len,
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 (pStb->ast2Len > 0) {
if (mndConvertRSmaTask(pStb->pAst2, pStb->uid, 0, 0, &req.pRSmaParam.qmsg2, &req.pRSmaParam.qmsg2Len,
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;
}

4
source/dnode/mnode/impl/src/mndSubscribe.c
View File

@ -402,7 +402,8 @@ static int32_t mndDoRebalance(SMnode *pMnode, const SMqRebInputObj *pInput, SMqR
}
static int32_t mndPersistRebResult(SMnode *pMnode, SRpcMsg *pMsg, const SMqRebOutputObj *pOutput) {
STrans *pTrans = mndTransCreate(pMnode, TRN_POLICY_ROLLBACK, TRN_CONFLICT_NOTHING, pMsg);
STrans *pTrans = mndTransCreate(pMnode, TRN_POLICY_ROLLBACK, TRN_CONFLICT_DB_INSIDE, pMsg);
mndTransSetDbName(pTrans, pOutput->pSub->dbName);
if (pTrans == NULL) return -1;
// make txn:
@ -547,6 +548,7 @@ static int32_t mndProcessRebalanceReq(SRpcMsg *pMsg) {
taosRLockLatch(&pTopic->lock);
rebOutput.pSub = mndCreateSub(pMnode, pTopic, pRebInfo->key);
memcpy(rebOutput.pSub->dbName, pTopic->db, TSDB_DB_FNAME_LEN);
ASSERT(taosHashGetSize(rebOutput.pSub->consumerHash) == 0);
taosRUnLockLatch(&pTopic->lock);

14
source/dnode/mnode/impl/src/mndSync.c
View File

@ -96,10 +96,18 @@ void mndRestoreFinish(struct SSyncFSM *pFsm) {
}
}
void mndReConfig(struct SSyncFSM *pFsm, SSyncCfg newCfg, SReConfigCbMeta cbMeta) {
void mndReConfig(struct SSyncFSM *pFsm, const SRpcMsg *pMsg, SReConfigCbMeta cbMeta) {
SMnode *pMnode = pFsm->data;
SSyncMgmt *pMgmt = &pMnode->syncMgmt;
#if 0
// send response
SRpcMsg rpcMsg = {.msgType = pMsg->msgType, .contLen = pMsg->contLen, .conn.applyIndex = cbMeta.index};
rpcMsg.pCont = rpcMallocCont(rpcMsg.contLen);
memcpy(rpcMsg.pCont, pMsg->pCont, pMsg->contLen);
syncGetAndDelRespRpc(pMnode->syncMgmt.sync, cbMeta.seqNum, &rpcMsg.info);
#endif
pMgmt->errCode = cbMeta.code;
mInfo("trans:-1, sync reconfig is proposed, saved:%d code:0x%x, index:%" PRId64 " term:%" PRId64, pMgmt->transId,
cbMeta.code, cbMeta.index, cbMeta.term);
@ -192,14 +200,14 @@ int32_t mndInitSync(SMnode *pMnode) {
return -1;
}
mDebug("mnode sync is opened, id:%" PRId64, pMgmt->sync);
mDebug("mnode-sync is opened, id:%" PRId64, pMgmt->sync);
return 0;
}
void mndCleanupSync(SMnode *pMnode) {
SSyncMgmt *pMgmt = &pMnode->syncMgmt;
syncStop(pMgmt->sync);
mDebug("mnode sync is stopped, id:%" PRId64, pMgmt->sync);
mDebug("mnode-sync is stopped, id:%" PRId64, pMgmt->sync);
tsem_destroy(&pMgmt->syncSem);
memset(pMgmt, 0, sizeof(SSyncMgmt));

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

@ -87,29 +87,21 @@ int32_t mndCheckColAndTagModifiable(SMnode *pMnode, int64_t suid, col_id_t colId
SNode *pAst = NULL;
if (nodesStringToNode(pTopic->ast, &pAst) != 0) {
ASSERT(0);
return false;
return -1;
}
SHashObj *pColHash = NULL;
SNodeList *pNodeList = NULL;
nodesCollectColumns((SSelectStmt *)pAst, SQL_CLAUSE_FROM, NULL, COLLECT_COL_TYPE_ALL, &pNodeList);
SNode *pNode = NULL;
FOREACH(pNode, pNodeList) {
SColumnNode *pCol = (SColumnNode *)pNode;
if (pCol->tableId != suid) goto NEXT;
if (pColHash == NULL) {
pColHash = taosHashInit(0, taosGetDefaultHashFunction(TSDB_DATA_TYPE_SMALLINT), false, HASH_NO_LOCK);
}
if (pCol->colId > 0) {
taosHashPut(pColHash, &pCol->colId, sizeof(int16_t), NULL, 0);
}
mTrace("topic:%s, colId:%d is used", pTopic->name, pCol->colId);
}
if (taosHashGet(pColHash, &colId, sizeof(int16_t)) != NULL) {
if (pCol->colId > 0 && pCol->colId == colId) {
found = true;
goto NEXT;
}
mTrace("topic:%s, colId:%d is used", pTopic->name, pCol->colId);
}
NEXT:
sdbRelease(pSdb, pTopic);
@ -563,7 +555,7 @@ static int32_t mndProcessDropTopicReq(SRpcMsg *pReq) {
mndReleaseConsumer(pMnode, pConsumer);
mndReleaseTopic(pMnode, pTopic);
terrno = TSDB_CODE_MND_TOPIC_SUBSCRIBED;
mError("topic:%s, failed to drop since subscribed by consumer %ld from cgroup %s", dropReq.name,
mError("topic:%s, failed to drop since subscribed by consumer %ld in consumer group %s", dropReq.name,
pConsumer->consumerId, pConsumer->cgroup);
return -1;
}
@ -580,7 +572,8 @@ static int32_t mndProcessDropTopicReq(SRpcMsg *pReq) {
}
#endif
STrans *pTrans = mndTransCreate(pMnode, TRN_POLICY_ROLLBACK, TRN_CONFLICT_NOTHING, pReq);
STrans *pTrans = mndTransCreate(pMnode, TRN_POLICY_ROLLBACK, TRN_CONFLICT_DB_INSIDE, pReq);
mndTransSetDbName(pTrans, pTopic->db);
if (pTrans == NULL) {
mError("topic:%s, failed to drop since %s", pTopic->name, terrstr());
return -1;

90
source/dnode/mnode/impl/src/mndTrans.c
View File

@ -297,7 +297,7 @@ static SSdbRow *mndTransActionDecode(SSdbRaw *pRaw) {
SDB_GET_INT32(pRaw, dataPos, &dataLen, _OVER)
action.pRaw = taosMemoryMalloc(dataLen);
if (action.pRaw == NULL) goto _OVER;
mTrace("raw:%p, is created", pData);
// mTrace("raw:%p, is created", pData);
SDB_GET_BINARY(pRaw, dataPos, (void *)action.pRaw, dataLen, _OVER);
if (taosArrayPush(pTrans->redoActions, &action) == NULL) goto _OVER;
action.pRaw = NULL;
@ -330,7 +330,7 @@ static SSdbRow *mndTransActionDecode(SSdbRaw *pRaw) {
SDB_GET_INT32(pRaw, dataPos, &dataLen, _OVER)
action.pRaw = taosMemoryMalloc(dataLen);
if (action.pRaw == NULL) goto _OVER;
mTrace("raw:%p, is created", pData);
// mTrace("raw:%p, is created", action.pRaw);
SDB_GET_BINARY(pRaw, dataPos, (void *)action.pRaw, dataLen, _OVER);
if (taosArrayPush(pTrans->undoActions, &action) == NULL) goto _OVER;
action.pRaw = NULL;
@ -363,7 +363,7 @@ static SSdbRow *mndTransActionDecode(SSdbRaw *pRaw) {
SDB_GET_INT32(pRaw, dataPos, &dataLen, _OVER)
action.pRaw = taosMemoryMalloc(dataLen);
if (action.pRaw == NULL) goto _OVER;
mTrace("raw:%p, is created", action.pRaw);
// mTrace("raw:%p, is created", action.pRaw);
SDB_GET_BINARY(pRaw, dataPos, (void *)action.pRaw, dataLen, _OVER);
if (taosArrayPush(pTrans->commitActions, &action) == NULL) goto _OVER;
action.pRaw = NULL;
@ -781,7 +781,7 @@ static void mndTransSendRpcRsp(SMnode *pMnode, STrans *pTrans) {
sendRsp = true;
}
} else {
if (pTrans->stage == TRN_STAGE_REDO_ACTION && pTrans->failedTimes > 3) {
if (pTrans->stage == TRN_STAGE_REDO_ACTION && pTrans->failedTimes > 2) {
if (code == 0) code = TSDB_CODE_MND_TRANS_UNKNOW_ERROR;
sendRsp = true;
}
@ -791,7 +791,7 @@ static void mndTransSendRpcRsp(SMnode *pMnode, STrans *pTrans) {
mDebug("trans:%d, send rsp, code:0x%x stage:%s app:%p", pTrans->id, code, mndTransStr(pTrans->stage),
pTrans->rpcInfo.ahandle);
if (code == TSDB_CODE_RPC_NETWORK_UNAVAIL) {
code = TSDB_CODE_RPC_INDIRECT_NETWORK_UNAVAIL;
code = TSDB_CODE_MND_TRANS_NETWORK_UNAVAILL;
}
SRpcMsg rspMsg = {.code = code, .info = pTrans->rpcInfo};
@ -894,10 +894,19 @@ static int32_t mndTransWriteSingleLog(SMnode *pMnode, STrans *pTrans, STransActi
code = 0;
mDebug("trans:%d, %s:%d write to sdb, type:%s status:%s", pTrans->id, mndTransStr(pAction->stage), pAction->id,
sdbTableName(pAction->pRaw->type), sdbStatusName(pAction->pRaw->status));
pTrans->lastAction = pAction->id;
pTrans->lastMsgType = pAction->msgType;
pTrans->lastEpset = pAction->epSet;
pTrans->lastErrorNo = 0;
} else {
pAction->errCode = (terrno != 0) ? terrno : code;
mError("trans:%d, %s:%d failed to write sdb since %s, type:%s status:%s", pTrans->id, mndTransStr(pAction->stage),
pAction->id, terrstr(), sdbTableName(pAction->pRaw->type), sdbStatusName(pAction->pRaw->status));
pTrans->lastAction = pAction->id;
pTrans->lastMsgType = pAction->msgType;
pTrans->lastEpset = pAction->epSet;
pTrans->lastErrorNo = pAction->errCode;
}
return code;
@ -933,27 +942,48 @@ static int32_t mndTransSendSingleMsg(SMnode *pMnode, STrans *pTrans, STransActio
pAction->msgReceived = 0;
pAction->errCode = 0;
mDebug("trans:%d, %s:%d is sent, %s", pTrans->id, mndTransStr(pAction->stage), pAction->id, detail);
pTrans->lastAction = pAction->id;
pTrans->lastMsgType = pAction->msgType;
pTrans->lastEpset = pAction->epSet;
if (pTrans->lastErrorNo == 0) {
pTrans->lastErrorNo = TSDB_CODE_ACTION_IN_PROGRESS;
}
} else {
pAction->msgSent = 0;
pAction->msgReceived = 0;
pAction->errCode = (terrno != 0) ? terrno : code;
mError("trans:%d, %s:%d not send since %s, %s", pTrans->id, mndTransStr(pAction->stage), pAction->id, terrstr(),
detail);
pTrans->lastAction = pAction->id;
pTrans->lastMsgType = pAction->msgType;
pTrans->lastEpset = pAction->epSet;
pTrans->lastErrorNo = pAction->errCode;
}
return code;
}
static int32_t mndTransExecNullMsg(SMnode *pMnode, STrans *pTrans, STransAction *pAction) {
pAction->rawWritten = 0;
pAction->errCode = 0;
mDebug("trans:%d, %s:%d null action executed", pTrans->id, mndTransStr(pAction->stage), pAction->id);
pTrans->lastAction = pAction->id;
pTrans->lastMsgType = pAction->msgType;
pTrans->lastEpset = pAction->epSet;
pTrans->lastErrorNo == 0;
return 0;
}
static int32_t mndTransExecSingleAction(SMnode *pMnode, STrans *pTrans, STransAction *pAction) {
if (pAction->actionType == TRANS_ACTION_RAW) {
return mndTransWriteSingleLog(pMnode, pTrans, pAction);
} else if (pAction->actionType == TRANS_ACTION_MSG) {
return mndTransSendSingleMsg(pMnode, pTrans, pAction);
} else {
pAction->rawWritten = 0;
pAction->errCode = 0;
mDebug("trans:%d, %s:%d null action executed", pTrans->id, mndTransStr(pAction->stage), pAction->id);
return 0;
return mndTransExecNullMsg(pMnode, pTrans, pAction);
}
}
@ -994,19 +1024,19 @@ static int32_t mndTransExecuteActions(SMnode *pMnode, STrans *pTrans, SArray *pA
if (numOfExecuted == numOfActions) {
if (errCode == 0) {
pTrans->lastErrorAction = 0;
pTrans->lastAction = 0;
pTrans->lastErrorNo = 0;
pTrans->lastErrorMsgType = 0;
memset(&pTrans->lastErrorEpset, 0, sizeof(pTrans->lastErrorEpset));
pTrans->lastMsgType = 0;
memset(&pTrans->lastEpset, 0, sizeof(pTrans->lastEpset));
mDebug("trans:%d, all %d actions execute successfully", pTrans->id, numOfActions);
return 0;
} else {
mError("trans:%d, all %d actions executed, code:0x%x", pTrans->id, numOfActions, errCode & 0XFFFF);
if (pErrAction != NULL) {
pTrans->lastErrorMsgType = pErrAction->msgType;
pTrans->lastErrorAction = pErrAction->id;
pTrans->lastMsgType = pErrAction->msgType;
pTrans->lastAction = pErrAction->id;
pTrans->lastErrorNo = pErrAction->errCode;
pTrans->lastErrorEpset = pErrAction->epSet;
pTrans->lastEpset = pErrAction->epSet;
}
mndTransResetActions(pMnode, pTrans, pArray);
terrno = errCode;
@ -1073,15 +1103,15 @@ static int32_t mndTransExecuteRedoActionsSerial(SMnode *pMnode, STrans *pTrans)
}
if (code == 0) {
pTrans->lastErrorAction = 0;
pTrans->lastAction = 0;
pTrans->lastErrorNo = 0;
pTrans->lastErrorMsgType = 0;
memset(&pTrans->lastErrorEpset, 0, sizeof(pTrans->lastErrorEpset));
pTrans->lastMsgType = 0;
memset(&pTrans->lastEpset, 0, sizeof(pTrans->lastEpset));
} else {
pTrans->lastErrorMsgType = pAction->msgType;
pTrans->lastErrorAction = action;
pTrans->lastErrorNo = pAction->errCode;
pTrans->lastErrorEpset = pAction->epSet;
pTrans->lastMsgType = pAction->msgType;
pTrans->lastAction = action;
pTrans->lastErrorNo = code;
pTrans->lastEpset = pAction->epSet;
}
if (code == 0) {
@ -1432,23 +1462,21 @@ static int32_t mndRetrieveTrans(SRpcMsg *pReq, SShowObj *pShow, SSDataBlock *pBl
pColInfo = taosArrayGet(pBlock->pDataBlock, cols++);
colDataAppend(pColInfo, numOfRows, (const char *)&pTrans->lastExecTime, false);
char lastError[TSDB_TRANS_ERROR_LEN + VARSTR_HEADER_SIZE] = {0};
char lastInfo[TSDB_TRANS_ERROR_LEN + VARSTR_HEADER_SIZE] = {0};
char detail[TSDB_TRANS_ERROR_LEN] = {0};
if (pTrans->lastErrorNo != 0) {
int32_t len = snprintf(detail, sizeof(detail), "action:%d errno:0x%x(%s) ", pTrans->lastErrorAction,
int32_t len = snprintf(detail, sizeof(detail), "action:%d code:0x%x(%s) ", pTrans->lastAction,
pTrans->lastErrorNo & 0xFFFF, tstrerror(pTrans->lastErrorNo));
SEpSet epset = pTrans->lastErrorEpset;
SEpSet epset = pTrans->lastEpset;
if (epset.numOfEps > 0) {
len += snprintf(detail + len, sizeof(detail) - len, "msgType:%s numOfEps:%d inUse:%d ",
TMSG_INFO(pTrans->lastErrorMsgType), epset.numOfEps, epset.inUse);
for (int32_t i = 0; i < pTrans->lastErrorEpset.numOfEps; ++i) {
TMSG_INFO(pTrans->lastMsgType), epset.numOfEps, epset.inUse);
for (int32_t i = 0; i < pTrans->lastEpset.numOfEps; ++i) {
len += snprintf(detail + len, sizeof(detail) - len, "ep:%d-%s:%u ", i, epset.eps[i].fqdn, epset.eps[i].port);
}
}
}
STR_WITH_MAXSIZE_TO_VARSTR(lastError, detail, pShow->pMeta->pSchemas[cols].bytes);
STR_WITH_MAXSIZE_TO_VARSTR(lastInfo, detail, pShow->pMeta->pSchemas[cols].bytes);
pColInfo = taosArrayGet(pBlock->pDataBlock, cols++);
colDataAppend(pColInfo, numOfRows, (const char *)lastError, false);
colDataAppend(pColInfo, numOfRows, (const char *)lastInfo, false);
numOfRows++;
sdbRelease(pSdb, pTrans);

236
source/dnode/mnode/impl/src/mndVgroup.c
View File

@ -59,6 +59,10 @@ int32_t mndInitVgroup(SMnode *pMnode) {
mndSetMsgHandle(pMnode, TDMT_DND_DROP_VNODE_RSP, mndTransProcessRsp);
mndSetMsgHandle(pMnode, TDMT_VND_COMPACT_RSP, mndTransProcessRsp);
mndSetMsgHandle(pMnode, TDMT_MND_REDISTRIBUTE_VGROUP, mndProcessRedistributeVgroupMsg);
mndSetMsgHandle(pMnode, TDMT_MND_MERGE_VGROUP, mndProcessSplitVgroupMsg);
mndSetMsgHandle(pMnode, TDMT_MND_BALANCE_VGROUP, mndProcessBalanceVgroupMsg);
mndAddShowRetrieveHandle(pMnode, TSDB_MGMT_TABLE_VGROUP, mndRetrieveVgroups);
mndAddShowFreeIterHandle(pMnode, TSDB_MGMT_TABLE_VGROUP, mndCancelGetNextVgroup);
mndAddShowRetrieveHandle(pMnode, TSDB_MGMT_TABLE_VNODES, mndRetrieveVnodes);
@ -390,12 +394,24 @@ SArray *mndBuildDnodesArray(SMnode *pMnode, int32_t exceptDnodeId) {
return pArray;
}
static int32_t mndCompareDnodeId(int32_t *dnode1Id, int32_t *dnode2Id) { return *dnode1Id >= *dnode2Id ? 1 : 0; }
static int32_t mndCompareDnodeVnodes(SDnodeObj *pDnode1, SDnodeObj *pDnode2) {
float d1Score = (float)pDnode1->numOfVnodes / pDnode1->numOfSupportVnodes;
float d2Score = (float)pDnode2->numOfVnodes / pDnode2->numOfSupportVnodes;
return d1Score >= d2Score ? 1 : 0;
}
void mndSortVnodeGid(SVgObj *pVgroup) {
for (int32_t i = 0; i < pVgroup->replica; ++i) {
for (int32_t j = 0; j < pVgroup->replica - 1 - i; ++j) {
if (pVgroup->vnodeGid[j].dnodeId > pVgroup->vnodeGid[j + 1].dnodeId) {
TSWAP(pVgroup->vnodeGid[j], pVgroup->vnodeGid[j + 1]);
}
}
}
}
static int32_t mndGetAvailableDnode(SMnode *pMnode, SVgObj *pVgroup, SArray *pArray) {
SSdb *pSdb = pMnode->pSdb;
int32_t allocedVnodes = 0;
@ -430,6 +446,7 @@ static int32_t mndGetAvailableDnode(SMnode *pMnode, SVgObj *pVgroup, SArray *pAr
pDnode->numOfVnodes++;
}
mndSortVnodeGid(pVgroup);
return 0;
}
@ -1009,10 +1026,10 @@ static int32_t mndAddDecVgroupReplicaFromTrans(SMnode *pMnode, STrans *pTrans, S
if (pGid == NULL) return 0;
pVgroup->replica--;
memcpy(&delGid, pGid, sizeof(SVnodeGid));
memcpy(pGid, &pVgroup->vnodeGid[pVgroup->replica], sizeof(SVnodeGid));
memset(&pVgroup->vnodeGid[pVgroup->replica], 0, sizeof(SVnodeGid));
pVgroup->replica--;
if (mndAddAlterVnodeAction(pMnode, pTrans, pDb, pVgroup, TDMT_VND_ALTER_REPLICA) != 0) return -1;
if (mndAddDropVnodeAction(pMnode, pTrans, pDb, pVgroup, &delGid, true) != 0) return -1;
@ -1031,7 +1048,7 @@ static int32_t mndRedistributeVgroup(SMnode *pMnode, SRpcMsg *pReq, SDbObj *pDb,
pTrans = mndTransCreate(pMnode, TRN_POLICY_RETRY, TRN_CONFLICT_GLOBAL, pReq);
if (pTrans == NULL) goto _OVER;
mndTransSetSerial(pTrans);
mDebug("trans:%d, used to drop redistribute vgId:%d", pTrans->id, pVgroup->vgId);
mDebug("trans:%d, used to redistribute vgroup, vgId:%d", pTrans->id, pVgroup->vgId);
SVgObj newVg = {0};
memcpy(&newVg, pVgroup, sizeof(SVgObj));
@ -1040,11 +1057,38 @@ static int32_t mndRedistributeVgroup(SMnode *pMnode, SRpcMsg *pReq, SDbObj *pDb,
mInfo("vgId:%d, vnode:%d dnode:%d", newVg.vgId, i, newVg.vnodeGid[i].dnodeId);
}
if (pNew1 != NULL && pOld1 != NULL) {
int32_t numOfVnodes = mndGetVnodesNum(pMnode, pNew1->id);
if (numOfVnodes >= pNew1->numOfSupportVnodes) {
mError("vgId:%d, no enough vnodes in dnode:%d, numOfVnodes:%d support:%d", newVg.vgId, pNew1->id, numOfVnodes,
pNew1->numOfSupportVnodes);
terrno = TSDB_CODE_MND_NO_ENOUGH_DNODES;
goto _OVER;
}
if (mndAddIncVgroupReplicaToTrans(pMnode, pTrans, pDb, &newVg, pNew1->id) != 0) goto _OVER;
if (mndAddDecVgroupReplicaFromTrans(pMnode, pTrans, pDb, &newVg, pOld1->id) != 0) goto _OVER;
if (pNew2 != NULL) {
}
if (pNew2 != NULL && pOld2 != NULL) {
int32_t numOfVnodes = mndGetVnodesNum(pMnode, pNew2->id);
if (numOfVnodes >= pNew2->numOfSupportVnodes) {
mError("vgId:%d, no enough vnodes in dnode:%d, numOfVnodes:%d support:%d", newVg.vgId, pNew2->id, numOfVnodes,
pNew2->numOfSupportVnodes);
terrno = TSDB_CODE_MND_NO_ENOUGH_DNODES;
goto _OVER;
}
if (mndAddIncVgroupReplicaToTrans(pMnode, pTrans, pDb, &newVg, pNew2->id) != 0) goto _OVER;
if (mndAddDecVgroupReplicaFromTrans(pMnode, pTrans, pDb, &newVg, pOld2->id) != 0) goto _OVER;
}
if (pNew3 != NULL && pOld3 != NULL) {
int32_t numOfVnodes = mndGetVnodesNum(pMnode, pNew3->id);
if (numOfVnodes >= pNew3->numOfSupportVnodes) {
mError("vgId:%d, no enough vnodes in dnode:%d, numOfVnodes:%d support:%d", newVg.vgId, pNew3->id, numOfVnodes,
pNew3->numOfSupportVnodes);
terrno = TSDB_CODE_MND_NO_ENOUGH_DNODES;
goto _OVER;
}
if (mndAddIncVgroupReplicaToTrans(pMnode, pTrans, pDb, &newVg, pNew3->id) != 0) goto _OVER;
if (mndAddDecVgroupReplicaFromTrans(pMnode, pTrans, pDb, &newVg, pOld3->id) != 0) goto _OVER;
}
@ -1082,76 +1126,178 @@ static int32_t mndProcessRedistributeVgroupMsg(SRpcMsg *pReq) {
SDbObj *pDb = NULL;
int32_t code = -1;
int64_t curMs = taosGetTimestampMs();
SMDropMnodeReq redReq = {0};
int32_t newDnodeId[3] = {0};
int32_t oldDnodeId[3] = {0};
int32_t newIndex = -1;
int32_t oldIndex = -1;
#if 0
if (tDeserializeSCreateDropMQSBNodeReq(pReq->pCont, pReq->contLen, &dropReq) != 0) {
SRedistributeVgroupReq req = {0};
if (tDeserializeSRedistributeVgroupReq(pReq->pCont, pReq->contLen, &req) != 0) {
terrno = TSDB_CODE_INVALID_MSG;
goto _OVER;
}
#endif
mDebug("vgId:%d, start to redistribute", 2);
mInfo("vgId:%d, start to redistribute to dnode %d:%d:%d", req.vgId, req.dnodeId1, req.dnodeId2, req.dnodeId3);
pUser = mndAcquireUser(pMnode, pReq->conn.user);
if (pUser == NULL) {
terrno = TSDB_CODE_MND_NO_USER_FROM_CONN;
goto _OVER;
}
if (mndCheckNodeAuth(pUser) != 0) {
goto _OVER;
}
if (mndCheckNodeAuth(pUser) != 0) goto _OVER;
pVgroup = mndAcquireVgroup(pMnode, 2);
pVgroup = mndAcquireVgroup(pMnode, req.vgId);
if (pVgroup == NULL) goto _OVER;
pDb = mndAcquireDb(pMnode, pVgroup->dbName);
if (pDb == NULL) goto _OVER;
if (pVgroup->replica == 1) {
pNew1 = mndAcquireDnode(pMnode, 1);
pOld1 = mndAcquireDnode(pMnode, pVgroup->vnodeGid[0].dnodeId);
if (pNew1 == NULL || pOld1 == NULL) goto _OVER;
if (!mndIsDnodeOnline(pNew1, curMs) || !mndIsDnodeOnline(pOld1, curMs)) {
terrno = TSDB_CODE_NODE_OFFLINE;
if (req.dnodeId1 <= 0 || req.dnodeId2 > 0 || req.dnodeId3 > 0) {
terrno = TSDB_CODE_MND_INVALID_REPLICA;
goto _OVER;
}
if (pNew1 == pOld1) {
terrno = TSDB_CODE_MND_VGROUP_UN_CHANGED;
goto _OVER;
}
if (mndRedistributeVgroup(pMnode, pReq, pDb, pVgroup, pNew1, pOld1, NULL, NULL, NULL, NULL) != 0) goto _OVER;
}
if (pVgroup->replica == 3) {
pNew1 = mndAcquireDnode(pMnode, 1);
pNew2 = mndAcquireDnode(pMnode, 2);
pNew3 = mndAcquireDnode(pMnode, 3);
pOld1 = mndAcquireDnode(pMnode, pVgroup->vnodeGid[0].dnodeId);
pOld2 = mndAcquireDnode(pMnode, pVgroup->vnodeGid[1].dnodeId);
pOld3 = mndAcquireDnode(pMnode, pVgroup->vnodeGid[2].dnodeId);
if (pNew1 == NULL || pOld1 == NULL || pNew2 == NULL || pOld2 == NULL || pNew3 == NULL || pOld3 == NULL) goto _OVER;
if (!mndIsDnodeOnline(pNew1, curMs) || !mndIsDnodeOnline(pOld1, curMs) || !mndIsDnodeOnline(pNew2, curMs) ||
!mndIsDnodeOnline(pOld2, curMs) || !mndIsDnodeOnline(pNew3, curMs) || !mndIsDnodeOnline(pOld3, curMs)) {
terrno = TSDB_CODE_NODE_OFFLINE;
goto _OVER;
}
bool changed = true;
if (pNew1 != pOld1 || pNew1 != pOld2 || pNew1 != pOld3) changed = true;
if (pNew2 != pOld1 || pNew2 != pOld2 || pNew2 != pOld3) changed = true;
if (pNew3 != pOld1 || pNew3 != pOld2 || pNew3 != pOld3) changed = true;
if (!changed) {
if (req.dnodeId1 == pVgroup->vnodeGid[0].dnodeId) {
terrno = TSDB_CODE_MND_VGROUP_UN_CHANGED;
goto _OVER;
}
if (mndRedistributeVgroup(pMnode, pReq, pDb, pVgroup, pNew1, pOld1, pNew2, pOld2, pNew3, pOld3) != 0) goto _OVER;
pNew1 = mndAcquireDnode(pMnode, req.dnodeId1);
if (pNew1 == NULL) goto _OVER;
if (!mndIsDnodeOnline(pNew1, curMs)) {
terrno = TSDB_CODE_MND_HAS_OFFLINE_DNODE;
goto _OVER;
}
pOld1 = mndAcquireDnode(pMnode, pVgroup->vnodeGid[0].dnodeId);
if (pOld1 == NULL) goto _OVER;
if (!mndIsDnodeOnline(pOld1, curMs)) {
terrno = TSDB_CODE_MND_HAS_OFFLINE_DNODE;
goto _OVER;
}
code = mndRedistributeVgroup(pMnode, pReq, pDb, pVgroup, pNew1, pOld1, NULL, NULL, NULL, NULL);
} else if (pVgroup->replica == 3) {
if (req.dnodeId1 <= 0 || req.dnodeId2 <= 0 || req.dnodeId3 <= 0) {
terrno = TSDB_CODE_MND_INVALID_REPLICA;
goto _OVER;
}
if (req.dnodeId1 == req.dnodeId2 || req.dnodeId1 == req.dnodeId3 || req.dnodeId2 == req.dnodeId3) {
terrno = TSDB_CODE_MND_INVALID_REPLICA;
goto _OVER;
}
if (req.dnodeId1 != pVgroup->vnodeGid[0].dnodeId && req.dnodeId1 != pVgroup->vnodeGid[1].dnodeId &&
req.dnodeId1 != pVgroup->vnodeGid[2].dnodeId) {
newDnodeId[++newIndex] = req.dnodeId1;
mInfo("vgId:2, dnode:%d will be added", newDnodeId[newIndex]);
}
if (req.dnodeId2 != pVgroup->vnodeGid[0].dnodeId && req.dnodeId2 != pVgroup->vnodeGid[1].dnodeId &&
req.dnodeId2 != pVgroup->vnodeGid[2].dnodeId) {
newDnodeId[++newIndex] = req.dnodeId2;
mInfo("vgId:2, dnode:%d will be added", newDnodeId[newIndex]);
}
if (req.dnodeId3 != pVgroup->vnodeGid[0].dnodeId && req.dnodeId3 != pVgroup->vnodeGid[1].dnodeId &&
req.dnodeId3 != pVgroup->vnodeGid[2].dnodeId) {
newDnodeId[++newIndex] = req.dnodeId3;
mInfo("vgId:2, dnode:%d will be added", newDnodeId[newIndex]);
}
if (req.dnodeId1 != pVgroup->vnodeGid[0].dnodeId && req.dnodeId2 != pVgroup->vnodeGid[0].dnodeId &&
req.dnodeId3 != pVgroup->vnodeGid[0].dnodeId) {
oldDnodeId[++oldIndex] = pVgroup->vnodeGid[0].dnodeId;
mInfo("vgId:2, dnode:%d will be removed", oldDnodeId[oldIndex]);
}
if (req.dnodeId1 != pVgroup->vnodeGid[1].dnodeId && req.dnodeId2 != pVgroup->vnodeGid[1].dnodeId &&
req.dnodeId3 != pVgroup->vnodeGid[1].dnodeId) {
oldDnodeId[++oldIndex] = pVgroup->vnodeGid[1].dnodeId;
mInfo("vgId:2, dnode:%d will be removed", oldDnodeId[oldIndex]);
}
if (req.dnodeId1 != pVgroup->vnodeGid[2].dnodeId && req.dnodeId2 != pVgroup->vnodeGid[2].dnodeId &&
req.dnodeId3 != pVgroup->vnodeGid[2].dnodeId) {
oldDnodeId[++oldIndex] = pVgroup->vnodeGid[2].dnodeId;
mInfo("vgId:2, dnode:%d will be removed", oldDnodeId[oldIndex]);
}
if (newDnodeId[0] != 0) {
pNew1 = mndAcquireDnode(pMnode, newDnodeId[0]);
if (pNew1 == NULL) goto _OVER;
if (!mndIsDnodeOnline(pNew1, curMs)) {
terrno = TSDB_CODE_MND_HAS_OFFLINE_DNODE;
goto _OVER;
}
}
if (newDnodeId[1] != 0) {
pNew2 = mndAcquireDnode(pMnode, newDnodeId[1]);
if (pNew2 == NULL) goto _OVER;
if (!mndIsDnodeOnline(pNew2, curMs)) {
terrno = TSDB_CODE_MND_HAS_OFFLINE_DNODE;
goto _OVER;
}
}
if (newDnodeId[2] != 0) {
pNew3 = mndAcquireDnode(pMnode, newDnodeId[2]);
if (pNew3 == NULL) goto _OVER;
if (!mndIsDnodeOnline(pNew3, curMs)) {
terrno = TSDB_CODE_MND_HAS_OFFLINE_DNODE;
goto _OVER;
}
}
if (oldDnodeId[0] != 0) {
pOld1 = mndAcquireDnode(pMnode, oldDnodeId[0]);
if (pOld1 == NULL) goto _OVER;
if (!mndIsDnodeOnline(pOld1, curMs)) {
terrno = TSDB_CODE_MND_HAS_OFFLINE_DNODE;
goto _OVER;
}
}
if (oldDnodeId[1] != 0) {
pOld2 = mndAcquireDnode(pMnode, oldDnodeId[1]);
if (pOld2 == NULL) goto _OVER;
if (!mndIsDnodeOnline(pOld2, curMs)) {
terrno = TSDB_CODE_MND_HAS_OFFLINE_DNODE;
goto _OVER;
}
}
if (oldDnodeId[2] != 0) {
pOld3 = mndAcquireDnode(pMnode, oldDnodeId[2]);
if (pOld3 == NULL) goto _OVER;
if (!mndIsDnodeOnline(pOld3, curMs)) {
terrno = TSDB_CODE_MND_HAS_OFFLINE_DNODE;
goto _OVER;
}
}
if (pNew1 == NULL && pOld1 == NULL && pNew2 == NULL && pOld2 == NULL && pNew3 == NULL && pOld3 == NULL) {
terrno = TSDB_CODE_MND_VGROUP_UN_CHANGED;
goto _OVER;
}
code = mndRedistributeVgroup(pMnode, pReq, pDb, pVgroup, pNew1, pOld1, pNew2, pOld2, pNew3, pOld3);
} else {
terrno = TSDB_CODE_MND_INVALID_REPLICA;
goto _OVER;
}
if (code == 0) code = TSDB_CODE_ACTION_IN_PROGRESS;
_OVER:
if (code != 0 && code != TSDB_CODE_ACTION_IN_PROGRESS) {
mDebug("vgId:%d, failed to redistribute since %s", 1, terrstr());
mError("vgId:%d, failed to redistribute to dnode %d:%d:%d since %s", req.vgId, req.dnodeId1, req.dnodeId2,
req.dnodeId3, terrstr());
}
mndReleaseDnode(pMnode, pNew1);
@ -1303,9 +1449,7 @@ static int32_t mndProcessSplitVgroupMsg(SRpcMsg *pReq) {
goto _OVER;
}
if (mndCheckNodeAuth(pUser) != 0) {
goto _OVER;
}
if (mndCheckNodeAuth(pUser) != 0) goto _OVER;
code = mndSplitVgroup(pMnode, pReq, pDb, pVgroup);
if (code == 0) code = TSDB_CODE_ACTION_IN_PROGRESS;

7
source/dnode/mnode/sdb/src/sdb.c
View File

@ -163,7 +163,12 @@ void sdbSetApplyIndex(SSdb *pSdb, int64_t index) { pSdb->curVer = index; }
void sdbSetApplyTerm(SSdb *pSdb, int64_t term) { pSdb->curTerm = term; }
void sdbSetCurConfig(SSdb *pSdb, int64_t config) { pSdb->curConfig = config; }
void sdbSetCurConfig(SSdb *pSdb, int64_t config) {
if (pSdb->curConfig != config) {
mDebug("mnode sync config set from %" PRId64 " to %" PRId64, pSdb->curConfig, config);
pSdb->curConfig = config;
}
}
int64_t sdbGetApplyIndex(SSdb *pSdb) { return pSdb->curVer; }

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

@ -357,7 +357,7 @@ static int32_t sdbWriteFileImp(SSdb *pSdb) {
SdbEncodeFp encodeFp = pSdb->encodeFps[i];
if (encodeFp == NULL) continue;
mTrace("write %s to sdb file, total %d rows", sdbTableName(i), sdbGetSize(pSdb, i));
mDebug("write %s to sdb file, total %d rows", sdbTableName(i), sdbGetSize(pSdb, i));
SHashObj *hash = pSdb->hashObjs[i];
TdThreadRwlock *pLock = &pSdb->locks[i];
@ -432,8 +432,8 @@ static int32_t sdbWriteFileImp(SSdb *pSdb) {
} else {
pSdb->lastCommitVer = pSdb->curVer;
pSdb->lastCommitTerm = pSdb->curTerm;
mDebug("write sdb file successfully, ver:%" PRId64 " term:%" PRId64 " file:%s", pSdb->lastCommitVer,
pSdb->lastCommitTerm, curfile);
mDebug("write sdb file successfully, index:%" PRId64 " term:%" PRId64 " config:%" PRId64 " file:%s",
pSdb->lastCommitVer, pSdb->lastCommitTerm, pSdb->curConfig, curfile);
}
terrno = code;

2
source/dnode/mnode/sdb/src/sdbHash.c
View File

@ -83,6 +83,7 @@ const char *sdbStatusName(ESdbStatus status) {
}
void sdbPrintOper(SSdb *pSdb, SSdbRow *pRow, const char *oper) {
#if 0
EKeyType keyType = pSdb->keyTypes[pRow->type];
if (keyType == SDB_KEY_BINARY) {
@ -96,6 +97,7 @@ void sdbPrintOper(SSdb *pSdb, SSdbRow *pRow, const char *oper) {
pRow->refCount, oper, pRow->pObj, sdbStatusName(pRow->status));
} else {
}
#endif
}
static SHashObj *sdbGetHash(SSdb *pSdb, int32_t type) {

4
source/dnode/mnode/sdb/src/sdbRaw.c
View File

@ -37,13 +37,17 @@ SSdbRaw *sdbAllocRaw(ESdbType type, int8_t sver, int32_t dataLen) {
pRaw->sver = sver;
pRaw->dataLen = dataLen;
#if 0
mTrace("raw:%p, is created, len:%d table:%s", pRaw, dataLen, sdbTableName(type));
#endif
return pRaw;
}
void sdbFreeRaw(SSdbRaw *pRaw) {
if (pRaw != NULL) {
#if 0
mTrace("raw:%p, is freed", pRaw);
#endif
taosMemoryFree(pRaw);
}
}

4
source/dnode/mnode/sdb/src/sdbRow.c
View File

@ -23,7 +23,9 @@ SSdbRow *sdbAllocRow(int32_t objSize) {
return NULL;
}
#if 0
mTrace("row:%p, is created, len:%d", pRow->pObj, objSize);
#endif
return pRow;
}
@ -45,6 +47,8 @@ void sdbFreeRow(SSdb *pSdb, SSdbRow *pRow, bool callFunc) {
sdbPrintOper(pSdb, pRow, "free");
#if 0
mTrace("row:%p, is freed", pRow->pObj);
#endif
taosMemoryFreeClear(pRow);
}

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

@ -116,7 +116,7 @@ typedef void *tsdbReaderT;
#define BLOCK_LOAD_TABLE_SEQ_ORDER 2
#define BLOCK_LOAD_TABLE_RR_ORDER 3
tsdbReaderT *tsdbQueryTables(SVnode *pVnode, SQueryTableDataCond *pCond, STableListInfo *tableInfoGroup, uint64_t qId,
tsdbReaderT *tsdbReaderOpen(SVnode *pVnode, SQueryTableDataCond *pCond, STableListInfo *tableInfoGroup, uint64_t qId,
uint64_t taskId);
tsdbReaderT tsdbQueryCacheLast(SVnode *pVnode, SQueryTableDataCond *pCond, STableListInfo *groupList, uint64_t qId,
void *pMemRef);

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

@ -58,7 +58,7 @@ struct STqReadHandle {
SArray* pColIdList; // SArray<int16_t>
int32_t cachedSchemaVer;
int64_t cachedSchemaUid;
int64_t cachedSchemaSuid;
SSchemaWrapper* pSchemaWrapper;
STSchema* pSchema;
};

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

@ -119,7 +119,7 @@ int tsdbInsertData(STsdb* pTsdb, int64_t version, SSubmitReq* pMsg, SSu
int32_t tsdbInsertTableData(STsdb* pTsdb, int64_t version, SSubmitMsgIter* pMsgIter, SSubmitBlk* pBlock,
SSubmitBlkRsp* pRsp);
int32_t tsdbDeleteTableData(STsdb* pTsdb, int64_t version, tb_uid_t suid, tb_uid_t uid, TSKEY sKey, TSKEY eKey);
tsdbReaderT* tsdbQueryTables(SVnode* pVnode, SQueryTableDataCond* pCond, STableListInfo* tableList, uint64_t qId,
tsdbReaderT* tsdbReaderOpen(SVnode* pVnode, SQueryTableDataCond* pCond, STableListInfo* tableList, uint64_t qId,
uint64_t taskId);
tsdbReaderT tsdbQueryCacheLastT(STsdb* tsdb, SQueryTableDataCond* pCond, STableListInfo* tableList, uint64_t qId,
void* pMemRef);

12
source/dnode/vnode/src/tq/tq.c
View File

@ -130,7 +130,17 @@ int32_t tqProcessPollReq(STQ* pTq, SRpcMsg* pMsg, int32_t workerId) {
TD_VID(pTq->pVnode), pReq->currentOffset, fetchOffset);
STqHandle* pHandle = taosHashGet(pTq->handles, pReq->subKey, strlen(pReq->subKey));
ASSERT(pHandle);
/*ASSERT(pHandle);*/
if (pHandle == NULL) {
tqError("tmq poll: no consumer handle for consumer %ld in vg %d, subkey %s", consumerId, pTq->pVnode->config.vgId,
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);
return -1;
}
int32_t consumerEpoch = atomic_load_32(&pHandle->epoch);
while (consumerEpoch < reqEpoch) {

7
source/dnode/vnode/src/tq/tqRead.c
View File

@ -67,7 +67,7 @@ STqReadHandle* tqInitSubmitMsgScanner(SMeta* pMeta) {
pReadHandle->ver = -1;
pReadHandle->pColIdList = NULL;
pReadHandle->cachedSchemaVer = -1;
pReadHandle->cachedSchemaUid = -1;
pReadHandle->cachedSchemaSuid = -1;
pReadHandle->pSchema = NULL;
pReadHandle->pSchemaWrapper = NULL;
pReadHandle->tbIdHash = NULL;
@ -130,7 +130,8 @@ int32_t tqRetrieveDataBlock(SArray** ppCols, STqReadHandle* pHandle, uint64_t* p
// TODO set to real sversion
/*int32_t sversion = 1;*/
int32_t sversion = htonl(pHandle->pBlock->sversion);
if (pHandle->cachedSchemaVer != sversion || pHandle->cachedSchemaUid != pHandle->msgIter.suid) {
if (pHandle->cachedSchemaSuid == 0 || pHandle->cachedSchemaVer != sversion ||
pHandle->cachedSchemaSuid != pHandle->msgIter.suid) {
pHandle->pSchema = metaGetTbTSchema(pHandle->pVnodeMeta, pHandle->msgIter.uid, sversion);
if (pHandle->pSchema == NULL) {
tqWarn("cannot found tsschema for table: uid: %ld (suid: %ld), version %d, possibly dropped table",
@ -150,7 +151,7 @@ int32_t tqRetrieveDataBlock(SArray** ppCols, STqReadHandle* pHandle, uint64_t* p
return -1;
}
pHandle->cachedSchemaVer = sversion;
pHandle->cachedSchemaUid = pHandle->msgIter.suid;
pHandle->cachedSchemaSuid = pHandle->msgIter.suid;
}
STSchema* pTschema = pHandle->pSchema;

4
source/dnode/vnode/src/tsdb/tsdbRead.c
View File

@ -500,7 +500,7 @@ static int32_t setCurrentSchema(SVnode* pVnode, STsdbReadHandle* pTsdbReadHandle
return TSDB_CODE_SUCCESS;
}
tsdbReaderT* tsdbQueryTables(SVnode* pVnode, SQueryTableDataCond* pCond, STableListInfo* tableList, uint64_t qId,
tsdbReaderT* tsdbReaderOpen(SVnode* pVnode, SQueryTableDataCond* pCond, STableListInfo* tableList, uint64_t qId,
uint64_t taskId) {
STsdbReadHandle* pTsdbReadHandle = tsdbQueryTablesImpl(pVnode, pCond, qId, taskId);
if (pTsdbReadHandle == NULL) {
@ -642,7 +642,7 @@ tsdbReaderT tsdbQueryLastRow(SVnode* pVnode, SQueryTableDataCond* pCond, STableL
return NULL;
}
STsdbReadHandle* pTsdbReadHandle = (STsdbReadHandle*)tsdbQueryTables(pVnode, pCond, pList, qId, taskId);
STsdbReadHandle* pTsdbReadHandle = (STsdbReadHandle*)tsdbReaderOpen(pVnode, pCond, pList, qId, taskId);
if (pTsdbReadHandle == NULL) {
return NULL;
}

4
source/dnode/vnode/src/vnd/vnodeSvr.c
View File

@ -303,6 +303,10 @@ int32_t vnodeProcessSyncReq(SVnode *pVnode, SRpcMsg *pMsg, SRpcMsg **pRsp) {
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;
if (++vndTick % 10 == 1) {
vTrace("sync trace msg:%s, %s", TMSG_INFO(pMsg->msgType), syncNodeStr);
}
syncRpcMsgLog2(logBuf, pMsg);
taosMemoryFree(syncNodeStr);

12
source/dnode/vnode/src/vnd/vnodeSync.c
View File

@ -180,13 +180,22 @@ static int32_t vnodeSyncGetSnapshot(SSyncFSM *pFsm, SSnapshot *pSnapshot) {
return 0;
}
static void vnodeSyncReconfig(struct SSyncFSM *pFsm, SSyncCfg newCfg, SReConfigCbMeta cbMeta) {
static void vnodeSyncReconfig(struct SSyncFSM *pFsm, const SRpcMsg *pMsg, SReConfigCbMeta cbMeta) {
SVnode *pVnode = pFsm->data;
vInfo("vgId:%d, sync reconfig is confirmed", TD_VID(pVnode));
#if 0
// send response
SRpcMsg rpcMsg = {.msgType = pMsg->msgType, .contLen = pMsg->contLen, .conn.applyIndex = cbMeta.index};
rpcMsg.pCont = rpcMallocCont(rpcMsg.contLen);
memcpy(rpcMsg.pCont, pMsg->pCont, pMsg->contLen);
syncGetAndDelRespRpc(pVnode->sync, cbMeta.seqNum, &rpcMsg.info);
#endif
// todo rpc response here
// build rpc msg
// put into apply queue
vnodePostBlockMsg(pVnode, TDMT_VND_ALTER_REPLICA);
}
static void vnodeSyncCommitMsg(SSyncFSM *pFsm, const SRpcMsg *pMsg, SFsmCbMeta cbMeta) {
@ -212,6 +221,7 @@ static void vnodeSyncCommitMsg(SSyncFSM *pFsm, const SRpcMsg *pMsg, SFsmCbMeta c
memcpy(rpcMsg.pCont, pMsg->pCont, pMsg->contLen);
syncGetAndDelRespRpc(pVnode->sync, cbMeta.seqNum, &rpcMsg.info);
tmsgPutToQueue(&pVnode->msgCb, APPLY_QUEUE, &rpcMsg);
} else {
char logBuf[256] = {0};
snprintf(logBuf, sizeof(logBuf),

92
source/libs/catalog/src/catalog.c
View File

@ -13,16 +13,15 @@
* along with this program. If not, see <http://www.gnu.org/licenses/>.
*/
#include "trpc.h"
#include "query.h"
#include "tname.h"
#include "catalogInt.h"
#include "query.h"
#include "systable.h"
#include "tname.h"
#include "tref.h"
#include "trpc.h"
SCatalogMgmt gCtgMgmt = {0};
int32_t ctgRemoveTbMetaFromCache(SCatalog* pCtg, SName* pTableName, bool syncReq) {
int32_t code = 0;
STableMeta* tblMeta = NULL;
@ -53,7 +52,8 @@ _return:
CTG_RET(code);
}
int32_t ctgGetDBVgInfo(SCatalog* pCtg, void *pTrans, const SEpSet* pMgmtEps, const char* dbFName, SCtgDBCache** dbCache, SDBVgInfo **pInfo) {
int32_t ctgGetDBVgInfo(SCatalog* pCtg, void* pTrans, const SEpSet* pMgmtEps, const char* dbFName, SCtgDBCache** dbCache,
SDBVgInfo** pInfo) {
int32_t code = 0;
CTG_ERR_RET(ctgAcquireVgInfoFromCache(pCtg, dbFName, dbCache));
@ -119,8 +119,6 @@ int32_t ctgRefreshDBVgInfo(SCatalog* pCtg, void *pTrans, const SEpSet* pMgmtEps,
return TSDB_CODE_SUCCESS;
}
int32_t ctgRefreshTbMeta(CTG_PARAMS, SCtgTbMetaCtx* ctx, STableMetaOutput** pOutput, bool syncReq) {
SVgroupInfo vgroupInfo = {0};
int32_t code = 0;
@ -139,7 +137,8 @@ int32_t ctgRefreshTbMeta(CTG_PARAMS, SCtgTbMetaCtx* ctx, STableMetaOutput **pOut
if (CTG_FLAG_IS_SYS_DB(ctx->flag)) {
ctgDebug("will refresh tbmeta, supposed in information_schema, tbName:%s", tNameGetTableName(ctx->pName));
CTG_ERR_JRET(ctgGetTbMetaFromMnodeImpl(CTG_PARAMS_LIST(), (char *)ctx->pName->dbname, (char *)ctx->pName->tname, output, NULL));
CTG_ERR_JRET(ctgGetTbMetaFromMnodeImpl(CTG_PARAMS_LIST(), (char*)ctx->pName->dbname, (char*)ctx->pName->tname,
output, NULL));
} else if (CTG_FLAG_IS_STB(ctx->flag)) {
ctgDebug("will refresh tbmeta, supposed to be stb, tbName:%s", tNameGetTableName(ctx->pName));
@ -150,7 +149,8 @@ int32_t ctgRefreshTbMeta(CTG_PARAMS, SCtgTbMetaCtx* ctx, STableMetaOutput **pOut
CTG_ERR_JRET(ctgGetTbMetaFromVnode(CTG_PARAMS_LIST(), ctx->pName, &vgroupInfo, output, NULL));
}
} else {
ctgDebug("will refresh tbmeta, not supposed to be stb, tbName:%s, flag:%d", tNameGetTableName(ctx->pName), ctx->flag);
ctgDebug("will refresh tbmeta, not supposed to be stb, tbName:%s, flag:%d", tNameGetTableName(ctx->pName),
ctx->flag);
// if get from vnode failed or no table meta, will not try mnode
CTG_ERR_JRET(ctgGetTbMetaFromVnode(CTG_PARAMS_LIST(), ctx->pName, &vgroupInfo, output, NULL));
@ -192,9 +192,11 @@ int32_t ctgRefreshTbMeta(CTG_PARAMS, SCtgTbMetaCtx* ctx, STableMetaOutput **pOut
}
if (CTG_IS_META_TABLE(output->metaType)) {
ctgDebug("tbmeta got, dbFName:%s, tbName:%s, tbType:%d", output->dbFName, output->tbName, output->tbMeta->tableType);
ctgDebug("tbmeta got, dbFName:%s, tbName:%s, tbType:%d", output->dbFName, output->tbName,
output->tbMeta->tableType);
} else {
ctgDebug("tbmeta got, dbFName:%s, tbName:%s, tbType:%d, stbMetaGot:%d", output->dbFName, output->ctbName, output->ctbMeta.tableType, CTG_IS_META_BOTH(output->metaType));
ctgDebug("tbmeta got, dbFName:%s, tbName:%s, tbType:%d, stbMetaGot:%d", output->dbFName, output->ctbName,
output->ctbMeta.tableType, CTG_IS_META_BOTH(output->metaType));
}
if (pOutput) {
@ -221,7 +223,8 @@ int32_t ctgGetTbMetaFromCache(CTG_PARAMS, SCtgTbMetaCtx* ctx, STableMeta** pTabl
CTG_ERR_RET(ctgReadTbMetaFromCache(pCtg, ctx, pTableMeta));
if (*pTableMeta) {
if (CTG_FLAG_MATCH_STB(ctx->flag, (*pTableMeta)->tableType) && ((!CTG_FLAG_IS_FORCE_UPDATE(ctx->flag)) || (CTG_FLAG_IS_SYS_DB(ctx->flag)))) {
if (CTG_FLAG_MATCH_STB(ctx->flag, (*pTableMeta)->tableType) &&
((!CTG_FLAG_IS_FORCE_UPDATE(ctx->flag)) || (CTG_FLAG_IS_SYS_DB(ctx->flag)))) {
return TSDB_CODE_SUCCESS;
}
@ -235,7 +238,6 @@ int32_t ctgGetTbMetaFromCache(CTG_PARAMS, SCtgTbMetaCtx* ctx, STableMeta** pTabl
return TSDB_CODE_SUCCESS;
}
int32_t ctgGetTbMeta(CTG_PARAMS, SCtgTbMetaCtx* ctx, STableMeta** pTableMeta) {
int32_t code = 0;
STableMetaOutput* output = NULL;
@ -344,8 +346,8 @@ _return:
CTG_RET(code);
}
int32_t ctgChkAuth(SCatalog* pCtg, void *pTrans, const SEpSet* pMgmtEps, const char* user, const char* dbFName, AUTH_TYPE type, bool *pass) {
int32_t ctgChkAuth(SCatalog* pCtg, void* pTrans, const SEpSet* pMgmtEps, const char* user, const char* dbFName,
AUTH_TYPE type, bool* pass) {
bool inCache = false;
int32_t code = 0;
@ -495,7 +497,8 @@ int32_t catalogInit(SCatalogCfg *cfg) {
gCtgMgmt.cfg.stbRentSec = CTG_DEFAULT_RENT_SECOND;
}
gCtgMgmt.pCluster = taosHashInit(CTG_DEFAULT_CACHE_CLUSTER_NUMBER, taosGetDefaultHashFunction(TSDB_DATA_TYPE_BIGINT), false, HASH_ENTRY_LOCK);
gCtgMgmt.pCluster = taosHashInit(CTG_DEFAULT_CACHE_CLUSTER_NUMBER, taosGetDefaultHashFunction(TSDB_DATA_TYPE_BIGINT),
false, HASH_ENTRY_LOCK);
if (NULL == gCtgMgmt.pCluster) {
qError("taosHashInit %d cluster cache failed", CTG_DEFAULT_CACHE_CLUSTER_NUMBER);
CTG_ERR_RET(TSDB_CODE_CTG_INTERNAL_ERROR);
@ -521,7 +524,8 @@ int32_t catalogInit(SCatalogCfg *cfg) {
CTG_ERR_RET(ctgStartUpdateThread());
qDebug("catalog initialized, maxDb:%u, maxTbl:%u, dbRentSec:%u, stbRentSec:%u", gCtgMgmt.cfg.maxDBCacheNum, gCtgMgmt.cfg.maxTblCacheNum, gCtgMgmt.cfg.dbRentSec, gCtgMgmt.cfg.stbRentSec);
qDebug("catalog initialized, maxDb:%u, maxTbl:%u, dbRentSec:%u, stbRentSec:%u", gCtgMgmt.cfg.maxDBCacheNum,
gCtgMgmt.cfg.maxTblCacheNum, gCtgMgmt.cfg.dbRentSec, gCtgMgmt.cfg.stbRentSec);
return TSDB_CODE_SUCCESS;
}
@ -559,7 +563,8 @@ int32_t catalogGetHandle(uint64_t clusterId, SCatalog** catalogHandle) {
CTG_ERR_JRET(ctgMetaRentInit(&clusterCtg->dbRent, gCtgMgmt.cfg.dbRentSec, CTG_RENT_DB));
CTG_ERR_JRET(ctgMetaRentInit(&clusterCtg->stbRent, gCtgMgmt.cfg.stbRentSec, CTG_RENT_STABLE));
clusterCtg->dbCache = taosHashInit(gCtgMgmt.cfg.maxDBCacheNum, taosGetDefaultHashFunction(TSDB_DATA_TYPE_BINARY), false, HASH_ENTRY_LOCK);
clusterCtg->dbCache = taosHashInit(gCtgMgmt.cfg.maxDBCacheNum, taosGetDefaultHashFunction(TSDB_DATA_TYPE_BINARY),
false, HASH_ENTRY_LOCK);
if (NULL == clusterCtg->dbCache) {
qError("taosHashInit %d dbCache failed", CTG_DEFAULT_CACHE_DB_NUMBER);
CTG_ERR_JRET(TSDB_CODE_CTG_MEM_ERROR);
@ -645,7 +650,8 @@ _return:
CTG_API_LEAVE(code);
}
int32_t catalogGetDBVgInfo(SCatalog* pCtg, void *pTrans, const SEpSet* pMgmtEps, const char* dbFName, SArray** vgroupList) {
int32_t catalogGetDBVgInfo(SCatalog* pCtg, void* pTrans, const SEpSet* pMgmtEps, const char* dbFName,
SArray** vgroupList) {
CTG_API_ENTER();
if (NULL == pCtg || NULL == dbFName || NULL == pTrans || NULL == pMgmtEps || NULL == vgroupList) {
@ -684,7 +690,6 @@ _return:
CTG_API_LEAVE(code);
}
int32_t catalogUpdateDBVgInfo(SCatalog* pCtg, const char* dbFName, uint64_t dbId, SDBVgInfo* dbInfo) {
CTG_API_ENTER();
@ -702,7 +707,6 @@ _return:
CTG_API_LEAVE(code);
}
int32_t catalogRemoveDB(SCatalog* pCtg, const char* dbFName, uint64_t dbId) {
CTG_API_ENTER();
@ -761,7 +765,6 @@ _return:
CTG_API_LEAVE(code);
}
int32_t catalogRemoveStbMeta(SCatalog* pCtg, const char* dbFName, uint64_t dbId, const char* stbName, uint64_t suid) {
CTG_API_ENTER();
@ -784,7 +787,8 @@ _return:
CTG_API_LEAVE(code);
}
int32_t catalogGetTableMeta(SCatalog* pCtg, void *pTrans, const SEpSet* pMgmtEps, const SName* pTableName, STableMeta** pTableMeta) {
int32_t catalogGetTableMeta(SCatalog* pCtg, void* pTrans, const SEpSet* pMgmtEps, const SName* pTableName,
STableMeta** pTableMeta) {
CTG_API_ENTER();
SCtgTbMetaCtx ctx = {0};
@ -794,7 +798,8 @@ int32_t catalogGetTableMeta(SCatalog* pCtg, void *pTrans, const SEpSet* pMgmtEps
CTG_API_LEAVE(ctgGetTbMeta(pCtg, pTrans, pMgmtEps, &ctx, pTableMeta));
}
int32_t catalogGetSTableMeta(SCatalog* pCtg, void * pTrans, const SEpSet* pMgmtEps, const SName* pTableName, STableMeta** pTableMeta) {
int32_t catalogGetSTableMeta(SCatalog* pCtg, void* pTrans, const SEpSet* pMgmtEps, const SName* pTableName,
STableMeta** pTableMeta) {
CTG_API_ENTER();
SCtgTbMetaCtx ctx = {0};
@ -868,7 +873,6 @@ int32_t catalogChkTbMetaVersion(SCatalog* pCtg, void *pTrans, const SEpSet* pMgm
CTG_API_LEAVE(TSDB_CODE_SUCCESS);
}
int32_t catalogRefreshDBVgInfo(SCatalog* pCtg, void* pTrans, const SEpSet* pMgmtEps, const char* dbFName) {
CTG_API_ENTER();
@ -879,7 +883,8 @@ int32_t catalogRefreshDBVgInfo(SCatalog* pCtg, void *pTrans, const SEpSet* pMgmt
CTG_API_LEAVE(ctgRefreshDBVgInfo(pCtg, pTrans, pMgmtEps, dbFName));
}
int32_t catalogRefreshTableMeta(SCatalog* pCtg, void *pTrans, const SEpSet* pMgmtEps, const SName* pTableName, int32_t isSTable) {
int32_t catalogRefreshTableMeta(SCatalog* pCtg, void* pTrans, const SEpSet* pMgmtEps, const SName* pTableName,
int32_t isSTable) {
CTG_API_ENTER();
if (NULL == pCtg || NULL == pTrans || NULL == pMgmtEps || NULL == pTableName) {
@ -893,7 +898,8 @@ int32_t catalogRefreshTableMeta(SCatalog* pCtg, void *pTrans, const SEpSet* pMgm
CTG_API_LEAVE(ctgRefreshTbMeta(CTG_PARAMS_LIST(), &ctx, NULL, true));
}
int32_t catalogRefreshGetTableMeta(SCatalog* pCtg, void *pTrans, const SEpSet* pMgmtEps, const SName* pTableName, STableMeta** pTableMeta, int32_t isSTable) {
int32_t catalogRefreshGetTableMeta(SCatalog* pCtg, void* pTrans, const SEpSet* pMgmtEps, const SName* pTableName,
STableMeta** pTableMeta, int32_t isSTable) {
CTG_API_ENTER();
SCtgTbMetaCtx ctx = {0};
@ -903,7 +909,8 @@ int32_t catalogRefreshGetTableMeta(SCatalog* pCtg, void *pTrans, const SEpSet* p
CTG_API_LEAVE(ctgGetTbMeta(CTG_PARAMS_LIST(), &ctx, pTableMeta));
}
int32_t catalogGetTableDistVgInfo(SCatalog* pCtg, void *pTrans, const SEpSet* pMgmtEps, const SName* pTableName, SArray** pVgList) {
int32_t catalogGetTableDistVgInfo(SCatalog* pCtg, void* pTrans, const SEpSet* pMgmtEps, const SName* pTableName,
SArray** pVgList) {
CTG_API_ENTER();
if (NULL == pCtg || NULL == pTrans || NULL == pMgmtEps || NULL == pTableName || NULL == pVgList) {
@ -918,8 +925,8 @@ int32_t catalogGetTableDistVgInfo(SCatalog* pCtg, void *pTrans, const SEpSet* pM
CTG_API_LEAVE(ctgGetTbDistVgInfo(pCtg, pTrans, pMgmtEps, (SName*)pTableName, pVgList));
}
int32_t catalogGetTableHashVgroup(SCatalog *pCtg, void *pTrans, const SEpSet *pMgmtEps, const SName *pTableName, SVgroupInfo *pVgroup) {
int32_t catalogGetTableHashVgroup(SCatalog* pCtg, void* pTrans, const SEpSet* pMgmtEps, const SName* pTableName,
SVgroupInfo* pVgroup) {
CTG_API_ENTER();
if (CTG_IS_SYS_DBNAME(pTableName->dbname)) {
@ -952,8 +959,8 @@ _return:
CTG_API_LEAVE(code);
}
int32_t catalogGetAllMeta(SCatalog* pCtg, void *pTrans, const SEpSet* pMgmtEps, const SCatalogReq* pReq, SMetaData* pRsp) {
int32_t catalogGetAllMeta(SCatalog* pCtg, void* pTrans, const SEpSet* pMgmtEps, const SCatalogReq* pReq,
SMetaData* pRsp) {
CTG_API_ENTER();
if (NULL == pCtg || NULL == pTrans || NULL == pMgmtEps || NULL == pReq || NULL == pRsp) {
@ -1016,7 +1023,8 @@ _return:
CTG_API_LEAVE(code);
}
int32_t catalogAsyncGetAllMeta(SCatalog* pCtg, void *pTrans, const SEpSet* pMgmtEps, uint64_t reqId, const SCatalogReq* pReq, catalogCallback fp, void* param, int64_t* jobId) {
int32_t catalogAsyncGetAllMeta(SCatalog* pCtg, void* pTrans, const SEpSet* pMgmtEps, uint64_t reqId,
const SCatalogReq* pReq, catalogCallback fp, void* param, int64_t* jobId) {
CTG_API_ENTER();
if (NULL == pCtg || NULL == pTrans || NULL == pMgmtEps || NULL == pReq || NULL == fp || NULL == param) {
@ -1115,7 +1123,6 @@ int32_t catalogGetExpiredUsers(SCatalog* pCtg, SUserAuthVersion **users, uint32_
CTG_API_LEAVE(TSDB_CODE_SUCCESS);
}
int32_t catalogGetDBCfg(SCatalog* pCtg, void* pTrans, const SEpSet* pMgmtEps, const char* dbFName, SDbCfgInfo* pDbCfg) {
CTG_API_ENTER();
@ -1126,7 +1133,8 @@ int32_t catalogGetDBCfg(SCatalog* pCtg, void *pTrans, const SEpSet* pMgmtEps, co
CTG_API_LEAVE(ctgGetDBCfgFromMnode(CTG_PARAMS_LIST(), dbFName, pDbCfg, NULL));
}
int32_t catalogGetIndexMeta(SCatalog* pCtg, void *pTrans, const SEpSet* pMgmtEps, const char* indexName, SIndexInfo* pInfo) {
int32_t catalogGetIndexMeta(SCatalog* pCtg, void* pTrans, const SEpSet* pMgmtEps, const char* indexName,
SIndexInfo* pInfo) {
CTG_API_ENTER();
if (NULL == pCtg || NULL == pTrans || NULL == pMgmtEps || NULL == indexName || NULL == pInfo) {
@ -1136,7 +1144,8 @@ int32_t catalogGetIndexMeta(SCatalog* pCtg, void *pTrans, const SEpSet* pMgmtEps
CTG_API_LEAVE(ctgGetIndexInfoFromMnode(CTG_PARAMS_LIST(), indexName, pInfo, NULL));
}
int32_t catalogGetTableIndex(SCatalog* pCtg, void *pTrans, const SEpSet* pMgmtEps, const SName* pTableName, SArray** pRes) {
int32_t catalogGetTableIndex(SCatalog* pCtg, void* pTrans, const SEpSet* pMgmtEps, const SName* pTableName,
SArray** pRes) {
CTG_API_ENTER();
if (NULL == pCtg || NULL == pTrans || NULL == pMgmtEps || NULL == pTableName || NULL == pRes) {
@ -1146,8 +1155,8 @@ int32_t catalogGetTableIndex(SCatalog* pCtg, void *pTrans, const SEpSet* pMgmtEp
CTG_API_LEAVE(ctgGetTbIndexFromMnode(CTG_PARAMS_LIST(), (SName*)pTableName, pRes, NULL));
}
int32_t catalogGetUdfInfo(SCatalog* pCtg, void *pTrans, const SEpSet* pMgmtEps, const char* funcName, SFuncInfo* pInfo) {
int32_t catalogGetUdfInfo(SCatalog* pCtg, void* pTrans, const SEpSet* pMgmtEps, const char* funcName,
SFuncInfo* pInfo) {
CTG_API_ENTER();
if (NULL == pCtg || NULL == pTrans || NULL == pMgmtEps || NULL == funcName || NULL == pInfo) {
@ -1162,7 +1171,8 @@ _return:
CTG_API_LEAVE(code);
}
int32_t catalogChkAuth(SCatalog* pCtg, void *pTrans, const SEpSet* pMgmtEps, const char* user, const char* dbFName, AUTH_TYPE type, bool *pass) {
int32_t catalogChkAuth(SCatalog* pCtg, void* pTrans, const SEpSet* pMgmtEps, const char* user, const char* dbFName,
AUTH_TYPE type, bool* pass) {
CTG_API_ENTER();
if (NULL == pCtg || NULL == pTrans || NULL == pMgmtEps || NULL == user || NULL == dbFName || NULL == pass) {
@ -1187,7 +1197,6 @@ int32_t catalogUpdateUserAuthInfo(SCatalog* pCtg, SGetUserAuthRsp* pAuth) {
CTG_API_LEAVE(ctgUpdateUserEnqueue(pCtg, pAuth, false));
}
void catalogDestroy(void) {
qInfo("start to destroy catalog");
@ -1226,6 +1235,3 @@ void catalogDestroy(void) {
qInfo("catalog destroyed");
}

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

@ -525,7 +525,7 @@ int32_t ctgDumpTbIndexRes(SCtgTask* pTask) {
}
SMetaRes res = {.code = pTask->code, .pRes = pTask->res};
taosArrayPush(pJob->jobRes.pTableHash, &res);
taosArrayPush(pJob->jobRes.pTableIndex, &res);
return TSDB_CODE_SUCCESS;
}
@ -875,7 +875,9 @@ int32_t ctgHandleGetTbIndexRsp(SCtgTask* pTask, int32_t reqType, const SDataBuf
TSWAP(pTask->res, pTask->msgCtx.out);
_return:
if (TSDB_CODE_MND_DB_INDEX_NOT_EXIST == code) {
code = TSDB_CODE_SUCCESS;
}
ctgHandleTaskEnd(pTask, code);
CTG_RET(code);

11
source/libs/executor/inc/executorimpl.h
View File

@ -458,7 +458,6 @@ typedef struct STimeWindowSupp {
int64_t waterMark;
TSKEY maxTs;
SColumnInfoData timeWindowData; // query time window info for scalar function execution.
SHashObj *winMap;
} STimeWindowAggSupp;
typedef struct SIntervalAggOperatorInfo {
@ -809,6 +808,11 @@ SOperatorInfo* createSortedMergeOperatorInfo(SOperatorInfo** downstream, int32_t
SOperatorInfo* createIntervalOperatorInfo(SOperatorInfo* downstream, SExprInfo* pExprInfo, int32_t numOfCols,
SSDataBlock* pResBlock, SInterval* pInterval, int32_t primaryTsSlotId,
STimeWindowAggSupp *pTwAggSupp, SExecTaskInfo* pTaskInfo, bool isStream);
SOperatorInfo* createMergeIntervalOperatorInfo(SOperatorInfo* downstream, SExprInfo* pExprInfo, int32_t numOfCols,
SSDataBlock* pResBlock, SInterval* pInterval, int32_t primaryTsSlotId,
SExecTaskInfo* pTaskInfo);
SOperatorInfo* createStreamFinalIntervalOperatorInfo(SOperatorInfo* downstream,
SPhysiNode* pPhyNode, SExecTaskInfo* pTaskInfo, int32_t numOfChild);
SOperatorInfo* createStreamIntervalOperatorInfo(SOperatorInfo* downstream, SExprInfo* pExprInfo, int32_t numOfCols,
@ -903,10 +907,11 @@ SResultWindowInfo* getSessionTimeWindow(SArray* pWinInfos, TSKEY ts, int64_t gap
int32_t updateSessionWindowInfo(SResultWindowInfo* pWinInfo, TSKEY* pTs, int32_t rows,
int32_t start, int64_t gap, SHashObj* pStDeleted);
bool functionNeedToExecute(SqlFunctionCtx* pCtx);
int64_t getSmaWaterMark(int64_t interval, double filesFactor);
bool isSmaStream(int8_t triggerType);
int32_t compareTimeWindow(const void* p1, const void* p2, const void* param);
int32_t finalizeResultRowIntoResultDataBlock(SDiskbasedBuf* pBuf, SResultRowPosition* resultRowPosition,
SqlFunctionCtx* pCtx, SExprInfo* pExprInfo, int32_t numOfExprs, const int32_t* rowCellOffset,
SSDataBlock* pBlock, SExecTaskInfo* pTaskInfo);
#ifdef __cplusplus
}

146
source/libs/executor/src/executorimpl.c
View File

@ -1819,7 +1819,7 @@ void setResultRowInitCtx(SResultRow* pResult, SqlFunctionCtx* pCtx, int32_t numO
}
}
static void extractQualifiedTupleByFilterResult(SSDataBlock* pBlock, const int8_t* rowRes, bool keep, bool needFree);
static void extractQualifiedTupleByFilterResult(SSDataBlock* pBlock, const int8_t* rowRes, bool keep);
void doFilter(const SNode* pFilterNode, SSDataBlock* pBlock, bool needFree) {
if (pFilterNode == NULL) {
@ -1840,30 +1840,29 @@ void doFilter(const SNode* pFilterNode, SSDataBlock* pBlock, bool needFree) {
bool keep = filterExecute(filter, pBlock, &rowRes, NULL, param1.numOfCols);
filterFreeInfo(filter);
extractQualifiedTupleByFilterResult(pBlock, rowRes, keep, needFree);
extractQualifiedTupleByFilterResult(pBlock, rowRes, keep);
blockDataUpdateTsWindow(pBlock, 0);
}
void extractQualifiedTupleByFilterResult(SSDataBlock* pBlock, const int8_t* rowRes, bool keep, bool needFree) {
void extractQualifiedTupleByFilterResult(SSDataBlock* pBlock, const int8_t* rowRes, bool keep) {
if (keep) {
return;
}
if (rowRes != NULL) {
SSDataBlock* px = createOneDataBlock(pBlock, false);
blockDataEnsureCapacity(px, pBlock->info.rows);
SSDataBlock* px = createOneDataBlock(pBlock, true);
int32_t totalRows = pBlock->info.rows;
for (int32_t i = 0; i < pBlock->info.numOfCols; ++i) {
SColumnInfoData* pDst = taosArrayGet(px->pDataBlock, i);
SColumnInfoData* pSrc = taosArrayGet(pBlock->pDataBlock, i);
SColumnInfoData* pSrc = taosArrayGet(px->pDataBlock, i);
SColumnInfoData* pDst = taosArrayGet(pBlock->pDataBlock, i);
// it is a reserved column for scalar function, and no data in this column yet.
if (pSrc->pData == NULL) {
if (pDst->pData == NULL) {
continue;
}
colInfoDataCleanup(pDst, pBlock->info.rows);
int32_t numOfRows = 0;
for (int32_t j = 0; j < totalRows; ++j) {
if (rowRes[j] == 0) {
@ -1883,20 +1882,8 @@ void extractQualifiedTupleByFilterResult(SSDataBlock* pBlock, const int8_t* rowR
} else {
ASSERT(pBlock->info.rows == numOfRows);
}
}
SColumnInfoData tmp = *pSrc;
*pSrc = *pDst;
*pDst = tmp;
if (!needFree) {
if (IS_VAR_DATA_TYPE(pDst->info.type)) { // this elements do not need free
pDst->varmeta.offset = NULL;
} else {
pDst->nullbitmap = NULL;
}
pDst->pData = NULL;
}
}
blockDataDestroy(px); // fix memory leak
} else {
// do nothing
@ -1956,6 +1943,57 @@ static void doUpdateNumOfRows(SResultRow* pRow, int32_t numOfExprs, const int32_
}
}
int32_t finalizeResultRowIntoResultDataBlock(SDiskbasedBuf* pBuf, SResultRowPosition* resultRowPosition,
SqlFunctionCtx* pCtx, SExprInfo* pExprInfo, int32_t numOfExprs,
const int32_t* rowCellOffset, SSDataBlock* pBlock,
SExecTaskInfo* pTaskInfo) {
SFilePage* page = getBufPage(pBuf, resultRowPosition->pageId);
SResultRow* pRow = (SResultRow*)((char*)page + resultRowPosition->offset);
doUpdateNumOfRows(pRow, numOfExprs, rowCellOffset);
if (pRow->numOfRows == 0) {
releaseBufPage(pBuf, page);
return 0;
}
while (pBlock->info.rows + pRow->numOfRows > pBlock->info.capacity) {
int32_t code = blockDataEnsureCapacity(pBlock, pBlock->info.capacity * 1.25);
if (TAOS_FAILED(code)) {
releaseBufPage(pBuf, page);
qError("%s ensure result data capacity failed, code %s", GET_TASKID(pTaskInfo), tstrerror(code));
longjmp(pTaskInfo->env, code);
}
}
for (int32_t j = 0; j < numOfExprs; ++j) {
int32_t slotId = pExprInfo[j].base.resSchema.slotId;
pCtx[j].resultInfo = getResultCell(pRow, j, rowCellOffset);
if (pCtx[j].fpSet.finalize) {
int32_t code = pCtx[j].fpSet.finalize(&pCtx[j], pBlock);
if (TAOS_FAILED(code)) {
qError("%s build result data block error, code %s", GET_TASKID(pTaskInfo), tstrerror(code));
longjmp(pTaskInfo->env, code);
}
} else if (strcmp(pCtx[j].pExpr->pExpr->_function.functionName, "_select_value") == 0) {
// do nothing, todo refactor
} else {
// expand the result into multiple rows. E.g., _wstartts, top(k, 20)
// the _wstartts needs to copy to 20 following rows, since the results of top-k expands to 20 different rows.
SColumnInfoData* pColInfoData = taosArrayGet(pBlock->pDataBlock, slotId);
char* in = GET_ROWCELL_INTERBUF(pCtx[j].resultInfo);
for (int32_t k = 0; k < pRow->numOfRows; ++k) {
colDataAppend(pColInfoData, pBlock->info.rows + k, in, pCtx[j].resultInfo->isNullRes);
}
}
}
releaseBufPage(pBuf, page);
pBlock->info.rows += pRow->numOfRows;
return 0;
}
int32_t doCopyToSDataBlock(SExecTaskInfo* pTaskInfo, SSDataBlock* pBlock, SExprInfo* pExprInfo, SDiskbasedBuf* pBuf,
SGroupResInfo* pGroupResInfo, const int32_t* rowCellOffset, SqlFunctionCtx* pCtx,
int32_t numOfExprs) {
@ -2086,11 +2124,6 @@ static int32_t compressQueryColData(SColumnInfoData* pColRes, int32_t numOfRows,
}
int32_t doFillTimeIntervalGapsInResults(struct SFillInfo* pFillInfo, SSDataBlock* pBlock, int32_t capacity) {
// for(int32_t i = 0; i < pFillInfo->numOfCols; ++i) {
// SColumnInfoData* pColInfoData = taosArrayGet(pOutput->pDataBlock, i);
// p[i] = pColInfoData->pData + (pColInfoData->info.bytes * pOutput->info.rows);
// }
int32_t numOfRows = (int32_t)taosFillResultDataBlock(pFillInfo, pBlock, capacity - pBlock->info.rows);
pBlock->info.rows += numOfRows;
@ -3112,8 +3145,8 @@ static SSDataBlock* doMerge(SOperatorInfo* pOperator) {
return (pInfo->binfo.pRes->info.rows > 0) ? pInfo->binfo.pRes : NULL;
}
SSDataBlock* getSortedMergeBlockData(SSortHandle* pHandle, SSDataBlock* pDataBlock, int32_t capacity, SArray* pColMatchInfo,
SSortedMergeOperatorInfo *pInfo) {
SSDataBlock* getSortedMergeBlockData(SSortHandle* pHandle, SSDataBlock* pDataBlock, int32_t capacity,
SArray* pColMatchInfo, SSortedMergeOperatorInfo* pInfo) {
blockDataCleanup(pDataBlock);
SSDataBlock* p = tsortGetSortedDataBlock(pHandle);
@ -4671,28 +4704,35 @@ SOperatorInfo* createOperatorTree(SPhysiNode* pPhyNode, SExecTaskInfo* pTaskInfo
.waterMark = pIntervalPhyNode->window.watermark,
.calTrigger = pIntervalPhyNode->window.triggerType,
.maxTs = INT64_MIN,
.winMap = NULL,
};
if (isSmaStream(pIntervalPhyNode->window.triggerType)) {
if (FLT_LESS(pIntervalPhyNode->window.filesFactor, 1.000000)) {
as.calTrigger = STREAM_TRIGGER_AT_ONCE_SMA;
} else {
_hash_fn_t hashFn = taosGetDefaultHashFunction(TSDB_DATA_TYPE_TIMESTAMP);
as.winMap = taosHashInit(64, hashFn, true, HASH_NO_LOCK);
as.waterMark = getSmaWaterMark(interval.interval, pIntervalPhyNode->window.filesFactor);
as.calTrigger = STREAM_TRIGGER_WINDOW_CLOSE_SMA;
}
}
int32_t tsSlotId = ((SColumnNode*)pIntervalPhyNode->window.pTspk)->slotId;
bool isStream = (QUERY_NODE_PHYSICAL_PLAN_STREAM_INTERVAL == type);
pOptr = createIntervalOperatorInfo(ops[0], pExprInfo, num, pResBlock, &interval, tsSlotId, &as, pTaskInfo, isStream);
pOptr =
createIntervalOperatorInfo(ops[0], pExprInfo, num, pResBlock, &interval, tsSlotId, &as, pTaskInfo, isStream);
} else if (QUERY_NODE_PHYSICAL_PLAN_MERGE_INTERVAL == type) {
SMergeIntervalPhysiNode * pIntervalPhyNode = (SMergeIntervalPhysiNode*)pPhyNode;
SExprInfo* pExprInfo = createExprInfo(pIntervalPhyNode->window.pFuncs, NULL, &num);
SSDataBlock* pResBlock = createResDataBlock(pPhyNode->pOutputDataBlockDesc);
SInterval interval = {.interval = pIntervalPhyNode->interval,
.sliding = pIntervalPhyNode->sliding,
.intervalUnit = pIntervalPhyNode->intervalUnit,
.slidingUnit = pIntervalPhyNode->slidingUnit,
.offset = pIntervalPhyNode->offset,
.precision = ((SColumnNode*)pIntervalPhyNode->window.pTspk)->node.resType.precision};
int32_t tsSlotId = ((SColumnNode*)pIntervalPhyNode->window.pTspk)->slotId;
pOptr = createMergeIntervalOperatorInfo(ops[0], pExprInfo, num, pResBlock, &interval, tsSlotId, pTaskInfo);
} else if (QUERY_NODE_PHYSICAL_PLAN_STREAM_SEMI_INTERVAL == type) {
int32_t children = 8;
qDebug("[******]create Semi");
int32_t children = 0;
pOptr = createStreamFinalIntervalOperatorInfo(ops[0], pPhyNode, pTaskInfo, children);
} else if (QUERY_NODE_PHYSICAL_PLAN_STREAM_FINAL_INTERVAL == type) {
int32_t children = 0;
qDebug("[******]create Final");
int32_t children = 1;
pOptr = createStreamFinalIntervalOperatorInfo(ops[0], pPhyNode, pTaskInfo, children);
} else if (QUERY_NODE_PHYSICAL_PLAN_SORT == type) {
SSortPhysiNode* pSortPhyNode = (SSortPhysiNode*)pPhyNode;
@ -5067,7 +5107,7 @@ tsdbReaderT doCreateDataReader(STableScanPhysiNode* pTableScanNode, SReadHandle*
goto _error;
}
return tsdbQueryTables(pHandle->vnode, &cond, pTableListInfo, queryId, taskId);
return tsdbReaderOpen(pHandle->vnode, &cond, pTableListInfo, queryId, taskId);
_error:
terrno = code;
@ -5349,7 +5389,8 @@ int32_t getOperatorExplainExecInfo(SOperatorInfo* operatorInfo, SExplainExecInfo
return TSDB_CODE_SUCCESS;
}
int32_t initStreamAggSupporter(SStreamAggSupporter* pSup, const char* pKey, SqlFunctionCtx* pCtx, int32_t numOfOutput, size_t size) {
int32_t initStreamAggSupporter(SStreamAggSupporter* pSup, const char* pKey, SqlFunctionCtx* pCtx, int32_t numOfOutput,
size_t size) {
pSup->resultRowSize = getResultRowSize(pCtx, numOfOutput);
pSup->keySize = sizeof(int64_t) + sizeof(TSKEY);
pSup->pKeyBuf = taosMemoryCalloc(1, pSup->keySize);
@ -5376,16 +5417,3 @@ int32_t initStreamAggSupporter(SStreamAggSupporter* pSup, const char* pKey, SqlF
return code;
}
int64_t getSmaWaterMark(int64_t interval, double filesFactor) {
int64_t waterMark = 0;
ASSERT(FLT_GREATEREQUAL(filesFactor, 0.000000));
waterMark = -1 * filesFactor;
return waterMark;
}
bool isSmaStream(int8_t triggerType) {
if (triggerType == STREAM_TRIGGER_AT_ONCE || triggerType == STREAM_TRIGGER_WINDOW_CLOSE) {
return false;
}
return true;
}

19
source/libs/executor/src/scanoperator.c
View File

@ -809,11 +809,18 @@ static void getUpdateDataBlock(SStreamBlockScanInfo* pInfo, bool invertible, SSD
// return p;
SColumnInfoData* pCol = (SColumnInfoData*)taosArrayGet(pUpdateBlock->pDataBlock, pInfo->primaryTsIndex);
ASSERT(pCol->info.type == TSDB_DATA_TYPE_TIMESTAMP);
colInfoDataEnsureCapacity(pCol, 0, size);
blockDataEnsureCapacity(pUpdateBlock, size);
for (int32_t i = 0; i < size; i++) {
TSKEY* pTs = (TSKEY*)taosArrayGet(pInfo->tsArray, i);
colDataAppend(pCol, i, (char*)pTs, false);
}
for (int32_t i = 0; i < pUpdateBlock->info.numOfCols; i++) {
if (i == pInfo->primaryTsIndex) {
continue;
}
SColumnInfoData* pCol = (SColumnInfoData*)taosArrayGet(pUpdateBlock->pDataBlock, i);
colDataAppendNNULL(pCol, 0, size);
}
pUpdateBlock->info.rows = size;
pUpdateBlock->info.type = STREAM_REPROCESS;
blockDataUpdateTsWindow(pUpdateBlock, 0);
@ -841,7 +848,9 @@ static SSDataBlock* doStreamBlockScan(SOperatorInfo* pOperator) {
}
int32_t current = pInfo->validBlockIndex++;
return taosArrayGetP(pInfo->pBlockLists, current);
SSDataBlock* pBlock = taosArrayGetP(pInfo->pBlockLists, current);
blockDataUpdateTsWindow(pBlock, 0);
return pBlock;
} else {
if (pInfo->scanMode == STREAM_SCAN_FROM_RES) {
blockDataDestroy(pInfo->pUpdateRes);
@ -940,7 +949,7 @@ static SSDataBlock* doStreamBlockScan(SOperatorInfo* pOperator) {
}
doFilter(pInfo->pCondition, pInfo->pRes, false);
blockDataUpdateTsWindow(pInfo->pRes, 0);
blockDataUpdateTsWindow(pInfo->pRes, pInfo->primaryTsIndex);
break;
}
@ -1019,10 +1028,6 @@ SOperatorInfo* createStreamScanOperatorInfo(void* pDataReader, SReadHandle* pHan
goto _error;
}
if (isSmaStream(pTableScanNode->triggerType)) {
pTwSup->waterMark = getSmaWaterMark(pSTInfo->interval.interval, pTableScanNode->filesFactor);
}
if (pSTInfo->interval.interval > 0 && pDataReader) {
pInfo->pUpdateInfo = updateInfoInitP(&pSTInfo->interval, pTwSup->waterMark);
} else {

430
source/libs/executor/src/timewindowoperator.c
View File

@ -750,14 +750,15 @@ int64_t getReskey(void* data, int32_t index) {
return *(int64_t*)pos->key;
}
static int32_t saveResult(SResultRow* result, uint64_t groupId, SArray* pUpdated) {
static int32_t saveResult(int64_t ts, int32_t pageId, int32_t offset, uint64_t groupId,
SArray* pUpdated) {
int32_t size = taosArrayGetSize(pUpdated);
int32_t index = binarySearch(pUpdated, size, result->win.skey, TSDB_ORDER_DESC, getReskey);
int32_t index = binarySearch(pUpdated, size, ts, TSDB_ORDER_DESC, getReskey);
if (index == -1) {
index = 0;
} else {
TSKEY resTs = getReskey(pUpdated, index);
if (resTs < result->win.skey) {
if (resTs < ts) {
index++;
} else {
return TSDB_CODE_SUCCESS;
@ -769,14 +770,18 @@ static int32_t saveResult(SResultRow* result, uint64_t groupId, SArray* pUpdated
return TSDB_CODE_OUT_OF_MEMORY;
}
newPos->groupId = groupId;
newPos->pos = (SResultRowPosition){.pageId = result->pageId, .offset = result->offset};
*(int64_t*)newPos->key = result->win.skey;
newPos->pos = (SResultRowPosition){.pageId = pageId, .offset = offset};
*(int64_t*)newPos->key = ts;
if (taosArrayInsert(pUpdated, index, &newPos) == NULL) {
return TSDB_CODE_OUT_OF_MEMORY;
}
return TSDB_CODE_SUCCESS;
}
static int32_t saveResultRow(SResultRow* result, uint64_t groupId, SArray* pUpdated) {
return saveResult(result->win.skey, result->pageId, result->offset, groupId, pUpdated);
}
static void removeResult(SArray* pUpdated, TSKEY key) {
int32_t size = taosArrayGetSize(pUpdated);
int32_t index = binarySearch(pUpdated, size, key, TSDB_ORDER_DESC, getReskey);
@ -818,12 +823,8 @@ static void hashIntervalAgg(SOperatorInfo* pOperatorInfo, SResultRowInfo* pResul
}
if (pInfo->execModel == OPTR_EXEC_MODEL_STREAM) {
if (pInfo->twAggSup.calTrigger == STREAM_TRIGGER_AT_ONCE ||
pInfo->twAggSup.calTrigger == STREAM_TRIGGER_AT_ONCE_SMA) {
saveResult(pResult, tableGroupId, pUpdated);
}
if (pInfo->twAggSup.winMap) {
taosHashRemove(pInfo->twAggSup.winMap, &win.skey, sizeof(TSKEY));
if (pInfo->twAggSup.calTrigger == STREAM_TRIGGER_AT_ONCE) {
saveResultRow(pResult, tableGroupId, pUpdated);
}
}
@ -872,12 +873,8 @@ static void hashIntervalAgg(SOperatorInfo* pOperatorInfo, SResultRowInfo* pResul
}
if (pInfo->execModel == OPTR_EXEC_MODEL_STREAM) {
if (pInfo->twAggSup.calTrigger == STREAM_TRIGGER_AT_ONCE ||
pInfo->twAggSup.calTrigger == STREAM_TRIGGER_AT_ONCE_SMA) {
saveResult(pResult, tableGroupId, pUpdated);
}
if (pInfo->twAggSup.winMap) {
taosHashRemove(pInfo->twAggSup.winMap, &win.skey, sizeof(TSKEY));
if (pInfo->twAggSup.calTrigger == STREAM_TRIGGER_AT_ONCE) {
saveResultRow(pResult, tableGroupId, pUpdated);
}
}
@ -1251,6 +1248,23 @@ static void doClearWindows(SAggSupporter* pSup, SOptrBasicInfo* pBinfo, SInterva
}
}
static int32_t getAllIntervalWindow(SHashObj* pHashMap, SArray* resWins) {
void* pIte = NULL;
size_t keyLen = 0;
while ((pIte = taosHashIterate(pHashMap, pIte)) != NULL) {
void* key = taosHashGetKey(pIte, &keyLen);
uint64_t groupId = *(uint64_t*)key;
ASSERT(keyLen == GET_RES_WINDOW_KEY_LEN(sizeof(TSKEY)));
TSKEY ts = *(int64_t*)((char*)key + sizeof(uint64_t));
SResultRowPosition* pPos = (SResultRowPosition*)pIte;
int32_t code = saveResult(ts, pPos->pageId, pPos->offset, groupId, resWins);
if (code != TSDB_CODE_SUCCESS) {
return code;
}
}
return TSDB_CODE_SUCCESS;
}
static int32_t closeIntervalWindow(SHashObj* pHashMap, STimeWindowAggSupp* pSup, SInterval* pInterval,
SArray* closeWins) {
void* pIte = NULL;
@ -1264,28 +1278,16 @@ static int32_t closeIntervalWindow(SHashObj* pHashMap, STimeWindowAggSupp* pSup,
dumyInfo.cur.pageId = -1;
STimeWindow win = getActiveTimeWindow(NULL, &dumyInfo, ts, pInterval, pInterval->precision, NULL);
if (win.ekey < pSup->maxTs - pSup->waterMark) {
if (pSup->calTrigger == STREAM_TRIGGER_WINDOW_CLOSE_SMA) {
if (taosHashGet(pSup->winMap, &win.skey, sizeof(TSKEY))) {
continue;
}
}
char keyBuf[GET_RES_WINDOW_KEY_LEN(sizeof(TSKEY))];
SET_RES_WINDOW_KEY(keyBuf, &ts, sizeof(TSKEY), groupId);
if (pSup->calTrigger != STREAM_TRIGGER_AT_ONCE_SMA && pSup->calTrigger != STREAM_TRIGGER_WINDOW_CLOSE_SMA) {
taosHashRemove(pHashMap, keyBuf, keyLen);
SResultRowPosition* pPos = (SResultRowPosition*)pIte;
if (pSup->calTrigger == STREAM_TRIGGER_WINDOW_CLOSE) {
int32_t code = saveResult(ts, pPos->pageId, pPos->offset, groupId, closeWins);
if (code != TSDB_CODE_SUCCESS) {
return code;
}
SResKeyPos* pos = taosMemoryMalloc(sizeof(SResKeyPos) + sizeof(uint64_t));
if (pos == NULL) {
return TSDB_CODE_OUT_OF_MEMORY;
}
pos->groupId = groupId;
pos->pos = *(SResultRowPosition*)pIte;
*(int64_t*)pos->key = ts;
if (!taosArrayPush(closeWins, &pos)) {
taosMemoryFree(pos);
return TSDB_CODE_OUT_OF_MEMORY;
}
taosHashPut(pSup->winMap, &win.skey, sizeof(TSKEY), NULL, 0);
}
}
return TSDB_CODE_SUCCESS;
@ -1312,8 +1314,6 @@ static SSDataBlock* doStreamIntervalAgg(SOperatorInfo* pOperator) {
SOperatorInfo* downstream = pOperator->pDownstream[0];
SArray* pUpdated = taosArrayInit(4, POINTER_BYTES);
SArray* pClosed = taosArrayInit(4, POINTER_BYTES);
while (1) {
SSDataBlock* pBlock = downstream->fpSet.getNextFn(downstream);
if (pBlock == NULL) {
@ -1332,22 +1332,18 @@ static SSDataBlock* doStreamIntervalAgg(SOperatorInfo* pOperator) {
doClearWindows(&pInfo->aggSup, &pInfo->binfo, &pInfo->interval, 0, pOperator->numOfExprs, pBlock, NULL);
qDebug("%s clear existed time window results for updates checked", GET_TASKID(pTaskInfo));
continue;
} else if (pBlock->info.type == STREAM_GET_ALL &&
pInfo->twAggSup.calTrigger == STREAM_TRIGGER_MAX_DELAY) {
getAllIntervalWindow(pInfo->aggSup.pResultRowHashTable, pUpdated);
continue;
}
pInfo->twAggSup.maxTs = TMAX(pInfo->twAggSup.maxTs, pBlock->info.window.ekey);
hashIntervalAgg(pOperator, &pInfo->binfo.resultRowInfo, pBlock, MAIN_SCAN, pUpdated);
}
closeIntervalWindow(pInfo->aggSup.pResultRowHashTable, &pInfo->twAggSup, &pInfo->interval, pUpdated);
closeIntervalWindow(pInfo->aggSup.pResultRowHashTable, &pInfo->twAggSup, &pInfo->interval, pClosed);
finalizeUpdatedResult(pOperator->numOfExprs, pInfo->aggSup.pResultBuf, pClosed, pInfo->binfo.rowCellInfoOffset);
if (pInfo->twAggSup.calTrigger == STREAM_TRIGGER_WINDOW_CLOSE ||
pInfo->twAggSup.calTrigger == STREAM_TRIGGER_WINDOW_CLOSE_SMA) {
taosArrayAddAll(pUpdated, pClosed);
}
taosArrayDestroy(pClosed);
finalizeUpdatedResult(pOperator->numOfExprs, pInfo->aggSup.pResultBuf, pUpdated, pInfo->binfo.rowCellInfoOffset);
initMultiResInfoFromArrayList(&pInfo->groupResInfo, pUpdated);
blockDataEnsureCapacity(pInfo->binfo.pRes, pOperator->resultInfo.capacity);
doBuildResultDatablock(pOperator, &pInfo->binfo, &pInfo->groupResInfo, pInfo->aggSup.pResultBuf);
@ -1881,8 +1877,8 @@ _error:
return NULL;
}
static void doHashInterval(SOperatorInfo* pOperatorInfo, SSDataBlock* pSDataBlock,
int32_t tableGroupId, SArray* pUpdated) {
static void doHashInterval(SOperatorInfo* pOperatorInfo, SSDataBlock* pSDataBlock, int32_t tableGroupId,
SArray* pUpdated) {
SStreamFinalIntervalOperatorInfo* pInfo = (SStreamFinalIntervalOperatorInfo*)pOperatorInfo->info;
SResultRowInfo* pResultRowInfo = &(pInfo->binfo.resultRowInfo);
SExecTaskInfo* pTaskInfo = pOperatorInfo->pTaskInfo;
@ -1914,17 +1910,19 @@ static void doHashInterval(SOperatorInfo* pOperatorInfo, SSDataBlock* pSDataBloc
pos->groupId = tableGroupId;
pos->pos = (SResultRowPosition){.pageId = pResult->pageId, .offset = pResult->offset};
*(int64_t*)pos->key = pResult->win.skey;
forwardRows = getNumOfRowsInTimeWindow(&pSDataBlock->info, tsCols, startPos,
nextWin.ekey, binarySearchForKey, NULL, TSDB_ORDER_ASC);
forwardRows = getNumOfRowsInTimeWindow(&pSDataBlock->info, tsCols, startPos, nextWin.ekey, binarySearchForKey, NULL,
TSDB_ORDER_ASC);
if (pInfo->twAggSup.calTrigger == STREAM_TRIGGER_AT_ONCE && pUpdated) {
saveResult(pResult, tableGroupId, pUpdated);
saveResultRow(pResult, tableGroupId, pUpdated);
}
// window start(end) key interpolation
// doWindowBorderInterpolation(pInfo, pSDataBlock, numOfOutput, pInfo->binfo.pCtx, pResult, &nextWin, startPos, forwardRows);
// doWindowBorderInterpolation(pInfo, pSDataBlock, numOfOutput, pInfo->binfo.pCtx, pResult, &nextWin, startPos,
// forwardRows);
updateTimeWindowInfo(&pInfo->twAggSup.timeWindowData, &nextWin, true);
doApplyFunctions(pTaskInfo, pInfo->binfo.pCtx, &nextWin, &pInfo->twAggSup.timeWindowData, startPos, forwardRows,
tsCols, pSDataBlock->info.rows, numOfOutput, TSDB_ORDER_ASC);
int32_t prevEndPos = (forwardRows - 1) * step + startPos;
ASSERT(pSDataBlock->info.window.skey > 0 && pSDataBlock->info.window.ekey > 0);
startPos = getNextQualifiedWindow(&pInfo->interval, &nextWin, &pSDataBlock->info, tsCols, prevEndPos, pInfo->order);
if (startPos < 0) {
break;
@ -2003,17 +2001,13 @@ static void copyUpdateDataBlock(SSDataBlock* pDest, SSDataBlock* pSource, int32_
}
static int32_t getChildIndex(SSDataBlock* pBlock) {
// if (pBlock->info.type != STREAM_INVALID && pBlock->info.rows < 4) { // for test
// return pBlock->info.rows - 1;
// }
return 0;
return pBlock->info.childId;
}
static SSDataBlock* doStreamFinalIntervalAgg(SOperatorInfo* pOperator) {
SStreamFinalIntervalOperatorInfo* pInfo = pOperator->info;
SOperatorInfo* downstream = pOperator->pDownstream[0];
SArray* pUpdated = taosArrayInit(4, POINTER_BYTES);
SArray* pClosed = taosArrayInit(4, POINTER_BYTES);
if (pOperator->status == OP_EXEC_DONE) {
return NULL;
@ -2051,10 +2045,10 @@ static SSDataBlock* doStreamFinalIntervalAgg(SOperatorInfo* pOperator) {
int32_t childIndex = getChildIndex(pBlock);
SOperatorInfo* pChildOp = taosArrayGetP(pInfo->pChildren, childIndex);
SIntervalAggOperatorInfo* pChildInfo = pChildOp->info;
doClearWindows(&pChildInfo->aggSup, &pChildInfo->binfo, &pChildInfo->interval,
pChildInfo->primaryTsIndex, pChildOp->numOfExprs, pBlock, NULL);
rebuildIntervalWindow(pInfo, pUpWins, pInfo->binfo.pRes->info.groupId,
pOperator->numOfExprs, pOperator->pTaskInfo);
doClearWindows(&pChildInfo->aggSup, &pChildInfo->binfo, &pChildInfo->interval, pChildInfo->primaryTsIndex,
pChildOp->numOfExprs, pBlock, NULL);
rebuildIntervalWindow(pInfo, pUpWins, pInfo->binfo.pRes->info.groupId, pOperator->numOfExprs,
pOperator->pTaskInfo);
taosArrayDestroy(pUpWins);
continue;
}
@ -2062,7 +2056,12 @@ static SSDataBlock* doStreamFinalIntervalAgg(SOperatorInfo* pOperator) {
copyUpdateDataBlock(pInfo->pUpdateRes, pBlock, pInfo->primaryTsIndex);
taosArrayDestroy(pUpWins);
break;
} else if (pBlock->info.type == STREAM_GET_ALL && isFinalInterval(pInfo) &&
pInfo->twAggSup.calTrigger == STREAM_TRIGGER_MAX_DELAY) {
getAllIntervalWindow(pInfo->aggSup.pResultRowHashTable, pUpdated);
continue;
}
if (isFinalInterval(pInfo)) {
int32_t chIndex = getChildIndex(pBlock);
int32_t size = taosArrayGetSize(pInfo->pChildren);
@ -2084,15 +2083,8 @@ static SSDataBlock* doStreamFinalIntervalAgg(SOperatorInfo* pOperator) {
}
if (isFinalInterval(pInfo)) {
closeIntervalWindow(pInfo->aggSup.pResultRowHashTable, &pInfo->twAggSup,
&pInfo->interval, pClosed);
finalizeUpdatedResult(pOperator->numOfExprs, pInfo->aggSup.pResultBuf, pClosed,
pInfo->binfo.rowCellInfoOffset);
if (pInfo->twAggSup.calTrigger == STREAM_TRIGGER_WINDOW_CLOSE) {
taosArrayAddAll(pUpdated, pClosed);
closeIntervalWindow(pInfo->aggSup.pResultRowHashTable, &pInfo->twAggSup, &pInfo->interval, pUpdated);
}
}
taosArrayDestroy(pClosed);
finalizeUpdatedResult(pOperator->numOfExprs, pInfo->aggSup.pResultBuf, pUpdated, pInfo->binfo.rowCellInfoOffset);
initMultiResInfoFromArrayList(&pInfo->groupResInfo, pUpdated);
@ -2111,8 +2103,8 @@ static SSDataBlock* doStreamFinalIntervalAgg(SOperatorInfo* pOperator) {
return pInfo->binfo.pRes;
}
SOperatorInfo* createStreamFinalIntervalOperatorInfo(SOperatorInfo* downstream,
SPhysiNode* pPhyNode, SExecTaskInfo* pTaskInfo, int32_t numOfChild) {
SOperatorInfo* createStreamFinalIntervalOperatorInfo(SOperatorInfo* downstream, SPhysiNode* pPhyNode,
SExecTaskInfo* pTaskInfo, int32_t numOfChild) {
SIntervalPhysiNode* pIntervalPhyNode = (SIntervalPhysiNode*)pPhyNode;
SStreamFinalIntervalOperatorInfo* pInfo = taosMemoryCalloc(1, sizeof(SStreamFinalIntervalOperatorInfo));
SOperatorInfo* pOperator = taosMemoryCalloc(1, sizeof(SOperatorInfo));
@ -2125,12 +2117,12 @@ SOperatorInfo* createStreamFinalIntervalOperatorInfo(SOperatorInfo* downstream,
.intervalUnit = pIntervalPhyNode->intervalUnit,
.slidingUnit = pIntervalPhyNode->slidingUnit,
.offset = pIntervalPhyNode->offset,
.precision =
((SColumnNode*)pIntervalPhyNode->window.pTspk)->node.resType.precision};
pInfo->twAggSup = (STimeWindowAggSupp){.waterMark = pIntervalPhyNode->window.watermark,
.precision = ((SColumnNode*)pIntervalPhyNode->window.pTspk)->node.resType.precision};
pInfo->twAggSup = (STimeWindowAggSupp){
.waterMark = pIntervalPhyNode->window.watermark,
.calTrigger = pIntervalPhyNode->window.triggerType,
.maxTs = INT64_MIN,
.winMap = NULL, };
};
pInfo->primaryTsIndex = ((SColumnNode*)pIntervalPhyNode->window.pTspk)->slotId;
size_t keyBufSize = sizeof(int64_t) + sizeof(int64_t) + POINTER_BYTES;
initResultSizeInfo(pOperator, 4096);
@ -2162,7 +2154,7 @@ SOperatorInfo* createStreamFinalIntervalOperatorInfo(SOperatorInfo* downstream,
if (!isFinalInterval(pInfo)) {
pInfo->twAggSup.calTrigger = STREAM_TRIGGER_AT_ONCE;
}
pInfo->pUpdateRes = createResDataBlock(pPhyNode->pOutputDataBlockDesc);\
pInfo->pUpdateRes = createResDataBlock(pPhyNode->pOutputDataBlockDesc);
pInfo->pUpdateRes->info.type = STREAM_REPROCESS;
blockDataEnsureCapacity(pInfo->pUpdateRes, 128);
pInfo->pPhyNode = nodesCloneNode(pPhyNode);
@ -2176,9 +2168,9 @@ SOperatorInfo* createStreamFinalIntervalOperatorInfo(SOperatorInfo* downstream,
pOperator->numOfExprs = numOfCols;
pOperator->info = pInfo;
pOperator->fpSet = createOperatorFpSet(NULL, doStreamFinalIntervalAgg, NULL, NULL,
destroyStreamFinalIntervalOperatorInfo, aggEncodeResultRow, aggDecodeResultRow,
NULL);
pOperator->fpSet =
createOperatorFpSet(NULL, doStreamFinalIntervalAgg, NULL, NULL, destroyStreamFinalIntervalOperatorInfo,
aggEncodeResultRow, aggDecodeResultRow, NULL);
code = appendDownstream(pOperator, &downstream, 1);
if (code != TSDB_CODE_SUCCESS) {
@ -2218,8 +2210,7 @@ void destroyStreamSessionAggOperatorInfo(void* param, int32_t numOfOutput) {
}
}
int32_t initBiasicInfo(SOptrBasicInfo* pBasicInfo, SExprInfo* pExprInfo,
int32_t numOfCols, SSDataBlock* pResultBlock) {
int32_t initBiasicInfo(SOptrBasicInfo* pBasicInfo, SExprInfo* pExprInfo, int32_t numOfCols, SSDataBlock* pResultBlock) {
pBasicInfo->pCtx = createSqlFunctionCtx(pExprInfo, numOfCols, &pBasicInfo->rowCellInfoOffset);
pBasicInfo->pRes = pResultBlock;
for (int32_t i = 0; i < numOfCols; ++i) {
@ -3144,7 +3135,6 @@ SOperatorInfo* createStreamStateAggOperatorInfo(SOperatorInfo* downstream, SPhys
.waterMark = pStateNode->window.watermark,
.calTrigger = pStateNode->window.triggerType,
.maxTs = INT64_MIN,
.winMap = NULL,
};
initExecTimeWindowInfo(&pInfo->twAggSup.timeWindowData, &pTaskInfo->window);
@ -3197,3 +3187,281 @@ _error:
pTaskInfo->code = code;
return NULL;
}
typedef struct SMergeIntervalAggOperatorInfo {
SIntervalAggOperatorInfo intervalAggOperatorInfo;
SHashObj* groupIntervalHash;
bool hasGroupId;
uint64_t groupId;
SSDataBlock* prefetchedBlock;
bool inputBlocksFinished;
} SMergeIntervalAggOperatorInfo;
void destroyMergeIntervalOperatorInfo(void* param, int32_t numOfOutput) {
SMergeIntervalAggOperatorInfo* miaInfo = (SMergeIntervalAggOperatorInfo*)param;
taosHashCleanup(miaInfo->groupIntervalHash);
destroyIntervalOperatorInfo(&miaInfo->intervalAggOperatorInfo, numOfOutput);
}
static int32_t outputPrevIntervalResult(SOperatorInfo* pOperatorInfo, uint64_t tableGroupId, SSDataBlock* pResultBlock,
STimeWindow* newWin) {
SMergeIntervalAggOperatorInfo* miaInfo = pOperatorInfo->info;
SIntervalAggOperatorInfo* iaInfo = &miaInfo->intervalAggOperatorInfo;
SExecTaskInfo* pTaskInfo = pOperatorInfo->pTaskInfo;
bool ascScan = (iaInfo->order == TSDB_ORDER_ASC);
STimeWindow* prevWin = taosHashGet(miaInfo->groupIntervalHash, &tableGroupId, sizeof(tableGroupId));
if (prevWin == NULL) {
taosHashPut(miaInfo->groupIntervalHash, &tableGroupId, sizeof(tableGroupId), newWin, sizeof(STimeWindow));
return 0;
}
if (newWin == NULL || (ascScan && newWin->skey > prevWin->ekey || (!ascScan) && newWin->skey < prevWin->ekey) ) {
SET_RES_WINDOW_KEY(iaInfo->aggSup.keyBuf, &prevWin->skey, TSDB_KEYSIZE, tableGroupId);
SResultRowPosition* p1 = (SResultRowPosition*)taosHashGet(iaInfo->aggSup.pResultRowHashTable, iaInfo->aggSup.keyBuf,
GET_RES_WINDOW_KEY_LEN(TSDB_KEYSIZE));
ASSERT(p1 != NULL);
finalizeResultRowIntoResultDataBlock(iaInfo->aggSup.pResultBuf, p1, iaInfo->binfo.pCtx, pOperatorInfo->pExpr,
pOperatorInfo->numOfExprs, iaInfo->binfo.rowCellInfoOffset, pResultBlock,
pTaskInfo);
taosHashRemove(iaInfo->aggSup.pResultRowHashTable, iaInfo->aggSup.keyBuf, GET_RES_WINDOW_KEY_LEN(TSDB_KEYSIZE));
if (newWin == NULL) {
taosHashRemove(miaInfo->groupIntervalHash, &tableGroupId, sizeof(tableGroupId));
} else {
taosHashPut(miaInfo->groupIntervalHash, &tableGroupId, sizeof(tableGroupId), newWin, sizeof(STimeWindow));
}
}
return 0;
}
static void doMergeIntervalAggImpl(SOperatorInfo* pOperatorInfo, SResultRowInfo* pResultRowInfo, SSDataBlock* pBlock,
int32_t scanFlag, SSDataBlock* pResultBlock) {
SMergeIntervalAggOperatorInfo* miaInfo = pOperatorInfo->info;
SIntervalAggOperatorInfo* iaInfo = &miaInfo->intervalAggOperatorInfo;
SExecTaskInfo* pTaskInfo = pOperatorInfo->pTaskInfo;
int32_t startPos = 0;
int32_t numOfOutput = pOperatorInfo->numOfExprs;
int64_t* tsCols = extractTsCol(pBlock, iaInfo);
uint64_t tableGroupId = pBlock->info.groupId;
bool ascScan = (iaInfo->order == TSDB_ORDER_ASC);
TSKEY blockStartTs = getStartTsKey(&pBlock->info.window, tsCols);
SResultRow* pResult = NULL;
STimeWindow win = getActiveTimeWindow(iaInfo->aggSup.pResultBuf, pResultRowInfo, blockStartTs, &iaInfo->interval,
iaInfo->interval.precision, &iaInfo->win);
int32_t ret =
setTimeWindowOutputBuf(pResultRowInfo, &win, (scanFlag == MAIN_SCAN), &pResult, tableGroupId, iaInfo->binfo.pCtx,
numOfOutput, iaInfo->binfo.rowCellInfoOffset, &iaInfo->aggSup, pTaskInfo);
if (ret != TSDB_CODE_SUCCESS || pResult == NULL) {
longjmp(pTaskInfo->env, TSDB_CODE_QRY_OUT_OF_MEMORY);
}
TSKEY ekey = ascScan ? win.ekey : win.skey;
int32_t forwardRows =
getNumOfRowsInTimeWindow(&pBlock->info, tsCols, startPos, ekey, binarySearchForKey, NULL, iaInfo->order);
ASSERT(forwardRows > 0);
// prev time window not interpolation yet.
if (iaInfo->timeWindowInterpo) {
SResultRowPosition pos = addToOpenWindowList(pResultRowInfo, pResult);
doInterpUnclosedTimeWindow(pOperatorInfo, numOfOutput, pResultRowInfo, pBlock, scanFlag, tsCols, &pos);
// restore current time window
ret = setTimeWindowOutputBuf(pResultRowInfo, &win, (scanFlag == MAIN_SCAN), &pResult, tableGroupId,
iaInfo->binfo.pCtx, numOfOutput, iaInfo->binfo.rowCellInfoOffset, &iaInfo->aggSup,
pTaskInfo);
if (ret != TSDB_CODE_SUCCESS) {
longjmp(pTaskInfo->env, TSDB_CODE_QRY_OUT_OF_MEMORY);
}
// window start key interpolation
doWindowBorderInterpolation(iaInfo, pBlock, numOfOutput, iaInfo->binfo.pCtx, pResult, &win, startPos, forwardRows);
}
updateTimeWindowInfo(&iaInfo->twAggSup.timeWindowData, &win, true);
doApplyFunctions(pTaskInfo, iaInfo->binfo.pCtx, &win, &iaInfo->twAggSup.timeWindowData, startPos, forwardRows, tsCols,
pBlock->info.rows, numOfOutput, iaInfo->order);
doCloseWindow(pResultRowInfo, iaInfo, pResult);
// output previous interval results after this interval (&win) is closed
outputPrevIntervalResult(pOperatorInfo, tableGroupId, pResultBlock, &win);
STimeWindow nextWin = win;
while (1) {
int32_t prevEndPos = forwardRows - 1 + startPos;
startPos = getNextQualifiedWindow(&iaInfo->interval, &nextWin, &pBlock->info, tsCols, prevEndPos, iaInfo->order);
if (startPos < 0) {
break;
}
// null data, failed to allocate more memory buffer
int32_t code = setTimeWindowOutputBuf(pResultRowInfo, &nextWin, (scanFlag == MAIN_SCAN), &pResult, tableGroupId,
iaInfo->binfo.pCtx, numOfOutput, iaInfo->binfo.rowCellInfoOffset,
&iaInfo->aggSup, pTaskInfo);
if (code != TSDB_CODE_SUCCESS || pResult == NULL) {
longjmp(pTaskInfo->env, TSDB_CODE_QRY_OUT_OF_MEMORY);
}
ekey = ascScan ? nextWin.ekey : nextWin.skey;
forwardRows =
getNumOfRowsInTimeWindow(&pBlock->info, tsCols, startPos, ekey, binarySearchForKey, NULL, iaInfo->order);
// window start(end) key interpolation
doWindowBorderInterpolation(iaInfo, pBlock, numOfOutput, iaInfo->binfo.pCtx, pResult, &nextWin, startPos,
forwardRows);
updateTimeWindowInfo(&iaInfo->twAggSup.timeWindowData, &nextWin, true);
doApplyFunctions(pTaskInfo, iaInfo->binfo.pCtx, &nextWin, &iaInfo->twAggSup.timeWindowData, startPos, forwardRows,
tsCols, pBlock->info.rows, numOfOutput, iaInfo->order);
doCloseWindow(pResultRowInfo, iaInfo, pResult);
// output previous interval results after this interval (&nextWin) is closed
outputPrevIntervalResult(pOperatorInfo, tableGroupId, pResultBlock, &nextWin);
}
if (iaInfo->timeWindowInterpo) {
saveDataBlockLastRow(iaInfo->pPrevValues, pBlock, iaInfo->pInterpCols);
}
}
static SSDataBlock* doMergeIntervalAgg(SOperatorInfo* pOperator) {
SExecTaskInfo* pTaskInfo = pOperator->pTaskInfo;
SMergeIntervalAggOperatorInfo* miaInfo = pOperator->info;
SIntervalAggOperatorInfo* iaInfo = &miaInfo->intervalAggOperatorInfo;
if (pOperator->status == OP_EXEC_DONE) {
return NULL;
}
SSDataBlock* pRes = iaInfo->binfo.pRes;
blockDataCleanup(pRes);
blockDataEnsureCapacity(pRes, pOperator->resultInfo.capacity);
if (!miaInfo->inputBlocksFinished) {
SOperatorInfo* downstream = pOperator->pDownstream[0];
int32_t scanFlag = MAIN_SCAN;
while (1) {
SSDataBlock* pBlock = NULL;
if (miaInfo->prefetchedBlock == NULL) {
pBlock = downstream->fpSet.getNextFn(downstream);
} else {
pBlock = miaInfo->prefetchedBlock;
miaInfo->groupId = pBlock->info.groupId;
}
if (pBlock == NULL) {
miaInfo->inputBlocksFinished = true;
break;
}
if (!miaInfo->hasGroupId) {
miaInfo->hasGroupId = true;
miaInfo->groupId = pBlock->info.groupId;
} else if (miaInfo->groupId != pBlock->info.groupId) {
miaInfo->prefetchedBlock = pBlock;
break;
}
getTableScanInfo(pOperator, &iaInfo->order, &scanFlag);
setInputDataBlock(pOperator, iaInfo->binfo.pCtx, pBlock, iaInfo->order, scanFlag, true);
STableQueryInfo* pTableQueryInfo = iaInfo->pCurrent;
setIntervalQueryRange(pTableQueryInfo, pBlock->info.window.skey, &pTaskInfo->window);
doMergeIntervalAggImpl(pOperator, &iaInfo->binfo.resultRowInfo, pBlock, scanFlag, pRes);
if (pRes->info.rows >= pOperator->resultInfo.threshold) {
break;
}
}
pRes->info.groupId = miaInfo->groupId;
} else {
void* p = taosHashIterate(miaInfo->groupIntervalHash, NULL);
if (p != NULL) {
size_t len = 0;
uint64_t* pKey = taosHashGetKey(p, &len);
outputPrevIntervalResult(pOperator, *pKey, pRes, NULL);
}
}
if (pRes->info.rows == 0) {
doSetOperatorCompleted(pOperator);
}
size_t rows = pRes->info.rows;
pOperator->resultInfo.totalRows += rows;
return (rows == 0) ? NULL : pRes;
}
SOperatorInfo* createMergeIntervalOperatorInfo(SOperatorInfo* downstream, SExprInfo* pExprInfo, int32_t numOfCols,
SSDataBlock* pResBlock, SInterval* pInterval, int32_t primaryTsSlotId,
SExecTaskInfo* pTaskInfo) {
SMergeIntervalAggOperatorInfo* miaInfo = taosMemoryCalloc(1, sizeof(SMergeIntervalAggOperatorInfo));
SOperatorInfo* pOperator = taosMemoryCalloc(1, sizeof(SOperatorInfo));
if (miaInfo == NULL || pOperator == NULL) {
goto _error;
}
SIntervalAggOperatorInfo* iaInfo = &miaInfo->intervalAggOperatorInfo;
iaInfo->win = pTaskInfo->window;
iaInfo->order = TSDB_ORDER_ASC;
iaInfo->interval = *pInterval;
iaInfo->execModel = pTaskInfo->execModel;
iaInfo->primaryTsIndex = primaryTsSlotId;
miaInfo->groupIntervalHash = taosHashInit(10, taosGetDefaultHashFunction(TSDB_DATA_TYPE_UBIGINT), true, HASH_NO_LOCK);
size_t keyBufSize = sizeof(int64_t) + sizeof(int64_t) + POINTER_BYTES;
initResultSizeInfo(pOperator, 4096);
int32_t code =
initAggInfo(&iaInfo->binfo, &iaInfo->aggSup, pExprInfo, numOfCols, pResBlock, keyBufSize, pTaskInfo->id.str);
initExecTimeWindowInfo(&iaInfo->twAggSup.timeWindowData, &iaInfo->win);
iaInfo->timeWindowInterpo = timeWindowinterpNeeded(iaInfo->binfo.pCtx, numOfCols, iaInfo);
if (iaInfo->timeWindowInterpo) {
iaInfo->binfo.resultRowInfo.openWindow = tdListNew(sizeof(SResultRowPosition));
}
// iaInfo->pTableQueryInfo = initTableQueryInfo(pTableGroupInfo);
if (code != TSDB_CODE_SUCCESS /* || iaInfo->pTableQueryInfo == NULL*/) {
goto _error;
}
initResultRowInfo(&iaInfo->binfo.resultRowInfo, (int32_t)1);
pOperator->name = "TimeMergeIntervalAggOperator";
pOperator->operatorType = QUERY_NODE_PHYSICAL_PLAN_MERGE_INTERVAL;
pOperator->blocking = false;
pOperator->status = OP_NOT_OPENED;
pOperator->pExpr = pExprInfo;
pOperator->pTaskInfo = pTaskInfo;
pOperator->numOfExprs = numOfCols;
pOperator->info = miaInfo;
pOperator->fpSet = createOperatorFpSet(operatorDummyOpenFn, doMergeIntervalAgg, NULL, NULL,
destroyMergeIntervalOperatorInfo, NULL, NULL, NULL);
code = appendDownstream(pOperator, &downstream, 1);
if (code != TSDB_CODE_SUCCESS) {
goto _error;
}
return pOperator;
_error:
destroyMergeIntervalOperatorInfo(miaInfo, numOfCols);
taosMemoryFreeClear(miaInfo);
taosMemoryFreeClear(pOperator);
pTaskInfo->code = code;
return NULL;
}

4
source/libs/function/inc/builtinsimpl.h
View File

@ -112,6 +112,7 @@ int32_t spreadFunctionMerge(SqlFunctionCtx* pCtx);
int32_t spreadFinalize(SqlFunctionCtx* pCtx, SSDataBlock* pBlock);
int32_t spreadPartialFinalize(SqlFunctionCtx* pCtx, SSDataBlock* pBlock);
int32_t getSpreadInfoSize();
int32_t spreadCombine(SqlFunctionCtx* pDestCtx, SqlFunctionCtx* pSourceCtx);
bool getElapsedFuncEnv(struct SFunctionNode* pFunc, SFuncExecEnv* pEnv);
bool elapsedFunctionSetup(SqlFunctionCtx *pCtx, SResultRowEntryInfo* pResultInfo);
@ -120,6 +121,7 @@ int32_t elapsedFunctionMerge(SqlFunctionCtx* pCtx);
int32_t elapsedFinalize(SqlFunctionCtx* pCtx, SSDataBlock* pBlock);
int32_t elapsedPartialFinalize(SqlFunctionCtx* pCtx, SSDataBlock* pBlock);
int32_t getElapsedInfoSize();
int32_t elapsedCombine(SqlFunctionCtx* pDestCtx, SqlFunctionCtx* pSourceCtx);
bool getHistogramFuncEnv(struct SFunctionNode* pFunc, SFuncExecEnv* pEnv);
bool histogramFunctionSetup(SqlFunctionCtx *pCtx, SResultRowEntryInfo* pResultInfo);
@ -128,6 +130,7 @@ int32_t histogramFunctionMerge(SqlFunctionCtx* pCtx);
int32_t histogramFinalize(SqlFunctionCtx* pCtx, SSDataBlock* pBlock);
int32_t histogramPartialFinalize(SqlFunctionCtx* pCtx, SSDataBlock* pBlock);
int32_t getHistogramInfoSize();
int32_t histogramCombine(SqlFunctionCtx* pDestCtx, SqlFunctionCtx* pSourceCtx);
bool getHLLFuncEnv(struct SFunctionNode* pFunc, SFuncExecEnv* pEnv);
int32_t hllFunction(SqlFunctionCtx* pCtx);
@ -135,6 +138,7 @@ int32_t hllFunctionMerge(SqlFunctionCtx* pCtx);
int32_t hllFinalize(SqlFunctionCtx* pCtx, SSDataBlock* pBlock);
int32_t hllPartialFinalize(SqlFunctionCtx* pCtx, SSDataBlock* pBlock);
int32_t getHLLInfoSize();
int32_t hllCombine(SqlFunctionCtx* pDestCtx, SqlFunctionCtx* pSourceCtx);
bool getStateFuncEnv(struct SFunctionNode* pFunc, SFuncExecEnv* pEnv);
bool stateFunctionSetup(SqlFunctionCtx *pCtx, SResultRowEntryInfo* pResultInfo);

53
source/libs/function/src/builtins.c
View File

@ -1112,7 +1112,7 @@ static bool validateHourRange(int8_t hour) {
}
static bool validateMinuteRange(int8_t hour, int8_t minute, char sign) {
if (minute == 0 || (minute == 30 && (hour == 3 || hour == 5) && sign == '-')) {
if (minute == 0 || (minute == 30 && (hour == 3 || hour == 5) && sign == '+')) {
return true;
}
@ -1447,6 +1447,7 @@ const SBuiltinFuncDefinition funcMgtBuiltins[] = {
.initFunc = apercentileFunctionSetup,
.processFunc = apercentileFunction,
.finalizeFunc = apercentileFinalize,
.invertFunc = NULL,
.combineFunc = apercentileCombine,
.pPartialFunc = "_apercentile_partial",
.pMergeFunc = "_apercentile_merge"
@ -1459,7 +1460,9 @@ const SBuiltinFuncDefinition funcMgtBuiltins[] = {
.getEnvFunc = getApercentileFuncEnv,
.initFunc = apercentileFunctionSetup,
.processFunc = apercentileFunction,
.finalizeFunc = apercentilePartialFinalize
.finalizeFunc = apercentilePartialFinalize,
.invertFunc = NULL,
.combineFunc = apercentileCombine,
},
{
.name = "_apercentile_merge",
@ -1469,7 +1472,9 @@ const SBuiltinFuncDefinition funcMgtBuiltins[] = {
.getEnvFunc = getApercentileFuncEnv,
.initFunc = functionSetup,
.processFunc = apercentileFunctionMerge,
.finalizeFunc = apercentileFinalize
.finalizeFunc = apercentileFinalize,
.invertFunc = NULL,
.combineFunc = apercentileCombine,
},
{
.name = "top",
@ -1503,6 +1508,8 @@ const SBuiltinFuncDefinition funcMgtBuiltins[] = {
.initFunc = spreadFunctionSetup,
.processFunc = spreadFunction,
.finalizeFunc = spreadFinalize,
.invertFunc = NULL,
.combineFunc = spreadCombine,
.pPartialFunc = "_spread_partial",
.pMergeFunc = "_spread_merge"
},
@ -1515,7 +1522,9 @@ const SBuiltinFuncDefinition funcMgtBuiltins[] = {
.getEnvFunc = getSpreadFuncEnv,
.initFunc = spreadFunctionSetup,
.processFunc = spreadFunction,
.finalizeFunc = spreadPartialFinalize
.finalizeFunc = spreadPartialFinalize,
.invertFunc = NULL,
.combineFunc = spreadCombine,
},
{
.name = "_spread_merge",
@ -1526,7 +1535,9 @@ const SBuiltinFuncDefinition funcMgtBuiltins[] = {
.getEnvFunc = getSpreadFuncEnv,
.initFunc = spreadFunctionSetup,
.processFunc = spreadFunctionMerge,
.finalizeFunc = spreadFinalize
.finalizeFunc = spreadFinalize,
.invertFunc = NULL,
.combineFunc = spreadCombine,
},
{
.name = "elapsed",
@ -1538,6 +1549,8 @@ const SBuiltinFuncDefinition funcMgtBuiltins[] = {
.initFunc = elapsedFunctionSetup,
.processFunc = elapsedFunction,
.finalizeFunc = elapsedFinalize,
.invertFunc = NULL,
.combineFunc = elapsedCombine,
.pPartialFunc = "_elapsed_partial",
.pMergeFunc = "_elapsed_merge"
},
@ -1550,7 +1563,9 @@ const SBuiltinFuncDefinition funcMgtBuiltins[] = {
.getEnvFunc = getElapsedFuncEnv,
.initFunc = elapsedFunctionSetup,
.processFunc = elapsedFunction,
.finalizeFunc = elapsedPartialFinalize
.finalizeFunc = elapsedPartialFinalize,
.invertFunc = NULL,
.combineFunc = elapsedCombine,
},
{
.name = "_elapsed_merge",
@ -1561,7 +1576,9 @@ const SBuiltinFuncDefinition funcMgtBuiltins[] = {
.getEnvFunc = getElapsedFuncEnv,
.initFunc = elapsedFunctionSetup,
.processFunc = elapsedFunctionMerge,
.finalizeFunc = elapsedFinalize
.finalizeFunc = elapsedFinalize,
.invertFunc = NULL,
.combineFunc = elapsedCombine,
},
{
.name = "last_row",
@ -1614,8 +1631,10 @@ const SBuiltinFuncDefinition funcMgtBuiltins[] = {
.initFunc = histogramFunctionSetup,
.processFunc = histogramFunction,
.finalizeFunc = histogramFinalize,
.invertFunc = NULL,
.combineFunc = histogramCombine,
.pPartialFunc = "_histogram_partial",
.pMergeFunc = "_histogram_merge"
.pMergeFunc = "_histogram_merge",
},
{
.name = "_histogram_partial",
@ -1625,7 +1644,9 @@ const SBuiltinFuncDefinition funcMgtBuiltins[] = {
.getEnvFunc = getHistogramFuncEnv,
.initFunc = histogramFunctionSetup,
.processFunc = histogramFunction,
.finalizeFunc = histogramPartialFinalize
.finalizeFunc = histogramPartialFinalize,
.invertFunc = NULL,
.combineFunc = histogramCombine,
},
{
.name = "_histogram_merge",
@ -1635,7 +1656,9 @@ const SBuiltinFuncDefinition funcMgtBuiltins[] = {
.getEnvFunc = getHistogramFuncEnv,
.initFunc = functionSetup,
.processFunc = histogramFunctionMerge,
.finalizeFunc = histogramFinalize
.finalizeFunc = histogramFinalize,
.invertFunc = NULL,
.combineFunc = histogramCombine,
},
{
.name = "hyperloglog",
@ -1646,6 +1669,8 @@ const SBuiltinFuncDefinition funcMgtBuiltins[] = {
.initFunc = functionSetup,
.processFunc = hllFunction,
.finalizeFunc = hllFinalize,
.invertFunc = NULL,
.combineFunc = hllCombine,
.pPartialFunc = "_hyperloglog_partial",
.pMergeFunc = "_hyperloglog_merge"
},
@ -1657,7 +1682,9 @@ const SBuiltinFuncDefinition funcMgtBuiltins[] = {
.getEnvFunc = getHLLFuncEnv,
.initFunc = functionSetup,
.processFunc = hllFunction,
.finalizeFunc = hllPartialFinalize
.finalizeFunc = hllPartialFinalize,
.invertFunc = NULL,
.combineFunc = hllCombine,
},
{
.name = "_hyperloglog_merge",
@ -1667,7 +1694,9 @@ const SBuiltinFuncDefinition funcMgtBuiltins[] = {
.getEnvFunc = getHLLFuncEnv,
.initFunc = functionSetup,
.processFunc = hllFunctionMerge,
.finalizeFunc = hllFinalize
.finalizeFunc = hllFinalize,
.invertFunc = NULL,
.combineFunc = hllCombine,
},
{
.name = "diff",

219
source/libs/function/src/builtinsimpl.c
View File

@ -2103,8 +2103,49 @@ int32_t apercentileFunction(SqlFunctionCtx* pCtx) {
return TSDB_CODE_SUCCESS;
}
static void apercentileTransferInfo(SAPercentileInfo* pInput, SAPercentileInfo* pOutput) {
pOutput->percent = pInput->percent;
pOutput->algo = pInput->algo;
if (pOutput->algo == APERCT_ALGO_TDIGEST) {
buildTDigestInfo(pInput);
tdigestAutoFill(pInput->pTDigest, COMPRESSION);
if(pInput->pTDigest->num_centroids == 0 && pInput->pTDigest->num_buffered_pts == 0) {
return;
}
buildTDigestInfo(pOutput);
TDigest *pTDigest = pOutput->pTDigest;
if(pTDigest->num_centroids <= 0) {
memcpy(pTDigest, pInput->pTDigest, (size_t)TDIGEST_SIZE(COMPRESSION));
tdigestAutoFill(pTDigest, COMPRESSION);
} else {
tdigestMerge(pTDigest, pInput->pTDigest);
}
} else {
buildHistogramInfo(pInput);
if (pInput->pHisto->numOfElems <= 0) {
return;
}
buildHistogramInfo(pOutput);
SHistogramInfo *pHisto = pOutput->pHisto;
if (pHisto->numOfElems <= 0) {
memcpy(pHisto, pInput->pHisto, sizeof(SHistogramInfo) + sizeof(SHistBin) * (MAX_HISTOGRAM_BIN + 1));
pHisto->elems = (SHistBin*) ((char *)pHisto + sizeof(SHistogramInfo));
} else {
pHisto->elems = (SHistBin*) ((char *)pHisto + sizeof(SHistogramInfo));
SHistogramInfo *pRes = tHistogramMerge(pHisto, pInput->pHisto, MAX_HISTOGRAM_BIN);
memcpy(pHisto, pRes, sizeof(SHistogramInfo) + sizeof(SHistBin) * MAX_HISTOGRAM_BIN);
pHisto->elems = (SHistBin*) ((char *)pHisto + sizeof(SHistogramInfo));
tHistogramDestroy(&pRes);
}
}
}
int32_t apercentileFunctionMerge(SqlFunctionCtx* pCtx) {
int32_t numOfElems = 0;
SResultRowEntryInfo* pResInfo = GET_RES_INFO(pCtx);
SInputColumnInfoData* pInput = &pCtx->input;
@ -2113,60 +2154,14 @@ int32_t apercentileFunctionMerge(SqlFunctionCtx* pCtx) {
ASSERT(pCol->info.type == TSDB_DATA_TYPE_BINARY);
SAPercentileInfo* pInfo = GET_ROWCELL_INTERBUF(pResInfo);
SAPercentileInfo* pInputInfo;
int32_t start = pInput->startRowIndex;
for (int32_t i = start; i < pInput->numOfRows + start; ++i) {
//if (colDataIsNull_s(pCol, i)) {
// continue;
//}
numOfElems += 1;
char* data = colDataGetData(pCol, i);
char* data = colDataGetData(pCol, start);
SAPercentileInfo* pInputInfo = (SAPercentileInfo *)varDataVal(data);
pInputInfo = (SAPercentileInfo *)varDataVal(data);
}
apercentileTransferInfo(pInputInfo, pInfo);
pInfo->percent = pInputInfo->percent;
pInfo->algo = pInputInfo->algo;
if (pInfo->algo == APERCT_ALGO_TDIGEST) {
buildTDigestInfo(pInputInfo);
tdigestAutoFill(pInputInfo->pTDigest, COMPRESSION);
if(pInputInfo->pTDigest->num_centroids == 0 && pInputInfo->pTDigest->num_buffered_pts == 0) {
return TSDB_CODE_SUCCESS;
}
buildTDigestInfo(pInfo);
TDigest *pTDigest = pInfo->pTDigest;
if(pTDigest->num_centroids <= 0) {
memcpy(pTDigest, pInputInfo->pTDigest, (size_t)TDIGEST_SIZE(COMPRESSION));
tdigestAutoFill(pTDigest, COMPRESSION);
} else {
tdigestMerge(pTDigest, pInputInfo->pTDigest);
}
} else {
buildHistogramInfo(pInputInfo);
if (pInputInfo->pHisto->numOfElems <= 0) {
return TSDB_CODE_SUCCESS;
}
buildHistogramInfo(pInfo);
SHistogramInfo *pHisto = pInfo->pHisto;
if (pHisto->numOfElems <= 0) {
memcpy(pHisto, pInputInfo->pHisto, sizeof(SHistogramInfo) + sizeof(SHistBin) * (MAX_HISTOGRAM_BIN + 1));
pHisto->elems = (SHistBin*) ((char *)pHisto + sizeof(SHistogramInfo));
} else {
pHisto->elems = (SHistBin*) ((char *)pHisto + sizeof(SHistogramInfo));
SHistogramInfo *pRes = tHistogramMerge(pHisto, pInputInfo->pHisto, MAX_HISTOGRAM_BIN);
memcpy(pHisto, pRes, sizeof(SHistogramInfo) + sizeof(SHistBin) * MAX_HISTOGRAM_BIN);
pHisto->elems = (SHistBin*) ((char *)pHisto + sizeof(SHistogramInfo));
tHistogramDestroy(&pRes);
}
}
SET_VAL(pResInfo, numOfElems, 1);
SET_VAL(pResInfo, 1, 1);
return TSDB_CODE_SUCCESS;
}
@ -2232,7 +2227,6 @@ int32_t apercentilePartialFinalize(SqlFunctionCtx* pCtx, SSDataBlock* pBlock) {
int32_t apercentileCombine(SqlFunctionCtx* pDestCtx, SqlFunctionCtx* pSourceCtx) {
SResultRowEntryInfo* pDResInfo = GET_RES_INFO(pDestCtx);
SAPercentileInfo* pDBuf = GET_ROWCELL_INTERBUF(pDResInfo);
int32_t type = pDestCtx->input.pData[0]->info.type;
SResultRowEntryInfo* pSResInfo = GET_RES_INFO(pSourceCtx);
SAPercentileInfo* pSBuf = GET_ROWCELL_INTERBUF(pSResInfo);
@ -3049,6 +3043,17 @@ _spread_over:
return TSDB_CODE_SUCCESS;
}
static void spreadTransferInfo(SSpreadInfo* pInput, SSpreadInfo* pOutput) {
pOutput->hasResult = pInput->hasResult;
if (pInput->max > pOutput->max) {
pOutput->max = pInput->max;
}
if (pInput->min < pOutput->min) {
pOutput->min = pInput->min;
}
}
int32_t spreadFunctionMerge(SqlFunctionCtx *pCtx) {
SInputColumnInfoData* pInput = &pCtx->input;
SColumnInfoData* pCol = pInput->pData[0];
@ -3061,14 +3066,7 @@ int32_t spreadFunctionMerge(SqlFunctionCtx *pCtx) {
char* data = colDataGetData(pCol, start);
pInputInfo = (SSpreadInfo *)varDataVal(data);
pInfo->hasResult = pInputInfo->hasResult;
if (pInputInfo->max > pInfo->max) {
pInfo->max = pInputInfo->max;
}
if (pInputInfo->min < pInfo->min) {
pInfo->min = pInputInfo->min;
}
spreadTransferInfo(pInputInfo, pInfo);
SET_VAL(GET_RES_INFO(pCtx), 1, 1);
@ -3101,6 +3099,17 @@ int32_t spreadPartialFinalize(SqlFunctionCtx* pCtx, SSDataBlock* pBlock) {
return pResInfo->numOfRes;
}
int32_t spreadCombine(SqlFunctionCtx* pDestCtx, SqlFunctionCtx* pSourceCtx) {
SResultRowEntryInfo* pDResInfo = GET_RES_INFO(pDestCtx);
SSpreadInfo* pDBuf = GET_ROWCELL_INTERBUF(pDResInfo);
SResultRowEntryInfo* pSResInfo = GET_RES_INFO(pSourceCtx);
SSpreadInfo* pSBuf = GET_ROWCELL_INTERBUF(pSResInfo);
spreadTransferInfo(pDBuf, pSBuf);
pDResInfo->numOfRes = TMAX(pDResInfo->numOfRes, pSResInfo->numOfRes);
return TSDB_CODE_SUCCESS;
}
int32_t getElapsedInfoSize() {
return (int32_t)sizeof(SElapsedInfo);
}
@ -3206,6 +3215,17 @@ _elapsed_over:
return TSDB_CODE_SUCCESS;
}
static void elapsedTransferInfo(SElapsedInfo* pInput, SElapsedInfo* pOutput) {
pOutput->timeUnit = pInput->timeUnit;
if (pOutput->min > pInput->min) {
pOutput->min = pInput->min;
}
if (pOutput->max < pInput->max) {
pOutput->max = pInput->max;
}
}
int32_t elapsedFunctionMerge(SqlFunctionCtx *pCtx) {
SInputColumnInfoData* pInput = &pCtx->input;
SColumnInfoData* pCol = pInput->pData[0];
@ -3217,14 +3237,7 @@ int32_t elapsedFunctionMerge(SqlFunctionCtx *pCtx) {
char* data = colDataGetData(pCol, start);
SElapsedInfo* pInputInfo = (SElapsedInfo *)varDataVal(data);
pInfo->timeUnit = pInputInfo->timeUnit;
if (pInfo->min > pInputInfo->min) {
pInfo->min = pInputInfo->min;
}
if (pInfo->max < pInputInfo->max) {
pInfo->max = pInputInfo->max;
}
elapsedTransferInfo(pInputInfo, pInfo);
SET_VAL(GET_RES_INFO(pCtx), 1, 1);
return TSDB_CODE_SUCCESS;
@ -3256,6 +3269,18 @@ int32_t elapsedPartialFinalize(SqlFunctionCtx* pCtx, SSDataBlock* pBlock) {
return pResInfo->numOfRes;
}
int32_t elapsedCombine(SqlFunctionCtx* pDestCtx, SqlFunctionCtx* pSourceCtx) {
SResultRowEntryInfo* pDResInfo = GET_RES_INFO(pDestCtx);
SElapsedInfo* pDBuf = GET_ROWCELL_INTERBUF(pDResInfo);
SResultRowEntryInfo* pSResInfo = GET_RES_INFO(pSourceCtx);
SElapsedInfo* pSBuf = GET_ROWCELL_INTERBUF(pSResInfo);
elapsedTransferInfo(pDBuf, pSBuf);
pDResInfo->numOfRes = TMAX(pDResInfo->numOfRes, pSResInfo->numOfRes);
return TSDB_CODE_SUCCESS;
}
int32_t getHistogramInfoSize() {
return (int32_t)sizeof(SHistoFuncInfo) + HISTOGRAM_MAX_BINS_NUM * sizeof(SHistoFuncBin);
}
@ -3470,6 +3495,17 @@ int32_t histogramFunction(SqlFunctionCtx *pCtx) {
return TSDB_CODE_SUCCESS;
}
static void histogramTransferInfo(SHistoFuncInfo* pInput, SHistoFuncInfo* pOutput) {
pOutput->normalized = pInput->normalized;
pOutput->numOfBins = pInput->numOfBins;
pOutput->totalCount += pInput->totalCount;
for (int32_t k = 0; k < pOutput->numOfBins; ++k) {
pOutput->bins[k].lower = pInput->bins[k].lower;
pOutput->bins[k].upper = pInput->bins[k].upper;
pOutput->bins[k].count += pInput->bins[k].count;
}
}
int32_t histogramFunctionMerge(SqlFunctionCtx *pCtx) {
SInputColumnInfoData* pInput = &pCtx->input;
SColumnInfoData* pCol = pInput->pData[0];
@ -3481,14 +3517,7 @@ int32_t histogramFunctionMerge(SqlFunctionCtx *pCtx) {
char* data = colDataGetData(pCol, start);
SHistoFuncInfo* pInputInfo = (SHistoFuncInfo *)varDataVal(data);
pInfo->normalized = pInputInfo->normalized;
pInfo->numOfBins = pInputInfo->numOfBins;
pInfo->totalCount += pInputInfo->totalCount;
for (int32_t k = 0; k < pInfo->numOfBins; ++k) {
pInfo->bins[k].lower = pInputInfo->bins[k].lower;
pInfo->bins[k].upper = pInputInfo->bins[k].upper;
pInfo->bins[k].count += pInputInfo->bins[k].count;
}
histogramTransferInfo(pInputInfo, pInfo);
SET_VAL(GET_RES_INFO(pCtx), pInfo->numOfBins, pInfo->numOfBins);
return TSDB_CODE_SUCCESS;
@ -3547,6 +3576,18 @@ int32_t histogramPartialFinalize(SqlFunctionCtx* pCtx, SSDataBlock* pBlock) {
return 1;
}
int32_t histogramCombine(SqlFunctionCtx* pDestCtx, SqlFunctionCtx* pSourceCtx) {
SResultRowEntryInfo* pDResInfo = GET_RES_INFO(pDestCtx);
SHistoFuncInfo* pDBuf = GET_ROWCELL_INTERBUF(pDResInfo);
SResultRowEntryInfo* pSResInfo = GET_RES_INFO(pSourceCtx);
SHistoFuncInfo* pSBuf = GET_ROWCELL_INTERBUF(pSResInfo);
histogramTransferInfo(pDBuf, pSBuf);
pDResInfo->numOfRes = TMAX(pDResInfo->numOfRes, pSResInfo->numOfRes);
return TSDB_CODE_SUCCESS;
}
int32_t getHLLInfoSize() {
return (int32_t)sizeof(SHLLInfo);
}
@ -3676,6 +3717,14 @@ int32_t hllFunction(SqlFunctionCtx *pCtx) {
return TSDB_CODE_SUCCESS;
}
static void hllTransferInfo(SHLLInfo* pInput, SHLLInfo* pOutput) {
for (int32_t k = 0; k < HLL_BUCKETS; ++k) {
if (pOutput->buckets[k] < pInput->buckets[k]) {
pOutput->buckets[k] = pInput->buckets[k];
}
}
}
int32_t hllFunctionMerge(SqlFunctionCtx *pCtx) {
SInputColumnInfoData* pInput = &pCtx->input;
SColumnInfoData* pCol = pInput->pData[0];
@ -3687,11 +3736,7 @@ int32_t hllFunctionMerge(SqlFunctionCtx *pCtx) {
char* data = colDataGetData(pCol, start);
SHLLInfo* pInputInfo = (SHLLInfo *)varDataVal(data);
for (int32_t k = 0; k < HLL_BUCKETS; ++k) {
if (pInfo->buckets[k] < pInputInfo->buckets[k]) {
pInfo->buckets[k] = pInputInfo->buckets[k];
}
}
hllTransferInfo(pInputInfo, pInfo);
SET_VAL(GET_RES_INFO(pCtx), 1, 1);
return TSDB_CODE_SUCCESS;
@ -3727,6 +3772,18 @@ int32_t hllPartialFinalize(SqlFunctionCtx* pCtx, SSDataBlock* pBlock) {
return pResInfo->numOfRes;
}
int32_t hllCombine(SqlFunctionCtx* pDestCtx, SqlFunctionCtx* pSourceCtx) {
SResultRowEntryInfo* pDResInfo = GET_RES_INFO(pDestCtx);
SHLLInfo* pDBuf = GET_ROWCELL_INTERBUF(pDResInfo);
SResultRowEntryInfo* pSResInfo = GET_RES_INFO(pSourceCtx);
SHLLInfo* pSBuf = GET_ROWCELL_INTERBUF(pSResInfo);
hllTransferInfo(pDBuf, pSBuf);
pDResInfo->numOfRes = TMAX(pDResInfo->numOfRes, pSResInfo->numOfRes);
return TSDB_CODE_SUCCESS;
}
bool getStateFuncEnv(SFunctionNode* UNUSED_PARAM(pFunc), SFuncExecEnv* pEnv) {
pEnv->calcMemSize = sizeof(SStateInfo);
return true;

9
source/libs/function/src/tpercentile.c
View File

@ -127,9 +127,11 @@ int32_t tBucketIntHash(tMemBucket *pBucket, const void *value) {
int64_t delta = v - pBucket->range.i64MinVal;
index = (delta % pBucket->numOfSlots);
} else {
double slotSpan = (double)span / pBucket->numOfSlots;
index = (int32_t)((v - pBucket->range.i64MinVal) / slotSpan);
if (v == pBucket->range.i64MaxVal) {
double slotSpan = ((double)span) / pBucket->numOfSlots;
uint64_t delta = v - pBucket->range.i64MinVal;
index = (int32_t)(delta / slotSpan);
if (v == pBucket->range.i64MaxVal || index == pBucket->numOfSlots) {
index -= 1;
}
}
@ -324,7 +326,6 @@ int32_t tMemBucketPut(tMemBucket *pBucket, const void *data, size_t size) {
int32_t bytes = pBucket->bytes;
for (int32_t i = 0; i < size; ++i) {
char *d = (char *) data + i * bytes;
int32_t index = (pBucket->hashFunc)(pBucket, d);
if (index < 0) {
continue;

83
source/libs/nodes/src/nodesCodeFuncs.c
View File

@ -2816,10 +2816,85 @@ static int32_t jsonToTableNode(const SJson* pJson, void* pObj) {
return code;
}
static const char* jkTableIndexInfoIntervalUnit = "IntervalUnit";
static const char* jkTableIndexInfoSlidingUnit = "SlidingUnit";
static const char* jkTableIndexInfoInterval = "Interval";
static const char* jkTableIndexInfoOffset = "Offset";
static const char* jkTableIndexInfoSliding = "Sliding";
static const char* jkTableIndexInfoDstTbUid = "DstTbUid";
static const char* jkTableIndexInfoDstVgId = "DstVgId";
static const char* jkTableIndexInfoEpSet = "EpSet";
static const char* jkTableIndexInfoExpr = "Expr";
static int32_t tableIndexInfoToJson(const void* pObj, SJson* pJson) {
const STableIndexInfo* pNode = (const STableIndexInfo*)pObj;
int32_t code = tjsonAddIntegerToObject(pJson, jkTableIndexInfoIntervalUnit, pNode->intervalUnit);
if (TSDB_CODE_SUCCESS == code) {
code = tjsonAddIntegerToObject(pJson, jkTableIndexInfoSlidingUnit, pNode->slidingUnit);
}
if (TSDB_CODE_SUCCESS == code) {
code = tjsonAddIntegerToObject(pJson, jkTableIndexInfoInterval, pNode->interval);
}
if (TSDB_CODE_SUCCESS == code) {
code = tjsonAddIntegerToObject(pJson, jkTableIndexInfoOffset, pNode->offset);
}
if (TSDB_CODE_SUCCESS == code) {
code = tjsonAddIntegerToObject(pJson, jkTableIndexInfoSliding, pNode->sliding);
}
if (TSDB_CODE_SUCCESS == code) {
code = tjsonAddIntegerToObject(pJson, jkTableIndexInfoDstTbUid, pNode->dstTbUid);
}
if (TSDB_CODE_SUCCESS == code) {
code = tjsonAddIntegerToObject(pJson, jkTableIndexInfoDstVgId, pNode->dstVgId);
}
if (TSDB_CODE_SUCCESS == code) {
code = tjsonAddObject(pJson, jkTableIndexInfoEpSet, epSetToJson, &pNode->epSet);
}
if (TSDB_CODE_SUCCESS == code) {
code = tjsonAddStringToObject(pJson, jkTableIndexInfoExpr, pNode->expr);
}
return code;
}
static int32_t jsonToTableIndexInfo(const SJson* pJson, void* pObj) {
STableIndexInfo* pNode = (STableIndexInfo*)pObj;
int32_t code = tjsonGetTinyIntValue(pJson, jkTableIndexInfoIntervalUnit, &pNode->intervalUnit);
if (TSDB_CODE_SUCCESS == code) {
code = tjsonGetTinyIntValue(pJson, jkTableIndexInfoSlidingUnit, &pNode->slidingUnit);
}
if (TSDB_CODE_SUCCESS == code) {
code = tjsonGetBigIntValue(pJson, jkTableIndexInfoInterval, &pNode->interval);
}
if (TSDB_CODE_SUCCESS == code) {
code = tjsonGetBigIntValue(pJson, jkTableIndexInfoOffset, &pNode->offset);
}
if (TSDB_CODE_SUCCESS == code) {
code = tjsonGetBigIntValue(pJson, jkTableIndexInfoSliding, &pNode->sliding);
}
if (TSDB_CODE_SUCCESS == code) {
code = tjsonGetBigIntValue(pJson, jkTableIndexInfoDstTbUid, &pNode->dstTbUid);
}
if (TSDB_CODE_SUCCESS == code) {
code = tjsonGetIntValue(pJson, jkTableIndexInfoDstVgId, &pNode->dstVgId);
}
if (TSDB_CODE_SUCCESS == code) {
code = tjsonToObject(pJson, jkTableIndexInfoEpSet, jsonToEpSet, &pNode->epSet);
}
if (TSDB_CODE_SUCCESS == code) {
code = tjsonDupStringValue(pJson, jkTableIndexInfoExpr, &pNode->expr);
}
return code;
}
static const char* jkRealTableMetaSize = "MetaSize";
static const char* jkRealTableMeta = "Meta";
static const char* jkRealTableVgroupsInfoSize = "VgroupsInfoSize";
static const char* jkRealTableVgroupsInfo = "VgroupsInfo";
static const char* jkRealTableSmaIndexes = "SmaIndexes";
static int32_t realTableNodeToJson(const void* pObj, SJson* pJson) {
const SRealTableNode* pNode = (const SRealTableNode*)pObj;
@ -2837,6 +2912,9 @@ static int32_t realTableNodeToJson(const void* pObj, SJson* pJson) {
if (TSDB_CODE_SUCCESS == code) {
code = tjsonAddObject(pJson, jkRealTableVgroupsInfo, vgroupsInfoToJson, pNode->pVgroupList);
}
if (TSDB_CODE_SUCCESS == code) {
code = tjsonAddTArray(pJson, jkRealTableSmaIndexes, tableIndexInfoToJson, pNode->pSmaIndexes);
}
return code;
}
@ -2858,6 +2936,10 @@ static int32_t jsonToRealTableNode(const SJson* pJson, void* pObj) {
if (TSDB_CODE_SUCCESS == code) {
code = tjsonMakeObject(pJson, jkRealTableVgroupsInfo, jsonToVgroupsInfo, (void**)&pNode->pVgroupList, objSize);
}
if (TSDB_CODE_SUCCESS == code) {
code =
tjsonToTArray(pJson, jkRealTableSmaIndexes, jsonToTableIndexInfo, &pNode->pSmaIndexes, sizeof(STableIndexInfo));
}
return code;
}
@ -4214,6 +4296,7 @@ int32_t nodesStringToList(const char* pStr, SNodeList** pList) {
tjsonDelete(pJson);
if (TSDB_CODE_SUCCESS != code) {
nodesDestroyList(*pList);
*pList = NULL;
terrno = code;
return code;
}

3
source/libs/parser/inc/parUtil.h
View File

@ -46,6 +46,7 @@ typedef struct SParseMetaCache {
SHashObj* pDbInfo; // key is tbFName, element is SDbInfo*
SHashObj* pUserAuth; // key is SUserAuthInfo serialized string, element is bool indicating whether or not to pass
SHashObj* pUdf; // key is funcName, element is SFuncInfo*
SHashObj* pTableIndex; // key is tbFName, element is SArray<STableIndexInfo>*
} SParseMetaCache;
int32_t generateSyntaxErrMsg(SMsgBuf* pBuf, int32_t errCode, ...);
@ -75,6 +76,7 @@ int32_t reserveUserAuthInCache(int32_t acctId, const char* pUser, const char* pD
SParseMetaCache* pMetaCache);
int32_t reserveUserAuthInCacheExt(const char* pUser, const SName* pName, AUTH_TYPE type, SParseMetaCache* pMetaCache);
int32_t reserveUdfInCache(const char* pFunc, SParseMetaCache* pMetaCache);
int32_t reserveTableIndexInCache(int32_t acctId, const char* pDb, const char* pTable, SParseMetaCache* pMetaCache);
int32_t getTableMetaFromCache(SParseMetaCache* pMetaCache, const SName* pName, STableMeta** pMeta);
int32_t getDbVgInfoFromCache(SParseMetaCache* pMetaCache, const char* pDbFName, SArray** pVgInfo);
int32_t getTableVgroupFromCache(SParseMetaCache* pMetaCache, const SName* pName, SVgroupInfo* pVgroup);
@ -84,6 +86,7 @@ int32_t getDbCfgFromCache(SParseMetaCache* pMetaCache, const char* pDbFName, SDb
int32_t getUserAuthFromCache(SParseMetaCache* pMetaCache, const char* pUser, const char* pDbFName, AUTH_TYPE type,
bool* pPass);
int32_t getUdfInfoFromCache(SParseMetaCache* pMetaCache, const char* pFunc, SFuncInfo* pInfo);
int32_t getTableIndexFromCache(SParseMetaCache* pMetaCache, const SName* pName, SArray** pIndexes);
#ifdef __cplusplus
}

17
source/libs/parser/src/parAstParser.c
View File

@ -91,6 +91,7 @@ abort_parse:
typedef struct SCollectMetaKeyCxt {
SParseContext* pParseCxt;
SParseMetaCache* pMetaCache;
SNode* pStmt;
} SCollectMetaKeyCxt;
static void destroyCollectMetaKeyCxt(SCollectMetaKeyCxt* pCxt) {
@ -114,6 +115,14 @@ static EDealRes collectMetaKeyFromFunction(SCollectMetaKeyFromExprCxt* pCxt, SFu
return TSDB_CODE_SUCCESS == pCxt->errCode ? DEAL_RES_CONTINUE : DEAL_RES_ERROR;
}
static bool needGetTableIndex(SNode* pStmt) {
if (QUERY_NODE_SELECT_STMT == nodeType(pStmt)) {
SSelectStmt* pSelect = (SSelectStmt*)pStmt;
return (NULL != pSelect->pWindow && QUERY_NODE_INTERVAL_WINDOW == nodeType(pSelect->pWindow));
}
return false;
}
static int32_t collectMetaKeyFromRealTableImpl(SCollectMetaKeyCxt* pCxt, SRealTableNode* pRealTable,
AUTH_TYPE authType) {
int32_t code = reserveTableMetaInCache(pCxt->pParseCxt->acctId, pRealTable->table.dbName, pRealTable->table.tableName,
@ -129,6 +138,10 @@ static int32_t collectMetaKeyFromRealTableImpl(SCollectMetaKeyCxt* pCxt, SRealTa
if (TSDB_CODE_SUCCESS == code) {
code = reserveDbVgInfoInCache(pCxt->pParseCxt->acctId, pRealTable->table.dbName, pCxt->pMetaCache);
}
if (TSDB_CODE_SUCCESS == code && needGetTableIndex(pCxt->pStmt)) {
code = reserveTableIndexInCache(pCxt->pParseCxt->acctId, pRealTable->table.dbName, pRealTable->table.tableName,
pCxt->pMetaCache);
}
return code;
}
@ -379,6 +392,7 @@ static int32_t collectMetaKeyFromDelete(SCollectMetaKeyCxt* pCxt, SDeleteStmt* p
}
static int32_t collectMetaKeyFromQuery(SCollectMetaKeyCxt* pCxt, SNode* pStmt) {
pCxt->pStmt = pStmt;
switch (nodeType(pStmt)) {
case QUERY_NODE_SET_OPERATOR:
return collectMetaKeyFromSetOperator(pCxt, (SSetOperator*)pStmt);
@ -449,7 +463,8 @@ static int32_t collectMetaKeyFromQuery(SCollectMetaKeyCxt* pCxt, SNode* pStmt) {
}
int32_t collectMetaKey(SParseContext* pParseCxt, SQuery* pQuery) {
SCollectMetaKeyCxt cxt = {.pParseCxt = pParseCxt, .pMetaCache = taosMemoryCalloc(1, sizeof(SParseMetaCache))};
SCollectMetaKeyCxt cxt = {
.pParseCxt = pParseCxt, .pMetaCache = taosMemoryCalloc(1, sizeof(SParseMetaCache)), .pStmt = pQuery->pRoot};
if (NULL == cxt.pMetaCache) {
return TSDB_CODE_OUT_OF_MEMORY;
}

86
source/libs/parser/src/parTranslater.c
View File

@ -41,6 +41,7 @@ typedef struct STranslateContext {
SHashObj* pTables;
SExplainOptions* pExplainOpt;
SParseMetaCache* pMetaCache;
bool createStream;
} STranslateContext;
typedef struct SFullDatabaseName {
@ -255,6 +256,23 @@ static int32_t getUdfInfo(STranslateContext* pCxt, SFunctionNode* pFunc) {
return code;
}
static int32_t getTableIndex(STranslateContext* pCxt, const SName* pName, SArray** pIndexes) {
SParseContext* pParCxt = pCxt->pParseCxt;
int32_t code = collectUseDatabase(pName, pCxt->pDbs);
if (TSDB_CODE_SUCCESS == code) {
code = collectUseTable(pName, pCxt->pTables);
}
if (pParCxt->async) {
code = getTableIndexFromCache(pCxt->pMetaCache, pName, pIndexes);
} else {
code = catalogGetTableIndex(pParCxt->pCatalog, pParCxt->pTransporter, &pParCxt->mgmtEpSet, pName, pIndexes);
}
if (TSDB_CODE_SUCCESS != code) {
parserError("getTableIndex error, code:%s, dbName:%s, tbName:%s", tstrerror(code), pName->dbname, pName->tname);
}
return code;
}
static int32_t initTranslateContext(SParseContext* pParseCxt, SParseMetaCache* pMetaCache, STranslateContext* pCxt) {
pCxt->pParseCxt = pParseCxt;
pCxt->errCode = TSDB_CODE_SUCCESS;
@ -329,6 +347,10 @@ static bool isIndefiniteRowsFunc(const SNode* pNode) {
return (QUERY_NODE_FUNCTION == nodeType(pNode) && fmIsIndefiniteRowsFunc(((SFunctionNode*)pNode)->funcId));
}
static bool isVectorFunc(const SNode* pNode) {
return (QUERY_NODE_FUNCTION == nodeType(pNode) && fmIsVectorFunc(((SFunctionNode*)pNode)->funcId));
}
static bool isDistinctOrderBy(STranslateContext* pCxt) {
return (SQL_CLAUSE_ORDER_BY == pCxt->currClause && pCxt->pCurrSelectStmt->isDistinct);
}
@ -1364,6 +1386,17 @@ static bool isSingleTable(SRealTableNode* pRealTable) {
return (TSDB_CHILD_TABLE == tableType || TSDB_NORMAL_TABLE == tableType);
}
static int32_t setTableIndex(STranslateContext* pCxt, SName* pName, SRealTableNode* pRealTable) {
if (pCxt->createStream) {
return TSDB_CODE_SUCCESS;
}
if (NULL != pCxt->pCurrSelectStmt && NULL != pCxt->pCurrSelectStmt->pWindow &&
QUERY_NODE_INTERVAL_WINDOW == nodeType(pCxt->pCurrSelectStmt->pWindow)) {
return getTableIndex(pCxt, pName, &pRealTable->pSmaIndexes);
}
return TSDB_CODE_SUCCESS;
}
static int32_t translateTable(STranslateContext* pCxt, SNode* pTable) {
int32_t code = TSDB_CODE_SUCCESS;
switch (nodeType(pTable)) {
@ -1380,6 +1413,9 @@ static int32_t translateTable(STranslateContext* pCxt, SNode* pTable) {
return generateSyntaxErrMsg(&pCxt->msgBuf, TSDB_CODE_PAR_TABLE_NOT_EXIST, pRealTable->table.tableName);
}
code = setTableVgroupList(pCxt, &name, pRealTable);
if (TSDB_CODE_SUCCESS == code) {
code = setTableIndex(pCxt, &name, pRealTable);
}
}
pRealTable->table.precision = pRealTable->pMeta->tableInfo.precision;
pRealTable->table.singleTable = isSingleTable(pRealTable);
@ -1803,7 +1839,7 @@ static int64_t getMonthsFromTimeVal(int64_t val, int32_t fromPrecision, char uni
return -1;
}
static int32_t checkIntervalWindow(STranslateContext* pCxt, SNode* pWhere, SIntervalWindowNode* pInterval) {
static int32_t checkIntervalWindow(STranslateContext* pCxt, SIntervalWindowNode* pInterval) {
uint8_t precision = ((SColumnNode*)pInterval->pCol)->node.resType.precision;
SValueNode* pInter = (SValueNode*)pInterval->pInterval;
@ -1845,7 +1881,15 @@ static int32_t checkIntervalWindow(STranslateContext* pCxt, SNode* pWhere, SInte
}
}
return translateFill(pCxt, pWhere, pInterval);
return TSDB_CODE_SUCCESS;
}
static int32_t translateIntervalWindow(STranslateContext* pCxt, SSelectStmt* pSelect, SIntervalWindowNode* pInterval) {
int32_t code = checkIntervalWindow(pCxt, pInterval);
if (TSDB_CODE_SUCCESS == code) {
code = translateFill(pCxt, pSelect->pWhere, pInterval);
}
return code;
}
static EDealRes checkStateExpr(SNode* pNode, void* pContext) {
@ -1867,13 +1911,13 @@ static EDealRes checkStateExpr(SNode* pNode, void* pContext) {
return DEAL_RES_CONTINUE;
}
static int32_t checkStateWindow(STranslateContext* pCxt, SStateWindowNode* pState) {
static int32_t translateStateWindow(STranslateContext* pCxt, SStateWindowNode* pState) {
nodesWalkExprPostOrder(pState->pExpr, checkStateExpr, pCxt);
// todo check for "function not support for state_window"
return pCxt->errCode;
}
static int32_t checkSessionWindow(STranslateContext* pCxt, SSessionWindowNode* pSession) {
static int32_t translateSessionWindow(STranslateContext* pCxt, SSessionWindowNode* pSession) {
if ('y' == pSession->pGap->unit || 'n' == pSession->pGap->unit || 0 == pSession->pGap->datum.i) {
return generateSyntaxErrMsg(&pCxt->msgBuf, TSDB_CODE_PAR_INTER_SESSION_GAP);
}
@ -1884,14 +1928,14 @@ static int32_t checkSessionWindow(STranslateContext* pCxt, SSessionWindowNode* p
return TSDB_CODE_SUCCESS;
}
static int32_t checkWindow(STranslateContext* pCxt, SSelectStmt* pSelect) {
static int32_t translateSpecificWindow(STranslateContext* pCxt, SSelectStmt* pSelect) {
switch (nodeType(pSelect->pWindow)) {
case QUERY_NODE_STATE_WINDOW:
return checkStateWindow(pCxt, (SStateWindowNode*)pSelect->pWindow);
return translateStateWindow(pCxt, (SStateWindowNode*)pSelect->pWindow);
case QUERY_NODE_SESSION_WINDOW:
return checkSessionWindow(pCxt, (SSessionWindowNode*)pSelect->pWindow);
return translateSessionWindow(pCxt, (SSessionWindowNode*)pSelect->pWindow);
case QUERY_NODE_INTERVAL_WINDOW:
return checkIntervalWindow(pCxt, pSelect->pWhere, (SIntervalWindowNode*)pSelect->pWindow);
return translateIntervalWindow(pCxt, pSelect, (SIntervalWindowNode*)pSelect->pWindow);
default:
break;
}
@ -1905,7 +1949,7 @@ static int32_t translateWindow(STranslateContext* pCxt, SSelectStmt* pSelect) {
pCxt->currClause = SQL_CLAUSE_WINDOW;
int32_t code = translateExpr(pCxt, &pSelect->pWindow);
if (TSDB_CODE_SUCCESS == code) {
code = checkWindow(pCxt, pSelect);
code = translateSpecificWindow(pCxt, pSelect);
}
return code;
}
@ -2724,7 +2768,8 @@ typedef struct SSampleAstInfo {
STableMeta* pRollupTableMeta;
} SSampleAstInfo;
static int32_t buildSampleAst(STranslateContext* pCxt, SSampleAstInfo* pInfo, char** pAst, int32_t* pLen) {
static int32_t buildSampleAst(STranslateContext* pCxt, SSampleAstInfo* pInfo, char** pAst, int32_t* pLen, char** pExpr,
int32_t* pExprLen) {
SSelectStmt* pSelect = nodesMakeNode(QUERY_NODE_SELECT_STMT);
if (NULL == pSelect) {
return TSDB_CODE_OUT_OF_MEMORY;
@ -2769,10 +2814,14 @@ static int32_t buildSampleAst(STranslateContext* pCxt, SSampleAstInfo* pInfo, ch
((SColumnNode*)pInterval->pCol)->colId = PRIMARYKEY_TIMESTAMP_COL_ID;
strcpy(((SColumnNode*)pInterval->pCol)->colName, PK_TS_COL_INTERNAL_NAME);
pCxt->createStream = true;
int32_t code = translateQuery(pCxt, (SNode*)pSelect);
if (TSDB_CODE_SUCCESS == code) {
code = nodesNodeToString(pSelect, false, pAst, pLen);
}
if (TSDB_CODE_SUCCESS == code && NULL != pExpr) {
code = nodesListToString(pSelect->pProjectionList, false, pExpr, pExprLen);
}
nodesDestroyNode(pSelect);
return code;
}
@ -2894,7 +2943,7 @@ static int32_t getRollupAst(STranslateContext* pCxt, SCreateTableStmt* pStmt, SR
SSampleAstInfo info = {0};
int32_t code = buildSampleAstInfoByTable(pCxt, pStmt, pRetension, precision, &info);
if (TSDB_CODE_SUCCESS == code) {
code = buildSampleAst(pCxt, &info, pAst, pLen);
code = buildSampleAst(pCxt, &info, pAst, pLen, NULL, NULL);
}
clearSampleAstInfo(&info);
return code;
@ -3152,10 +3201,6 @@ static int32_t getSmaIndexSql(STranslateContext* pCxt, char** pSql, int32_t* pLe
return TSDB_CODE_SUCCESS;
}
static int32_t getSmaIndexExpr(STranslateContext* pCxt, SCreateIndexStmt* pStmt, char** pExpr, int32_t* pLen) {
return nodesListToString(pStmt->pOptions->pFuncs, false, pExpr, pLen);
}
static int32_t buildSampleAstInfoByIndex(STranslateContext* pCxt, SCreateIndexStmt* pStmt, SSampleAstInfo* pInfo) {
pInfo->pDbName = pCxt->pParseCxt->db;
pInfo->pTableName = pStmt->tableName;
@ -3171,11 +3216,12 @@ static int32_t buildSampleAstInfoByIndex(STranslateContext* pCxt, SCreateIndexSt
return TSDB_CODE_SUCCESS;
}
static int32_t getSmaIndexAst(STranslateContext* pCxt, SCreateIndexStmt* pStmt, char** pAst, int32_t* pLen) {
static int32_t getSmaIndexAst(STranslateContext* pCxt, SCreateIndexStmt* pStmt, char** pAst, int32_t* pLen,
char** pExpr, int32_t* pExprLen) {
SSampleAstInfo info = {0};
int32_t code = buildSampleAstInfoByIndex(pCxt, pStmt, &info);
if (TSDB_CODE_SUCCESS == code) {
code = buildSampleAst(pCxt, &info, pAst, pLen);
code = buildSampleAst(pCxt, &info, pAst, pLen, pExpr, pExprLen);
}
clearSampleAstInfo(&info);
return code;
@ -3201,10 +3247,7 @@ static int32_t buildCreateSmaReq(STranslateContext* pCxt, SCreateIndexStmt* pStm
code = getSmaIndexSql(pCxt, &pReq->sql, &pReq->sqlLen);
}
if (TSDB_CODE_SUCCESS == code) {
code = getSmaIndexExpr(pCxt, pStmt, &pReq->expr, &pReq->exprLen);
}
if (TSDB_CODE_SUCCESS == code) {
code = getSmaIndexAst(pCxt, pStmt, &pReq->ast, &pReq->astLen);
code = getSmaIndexAst(pCxt, pStmt, &pReq->ast, &pReq->astLen, &pReq->expr, &pReq->exprLen);
}
return code;
@ -3487,6 +3530,7 @@ static int32_t buildCreateStreamReq(STranslateContext* pCxt, SCreateStreamStmt*
tNameExtractFullName(&name, pReq->targetStbFullName);
}
pCxt->createStream = true;
int32_t code = translateQuery(pCxt, pStmt->pQuery);
if (TSDB_CODE_SUCCESS == code) {
getSourceDatabase(pStmt->pQuery, pCxt->pParseCxt->acctId, pReq->sourceDB);

46
source/libs/parser/src/parUtil.c
View File

@ -542,6 +542,9 @@ int32_t buildCatalogReq(const SParseMetaCache* pMetaCache, SCatalogReq* pCatalog
if (TSDB_CODE_SUCCESS == code) {
code = buildUdfReq(pMetaCache->pUdf, &pCatalogReq->pUdf);
}
if (TSDB_CODE_SUCCESS == code) {
code = buildTableReq(pMetaCache->pTableIndex, &pCatalogReq->pTableIndex);
}
return code;
}
@ -628,6 +631,9 @@ int32_t putMetaDataToCache(const SCatalogReq* pCatalogReq, const SMetaData* pMet
if (TSDB_CODE_SUCCESS == code) {
code = putUdfToCache(pCatalogReq->pUdf, pMetaData->pUdfList, pMetaCache->pUdf);
}
if (TSDB_CODE_SUCCESS == code) {
code = putTableDataToCache(pCatalogReq->pTableIndex, pMetaData->pTableIndex, pMetaCache->pTableIndex);
}
return code;
}
@ -806,3 +812,43 @@ int32_t getUdfInfoFromCache(SParseMetaCache* pMetaCache, const char* pFunc, SFun
}
return code;
}
static void destroySmaIndex(void* p) { taosMemoryFree(((STableIndexInfo*)p)->expr); }
static SArray* smaIndexesDup(SArray* pSrc) {
SArray* pDst = taosArrayDup(pSrc);
if (NULL == pDst) {
return NULL;
}
int32_t size = taosArrayGetSize(pDst);
for (int32_t i = 0; i < size; ++i) {
((STableIndexInfo*)taosArrayGet(pDst, i))->expr = NULL;
}
for (int32_t i = 0; i < size; ++i) {
STableIndexInfo* pIndex = taosArrayGet(pDst, i);
pIndex->expr = taosMemoryStrDup(((STableIndexInfo*)taosArrayGet(pSrc, i))->expr);
if (NULL == pIndex->expr) {
taosArrayDestroyEx(pDst, destroySmaIndex);
return NULL;
}
}
return pDst;
}
int32_t reserveTableIndexInCache(int32_t acctId, const char* pDb, const char* pTable, SParseMetaCache* pMetaCache) {
return reserveTableReqInCache(acctId, pDb, pTable, &pMetaCache->pTableIndex);
}
int32_t getTableIndexFromCache(SParseMetaCache* pMetaCache, const SName* pName, SArray** pIndexes) {
char fullName[TSDB_TABLE_FNAME_LEN];
tNameExtractFullName(pName, fullName);
SArray* pSmaIndexes = NULL;
int32_t code = getMetaDataFromHash(fullName, strlen(fullName), pMetaCache->pTableIndex, (void**)&pSmaIndexes);
if (TSDB_CODE_SUCCESS == code && NULL != pSmaIndexes) {
*pIndexes = smaIndexesDup(pSmaIndexes);
if (NULL == *pIndexes) {
code = TSDB_CODE_OUT_OF_MEMORY;
}
}
return code;
}

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

@ -186,7 +186,7 @@ int32_t qExtractResultSchema(const SNode* pRoot, int32_t* numOfCols, SSchema** p
return extractResultSchema(pRoot, numOfCols, pSchema);
}
int32_t qSetSTableIdForRSma(SNode* pStmt, int64_t uid) {
int32_t qSetSTableIdForRsma(SNode* pStmt, int64_t uid) {
if (QUERY_NODE_SELECT_STMT == nodeType(pStmt)) {
SNode* pTable = ((SSelectStmt*)pStmt)->pFromTable;
if (QUERY_NODE_REAL_TABLE == nodeType(pTable)) {

6
source/libs/parser/test/mockCatalog.cpp
View File

@ -214,6 +214,11 @@ int32_t __catalogRefreshGetTableMeta(SCatalog* pCatalog, void* pTransporter, con
int32_t __catalogRemoveTableMeta(SCatalog* pCtg, SName* pTableName) { return 0; }
int32_t __catalogGetTableIndex(SCatalog* pCtg, void* pTrans, const SEpSet* pMgmtEps, const SName* pName,
SArray** pRes) {
return g_mockCatalogService->catalogGetTableIndex(pName, pRes);
}
void initMetaDataEnv() {
g_mockCatalogService.reset(new MockCatalogService());
@ -230,6 +235,7 @@ void initMetaDataEnv() {
stub.set(catalogGetUdfInfo, __catalogGetUdfInfo);
stub.set(catalogRefreshGetTableMeta, __catalogRefreshGetTableMeta);
stub.set(catalogRemoveTableMeta, __catalogRemoveTableMeta);
stub.set(catalogGetTableIndex, __catalogGetTableIndex);
// {
// AddrAny any("libcatalog.so");
// std::map<std::string,void*> result;

75
source/libs/parser/test/mockCatalogService.cpp
View File

@ -149,6 +149,20 @@ class MockCatalogServiceImpl {
return TSDB_CODE_SUCCESS;
}
int32_t catalogGetTableIndex(const SName* pTableName, SArray** pIndexes) const {
char tbFName[TSDB_TABLE_FNAME_LEN] = {0};
tNameExtractFullName(pTableName, tbFName);
auto it = index_.find(tbFName);
if (index_.end() == it) {
return TSDB_CODE_SUCCESS;
}
*pIndexes = taosArrayInit(it->second.size(), sizeof(STableIndexInfo));
for (const auto& index : it->second) {
taosArrayPush(*pIndexes, &index);
}
return TSDB_CODE_SUCCESS;
}
int32_t catalogGetAllMeta(const SCatalogReq* pCatalogReq, SMetaData* pMetaData) const {
int32_t code = getAllTableMeta(pCatalogReq->pTableMeta, &pMetaData->pTableMeta);
if (TSDB_CODE_SUCCESS == code) {
@ -169,6 +183,9 @@ class MockCatalogServiceImpl {
if (TSDB_CODE_SUCCESS == code) {
code = getAllUdf(pCatalogReq->pUdf, &pMetaData->pUdfList);
}
if (TSDB_CODE_SUCCESS == code) {
code = getAllTableIndex(pCatalogReq->pTableIndex, &pMetaData->pTableIndex);
}
return code;
}
@ -176,7 +193,7 @@ class MockCatalogServiceImpl {
int32_t numOfColumns, int32_t numOfTags) {
builder_ = TableBuilder::createTableBuilder(tableType, numOfColumns, numOfTags);
meta_[db][tbname] = builder_->table();
meta_[db][tbname]->schema->uid = id_++;
meta_[db][tbname]->schema->uid = getNextId();
return *(builder_.get());
}
@ -187,14 +204,11 @@ class MockCatalogServiceImpl {
}
meta_[db][tbname].reset(new MockTableMeta());
meta_[db][tbname]->schema = table.release();
meta_[db][tbname]->schema->uid = id_++;
meta_[db][tbname]->schema->uid = getNextId();
meta_[db][tbname]->schema->tableType = TSDB_CHILD_TABLE;
SVgroupInfo vgroup = {vgid, 0, 0, {0}, 0};
addEpIntoEpSet(&vgroup.epSet, "dnode_1", 6030);
addEpIntoEpSet(&vgroup.epSet, "dnode_2", 6030);
addEpIntoEpSet(&vgroup.epSet, "dnode_3", 6030);
vgroup.epSet.inUse = 0;
genEpSet(&vgroup.epSet);
meta_[db][tbname]->vgs.emplace_back(vgroup);
// super table
@ -268,10 +282,39 @@ class MockCatalogServiceImpl {
udf_.insert(std::make_pair(func, info));
}
void createSmaIndex(const SMCreateSmaReq* pReq) {
STableIndexInfo info;
info.intervalUnit = pReq->intervalUnit;
info.slidingUnit = pReq->slidingUnit;
info.interval = pReq->interval;
info.offset = pReq->offset;
info.sliding = pReq->sliding;
info.dstTbUid = getNextId();
info.dstVgId = pReq->dstVgId;
genEpSet(&info.epSet);
info.expr = strdup(pReq->expr);
auto it = index_.find(pReq->stb);
if (index_.end() == it) {
index_.insert(std::make_pair(std::string(pReq->stb), std::vector<STableIndexInfo>{info}));
} else {
it->second.push_back(info);
}
}
private:
typedef std::map<std::string, std::shared_ptr<MockTableMeta>> TableMetaCache;
typedef std::map<std::string, TableMetaCache> DbMetaCache;
typedef std::map<std::string, std::shared_ptr<SFuncInfo>> UdfMetaCache;
typedef std::map<std::string, std::vector<STableIndexInfo>> IndexMetaCache;
uint64_t getNextId() { return id_++; }
void genEpSet(SEpSet* pEpSet) {
addEpIntoEpSet(pEpSet, "dnode_1", 6030);
addEpIntoEpSet(pEpSet, "dnode_2", 6030);
addEpIntoEpSet(pEpSet, "dnode_3", 6030);
pEpSet->inUse = 0;
}
std::string toDbname(const std::string& dbFullName) const {
std::string::size_type n = dbFullName.find(".");
@ -463,10 +506,24 @@ class MockCatalogServiceImpl {
return TSDB_CODE_SUCCESS;
}
int32_t getAllTableIndex(SArray* pTableIndex, SArray** pTableIndexData) const {
if (NULL != pTableIndex) {
int32_t num = taosArrayGetSize(pTableIndex);
*pTableIndexData = taosArrayInit(num, sizeof(SMetaRes));
for (int32_t i = 0; i < num; ++i) {
SMetaRes res = {0};
res.code = catalogGetTableIndex((const SName*)taosArrayGet(pTableIndex, i), (SArray**)(&res.pRes));
taosArrayPush(*pTableIndexData, &res);
}
}
return TSDB_CODE_SUCCESS;
}
uint64_t id_;
std::unique_ptr<TableBuilder> builder_;
DbMetaCache meta_;
UdfMetaCache udf_;
IndexMetaCache index_;
};
MockCatalogService::MockCatalogService() : impl_(new MockCatalogServiceImpl()) {}
@ -490,6 +547,8 @@ void MockCatalogService::createFunction(const std::string& func, int8_t funcType
impl_->createFunction(func, funcType, outputType, outputLen, bufSize);
}
void MockCatalogService::createSmaIndex(const SMCreateSmaReq* pReq) { impl_->createSmaIndex(pReq); }
int32_t MockCatalogService::catalogGetTableMeta(const SName* pTableName, STableMeta** pTableMeta) const {
return impl_->catalogGetTableMeta(pTableName, pTableMeta);
}
@ -510,6 +569,10 @@ int32_t MockCatalogService::catalogGetUdfInfo(const std::string& funcName, SFunc
return impl_->catalogGetUdfInfo(funcName, pInfo);
}
int32_t MockCatalogService::catalogGetTableIndex(const SName* pTableName, SArray** pIndexes) const {
return impl_->catalogGetTableIndex(pTableName, pIndexes);
}
int32_t MockCatalogService::catalogGetAllMeta(const SCatalogReq* pCatalogReq, SMetaData* pMetaData) const {
return impl_->catalogGetAllMeta(pCatalogReq, pMetaData);
}

2
source/libs/parser/test/mockCatalogService.h
View File

@ -57,12 +57,14 @@ class MockCatalogService {
void createSubTable(const std::string& db, const std::string& stbname, const std::string& tbname, int16_t vgid);
void showTables() const;
void createFunction(const std::string& func, int8_t funcType, int8_t outputType, int32_t outputLen, int32_t bufSize);
void createSmaIndex(const SMCreateSmaReq* pReq);
int32_t catalogGetTableMeta(const SName* pTableName, STableMeta** pTableMeta) const;
int32_t catalogGetTableHashVgroup(const SName* pTableName, SVgroupInfo* vgInfo) const;
int32_t catalogGetTableDistVgInfo(const SName* pTableName, SArray** pVgList) const;
int32_t catalogGetDBVgInfo(const char* pDbFName, SArray** pVgList) const;
int32_t catalogGetUdfInfo(const std::string& funcName, SFuncInfo* pInfo) const;
int32_t catalogGetTableIndex(const SName* pTableName, SArray** pIndexes) const;
int32_t catalogGetAllMeta(const SCatalogReq* pCatalogReq, SMetaData* pMetaData) const;
private:

10
source/libs/planner/inc/planInt.h
View File

@ -36,11 +36,13 @@ extern "C" {
#define planTrace(param, ...) qTrace("PLAN: " param, __VA_ARGS__)
int32_t generateUsageErrMsg(char* pBuf, int32_t len, int32_t errCode, ...);
int32_t createColumnByRewriteExps(SNodeList* pExprs, SNodeList** pList);
int32_t createColumnByRewriteExprs(SNodeList* pExprs, SNodeList** pList);
int32_t createColumnByRewriteExpr(SNode* pExpr, SNodeList** pList);
int32_t replaceLogicNode(SLogicSubplan* pSubplan, SLogicNode* pOld, SLogicNode* pNew);
int32_t createLogicPlan(SPlanContext* pCxt, SLogicNode** pLogicNode);
int32_t optimizeLogicPlan(SPlanContext* pCxt, SLogicNode* pLogicNode);
int32_t splitLogicPlan(SPlanContext* pCxt, SLogicNode* pLogicNode, SLogicSubplan** pLogicSubplan);
int32_t createLogicPlan(SPlanContext* pCxt, SLogicSubplan** pLogicSubplan);
int32_t optimizeLogicPlan(SPlanContext* pCxt, SLogicSubplan* pLogicSubplan);
int32_t splitLogicPlan(SPlanContext* pCxt, SLogicSubplan* pLogicSubplan);
int32_t scaleOutLogicPlan(SPlanContext* pCxt, SLogicSubplan* pLogicSubplan, SQueryLogicPlan** pLogicPlan);
int32_t createPhysiPlan(SPlanContext* pCxt, SQueryLogicPlan* pLogicPlan, SQueryPlan** pPlan, SArray* pExecNodeList);

60
source/libs/planner/src/planLogicCreater.c
View File

@ -220,6 +220,7 @@ static int32_t makeScanLogicNode(SLogicPlanContext* pCxt, SRealTableNode* pRealT
}
TSWAP(pScan->pVgroupList, pRealTable->pVgroupList);
TSWAP(pScan->pSmaIndexes, pRealTable->pSmaIndexes);
pScan->tableId = pRealTable->pMeta->uid;
pScan->stableId = pRealTable->pMeta->suid;
pScan->tableType = pRealTable->pMeta->tableType;
@ -272,10 +273,10 @@ static int32_t createScanLogicNode(SLogicPlanContext* pCxt, SSelectStmt* pSelect
// set output
if (TSDB_CODE_SUCCESS == code) {
code = createColumnByRewriteExps(pScan->pScanCols, &pScan->node.pTargets);
code = createColumnByRewriteExprs(pScan->pScanCols, &pScan->node.pTargets);
}
if (TSDB_CODE_SUCCESS == code) {
code = createColumnByRewriteExps(pScan->pScanPseudoCols, &pScan->node.pTargets);
code = createColumnByRewriteExprs(pScan->pScanPseudoCols, &pScan->node.pTargets);
}
if (TSDB_CODE_SUCCESS == code) {
@ -441,10 +442,10 @@ static int32_t createAggLogicNode(SLogicPlanContext* pCxt, SSelectStmt* pSelect,
// set the output
if (TSDB_CODE_SUCCESS == code && NULL != pAgg->pGroupKeys) {
code = createColumnByRewriteExps(pAgg->pGroupKeys, &pAgg->node.pTargets);
code = createColumnByRewriteExprs(pAgg->pGroupKeys, &pAgg->node.pTargets);
}
if (TSDB_CODE_SUCCESS == code && NULL != pAgg->pAggFuncs) {
code = createColumnByRewriteExps(pAgg->pAggFuncs, &pAgg->node.pTargets);
code = createColumnByRewriteExprs(pAgg->pAggFuncs, &pAgg->node.pTargets);
}
if (TSDB_CODE_SUCCESS == code) {
@ -475,7 +476,7 @@ static int32_t createIndefRowsFuncLogicNode(SLogicPlanContext* pCxt, SSelectStmt
// set the output
if (TSDB_CODE_SUCCESS == code) {
code = createColumnByRewriteExps(pIdfRowsFunc->pVectorFuncs, &pIdfRowsFunc->node.pTargets);
code = createColumnByRewriteExprs(pIdfRowsFunc->pVectorFuncs, &pIdfRowsFunc->node.pTargets);
}
if (TSDB_CODE_SUCCESS == code) {
@ -505,7 +506,7 @@ static int32_t createWindowLogicNodeFinalize(SLogicPlanContext* pCxt, SSelectStm
}
if (TSDB_CODE_SUCCESS == code) {
code = createColumnByRewriteExps(pWindow->pFuncs, &pWindow->node.pTargets);
code = createColumnByRewriteExprs(pWindow->pFuncs, &pWindow->node.pTargets);
}
pSelect->hasAggFuncs = false;
@ -776,7 +777,7 @@ static int32_t createDistinctLogicNode(SLogicPlanContext* pCxt, SSelectStmt* pSe
// set the output
if (TSDB_CODE_SUCCESS == code) {
code = createColumnByRewriteExps(pAgg->pGroupKeys, &pAgg->node.pTargets);
code = createColumnByRewriteExprs(pAgg->pGroupKeys, &pAgg->node.pTargets);
}
if (TSDB_CODE_SUCCESS == code) {
@ -918,7 +919,7 @@ static int32_t createSetOpAggLogicNode(SLogicPlanContext* pCxt, SSetOperator* pS
// set the output
if (TSDB_CODE_SUCCESS == code) {
code = createColumnByRewriteExps(pAgg->pGroupKeys, &pAgg->node.pTargets);
code = createColumnByRewriteExprs(pAgg->pGroupKeys, &pAgg->node.pTargets);
}
if (TSDB_CODE_SUCCESS == code) {
@ -1040,7 +1041,7 @@ static int32_t createDeleteScanLogicNode(SLogicPlanContext* pCxt, SDeleteStmt* p
// set output
if (TSDB_CODE_SUCCESS == code) {
code = createColumnByRewriteExps(pScan->pScanCols, &pScan->node.pTargets);
code = createColumnByRewriteExprs(pScan->pScanCols, &pScan->node.pTargets);
}
if (TSDB_CODE_SUCCESS == code) {
@ -1064,7 +1065,7 @@ static int32_t createDeleteAggLogicNode(SLogicPlanContext* pCxt, SDeleteStmt* pD
}
// set the output
if (TSDB_CODE_SUCCESS == code) {
code = createColumnByRewriteExps(pAgg->pAggFuncs, &pAgg->node.pTargets);
code = createColumnByRewriteExprs(pAgg->pAggFuncs, &pAgg->node.pTargets);
}
if (TSDB_CODE_SUCCESS == code) {
@ -1138,11 +1139,40 @@ static int32_t createQueryLogicNode(SLogicPlanContext* pCxt, SNode* pStmt, SLogi
return TSDB_CODE_FAILED;
}
int32_t createLogicPlan(SPlanContext* pCxt, SLogicNode** pLogicNode) {
static void doSetLogicNodeParent(SLogicNode* pNode, SLogicNode* pParent) {
pNode->pParent = pParent;
SNode* pChild;
FOREACH(pChild, pNode->pChildren) { doSetLogicNodeParent((SLogicNode*)pChild, pNode); }
}
static void setLogicNodeParent(SLogicNode* pNode) { doSetLogicNodeParent(pNode, NULL); }
int32_t createLogicPlan(SPlanContext* pCxt, SLogicSubplan** pLogicSubplan) {
SLogicPlanContext cxt = {.pPlanCxt = pCxt};
int32_t code = createQueryLogicNode(&cxt, pCxt->pAstRoot, pLogicNode);
if (TSDB_CODE_SUCCESS != code) {
SLogicSubplan* pSubplan = (SLogicSubplan*)nodesMakeNode(QUERY_NODE_LOGIC_SUBPLAN);
if (NULL == pSubplan) {
return TSDB_CODE_OUT_OF_MEMORY;
}
pSubplan->id.queryId = pCxt->queryId;
pSubplan->id.groupId = 1;
pSubplan->id.subplanId = 1;
int32_t code = createQueryLogicNode(&cxt, pCxt->pAstRoot, &pSubplan->pNode);
if (TSDB_CODE_SUCCESS == code) {
setLogicNodeParent(pSubplan->pNode);
if (QUERY_NODE_LOGIC_PLAN_VNODE_MODIFY == nodeType(pSubplan->pNode)) {
pSubplan->subplanType = SUBPLAN_TYPE_MODIFY;
} else {
pSubplan->subplanType = SUBPLAN_TYPE_SCAN;
}
}
if (TSDB_CODE_SUCCESS == code) {
*pLogicSubplan = pSubplan;
} else {
nodesDestroyNode(pSubplan);
}
return code;
}
return TSDB_CODE_SUCCESS;
}

294
source/libs/planner/src/planOptimizer.c
View File

@ -32,8 +32,7 @@ typedef struct SOptimizeContext {
bool optimized;
} SOptimizeContext;
typedef int32_t (*FMatch)(SOptimizeContext* pCxt, SLogicNode* pLogicNode);
typedef int32_t (*FOptimize)(SOptimizeContext* pCxt, SLogicNode* pLogicNode);
typedef int32_t (*FOptimize)(SOptimizeContext* pCxt, SLogicSubplan* pLogicSubplan);
typedef struct SOptimizeRule {
char* pName;
@ -109,7 +108,6 @@ static bool osdMayBeOptimized(SLogicNode* pNode) {
}
if (QUERY_NODE_LOGIC_PLAN_WINDOW == nodeType(pNode->pParent)) {
return true;
// return (WINDOW_TYPE_INTERVAL == ((SWindowLogicNode*)pNode->pParent)->winType);
}
return !osdHaveNormalCol(((SAggLogicNode*)pNode->pParent)->pGroupKeys);
}
@ -231,9 +229,9 @@ static void setScanWindowInfo(SScanLogicNode* pScan) {
}
}
static int32_t osdOptimize(SOptimizeContext* pCxt, SLogicNode* pLogicNode) {
static int32_t osdOptimize(SOptimizeContext* pCxt, SLogicSubplan* pLogicSubplan) {
SOsdInfo info = {0};
int32_t code = osdMatch(pCxt, pLogicNode, &info);
int32_t code = osdMatch(pCxt, pLogicSubplan->pNode, &info);
if (TSDB_CODE_SUCCESS == code && info.pScan) {
setScanWindowInfo((SScanLogicNode*)info.pScan);
}
@ -635,8 +633,8 @@ static int32_t cpdPushCondition(SOptimizeContext* pCxt, SLogicNode* pLogicNode)
return code;
}
static int32_t cpdOptimize(SOptimizeContext* pCxt, SLogicNode* pLogicNode) {
return cpdPushCondition(pCxt, pLogicNode);
static int32_t cpdOptimize(SOptimizeContext* pCxt, SLogicSubplan* pLogicSubplan) {
return cpdPushCondition(pCxt, pLogicSubplan->pNode);
}
static bool opkIsPrimaryKeyOrderBy(SNodeList* pSortKeys) {
@ -745,26 +743,292 @@ static int32_t opkOptimizeImpl(SOptimizeContext* pCxt, SSortLogicNode* pSort) {
return code;
}
static int32_t opkOptimize(SOptimizeContext* pCxt, SLogicNode* pLogicNode) {
SSortLogicNode* pSort = (SSortLogicNode*)optFindPossibleNode(pLogicNode, opkSortMayBeOptimized);
static int32_t opkOptimize(SOptimizeContext* pCxt, SLogicSubplan* pLogicSubplan) {
SSortLogicNode* pSort = (SSortLogicNode*)optFindPossibleNode(pLogicSubplan->pNode, opkSortMayBeOptimized);
if (NULL == pSort) {
return TSDB_CODE_SUCCESS;
}
return opkOptimizeImpl(pCxt, pSort);
}
static const SOptimizeRule optimizeRuleSet[] = {{.pName = "OptimizeScanData", .optimizeFunc = osdOptimize},
static bool smaOptMayBeOptimized(SLogicNode* pNode) {
if (QUERY_NODE_LOGIC_PLAN_SCAN != nodeType(pNode) || NULL == pNode->pParent ||
QUERY_NODE_LOGIC_PLAN_WINDOW != nodeType(pNode->pParent) ||
WINDOW_TYPE_INTERVAL != ((SWindowLogicNode*)pNode->pParent)->winType) {
return false;
}
SScanLogicNode* pScan = (SScanLogicNode*)pNode;
if (0 == pScan->interval || NULL == pScan->pSmaIndexes || NULL != pScan->node.pConditions) {
return false;
}
return true;
}
static int32_t smaOptCreateMerge(SLogicNode* pChild, SNodeList* pMergeKeys, SNodeList* pTargets, SLogicNode** pOutput) {
SMergeLogicNode* pMerge = nodesMakeNode(QUERY_NODE_LOGIC_PLAN_MERGE);
if (NULL == pMerge) {
return TSDB_CODE_OUT_OF_MEMORY;
}
pMerge->node.precision = pChild->precision;
pMerge->numOfChannels = 2;
pMerge->pMergeKeys = pMergeKeys;
pMerge->node.pTargets = pTargets;
pMerge->pInputs = nodesCloneList(pChild->pTargets);
if (NULL == pMerge->pInputs) {
nodesDestroyNode(pMerge);
return TSDB_CODE_OUT_OF_MEMORY;
}
*pOutput = (SLogicNode*)pMerge;
return TSDB_CODE_SUCCESS;
}
static int32_t smaOptRecombinationNode(SLogicSubplan* pLogicSubplan, SLogicNode* pInterval, SLogicNode* pMerge,
SLogicNode* pSmaScan) {
int32_t code = nodesListMakeAppend(&pMerge->pChildren, pInterval);
if (TSDB_CODE_SUCCESS == code) {
code = nodesListMakeAppend(&pMerge->pChildren, pSmaScan);
}
if (TSDB_CODE_SUCCESS == code) {
code = replaceLogicNode(pLogicSubplan, pInterval, pMerge);
pSmaScan->pParent = pMerge;
pInterval->pParent = pMerge;
}
return code;
}
static int32_t smaOptCreateSmaScan(SScanLogicNode* pScan, STableIndexInfo* pIndex, SNodeList* pCols,
SLogicNode** pOutput) {
SScanLogicNode* pSmaScan = nodesMakeNode(QUERY_NODE_LOGIC_PLAN_SCAN);
if (NULL == pSmaScan) {
return TSDB_CODE_OUT_OF_MEMORY;
}
pSmaScan->pScanCols = pCols;
pSmaScan->tableType = TSDB_SUPER_TABLE;
pSmaScan->tableId = pIndex->dstTbUid;
pSmaScan->stableId = pIndex->dstTbUid;
pSmaScan->scanType = SCAN_TYPE_TABLE;
pSmaScan->scanSeq[0] = pScan->scanSeq[0];
pSmaScan->scanSeq[1] = pScan->scanSeq[1];
pSmaScan->scanRange = pScan->scanRange;
pSmaScan->dataRequired = FUNC_DATA_REQUIRED_DATA_LOAD;
pSmaScan->pVgroupList = taosMemoryCalloc(1, sizeof(SVgroupsInfo) + sizeof(SVgroupInfo));
if (NULL == pSmaScan->pVgroupList) {
nodesDestroyNode(pSmaScan);
return TSDB_CODE_OUT_OF_MEMORY;
}
pSmaScan->pVgroupList->numOfVgroups = 1;
pSmaScan->pVgroupList->vgroups[0].vgId = pIndex->dstVgId;
memcpy(&(pSmaScan->pVgroupList->vgroups[0].epSet), &pIndex->epSet, sizeof(SEpSet));
*pOutput = (SLogicNode*)pSmaScan;
return TSDB_CODE_SUCCESS;
}
static bool smaOptEqualInterval(SWindowLogicNode* pWindow, STableIndexInfo* pIndex) {
if (pWindow->interval != pIndex->interval || pWindow->intervalUnit != pIndex->intervalUnit ||
pWindow->offset != pIndex->offset || pWindow->sliding != pIndex->sliding ||
pWindow->slidingUnit != pIndex->slidingUnit) {
return false;
}
// todo time range
return true;
}
// #define SMA_TABLE_NAME "#sma_table"
// #define SMA_COL_NAME_PREFIX "#sma_col_"
static SNode* smaOptCreateSmaCol(SNode* pFunc, uint64_t tableId, int32_t colId) {
SColumnNode* pCol = nodesMakeNode(QUERY_NODE_COLUMN);
if (NULL == pCol) {
return NULL;
}
pCol->tableId = tableId;
pCol->tableType = TSDB_SUPER_TABLE;
pCol->colId = colId;
pCol->colType = COLUMN_TYPE_COLUMN;
snprintf(pCol->colName, sizeof(pCol->colName), "#sma_col_%d", pCol->colId);
// strcpy(pCol->tableName, SMA_TABLE_NAME);
// strcpy(pCol->tableAlias, SMA_TABLE_NAME);
pCol->node.resType = ((SExprNode*)pFunc)->resType;
strcpy(pCol->node.aliasName, ((SExprNode*)pFunc)->aliasName);
return (SNode*)pCol;
}
static int32_t smaOptFindSmaFunc(SNode* pQueryFunc, SNodeList* pSmaFuncs) {
int32_t index = 0;
SNode* pSmaFunc = NULL;
FOREACH(pSmaFunc, pSmaFuncs) {
if (nodesEqualNode(pQueryFunc, pSmaFunc)) {
return index;
}
++index;
}
return -1;
}
static int32_t smaOptCreateSmaCols(SNodeList* pFuncs, uint64_t tableId, SNodeList* pSmaFuncs, SNodeList** pOutput,
int32_t* pWStrartIndex) {
SNodeList* pCols = NULL;
SNode* pFunc = NULL;
int32_t code = TSDB_CODE_SUCCESS;
int32_t index = 0;
*pWStrartIndex = -1;
FOREACH(pFunc, pFuncs) {
if (FUNCTION_TYPE_WSTARTTS == ((SFunctionNode*)pFunc)->funcType) {
*pWStrartIndex = index;
}
int32_t smaFuncIndex = smaOptFindSmaFunc(pFunc, pSmaFuncs);
if (smaFuncIndex < 0) {
break;
} else {
code = nodesListMakeStrictAppend(&pCols, smaOptCreateSmaCol(pFunc, tableId, smaFuncIndex + 2));
if (TSDB_CODE_SUCCESS != code) {
break;
}
}
++index;
}
if (TSDB_CODE_SUCCESS == code) {
*pOutput = pCols;
} else {
nodesDestroyList(pCols);
}
return code;
}
static int32_t smaOptCouldApplyIndex(SWindowLogicNode* pWindow, STableIndexInfo* pIndex, SNodeList** pCols,
int32_t* pWStrartIndex) {
if (!smaOptEqualInterval(pWindow, pIndex)) {
return TSDB_CODE_SUCCESS;
}
SNodeList* pSmaFuncs = NULL;
int32_t code = nodesStringToList(pIndex->expr, &pSmaFuncs);
if (TSDB_CODE_SUCCESS == code) {
code = smaOptCreateSmaCols(pWindow->pFuncs, pIndex->dstTbUid, pSmaFuncs, pCols, pWStrartIndex);
}
nodesDestroyList(pSmaFuncs);
return code;
}
static SNode* smaOptCreateWStartTs() {
SFunctionNode* pWStart = nodesMakeNode(QUERY_NODE_FUNCTION);
if (NULL == pWStart) {
return NULL;
}
strcpy(pWStart->functionName, "_wstartts");
snprintf(pWStart->node.aliasName, sizeof(pWStart->node.aliasName), "%s.%p", pWStart->functionName, pWStart);
if (TSDB_CODE_SUCCESS != fmGetFuncInfo(pWStart, NULL, 0)) {
nodesDestroyNode(pWStart);
return NULL;
}
return (SNode*)pWStart;
}
static int32_t smaOptCreateMergeKey(SNode* pCol, SNodeList** pMergeKeys) {
SOrderByExprNode* pMergeKey = nodesMakeNode(QUERY_NODE_ORDER_BY_EXPR);
if (NULL == pMergeKey) {
return TSDB_CODE_OUT_OF_MEMORY;
}
pMergeKey->pExpr = nodesCloneNode(pCol);
if (NULL == pMergeKey->pExpr) {
nodesDestroyNode(pMergeKey);
return TSDB_CODE_OUT_OF_MEMORY;
}
pMergeKey->order = ORDER_ASC;
pMergeKey->nullOrder = NULL_ORDER_FIRST;
return nodesListMakeStrictAppend(pMergeKeys, pMergeKey);
}
static int32_t smaOptRewriteInterval(SWindowLogicNode* pInterval, int32_t wstrartIndex, SNodeList** pMergeKeys) {
if (wstrartIndex < 0) {
SNode* pWStart = smaOptCreateWStartTs();
if (NULL == pWStart) {
return TSDB_CODE_OUT_OF_MEMORY;
}
int32_t code = createColumnByRewriteExpr(pWStart, &pInterval->node.pTargets);
if (TSDB_CODE_SUCCESS != code) {
nodesDestroyNode(pWStart);
return code;
}
wstrartIndex = LIST_LENGTH(pInterval->node.pTargets) - 1;
}
return smaOptCreateMergeKey(nodesListGetNode(pInterval->node.pTargets, wstrartIndex), pMergeKeys);
}
static int32_t smaOptApplyIndex(SLogicSubplan* pLogicSubplan, SScanLogicNode* pScan, STableIndexInfo* pIndex,
SNodeList* pSmaCols, int32_t wstrartIndex) {
SWindowLogicNode* pInterval = (SWindowLogicNode*)pScan->node.pParent;
SNodeList* pMergeTargets = nodesCloneList(pInterval->node.pTargets);
if (NULL == pMergeTargets) {
return TSDB_CODE_OUT_OF_MEMORY;
}
SLogicNode* pSmaScan = NULL;
SLogicNode* pMerge = NULL;
SNodeList* pMergeKeys = NULL;
int32_t code = smaOptRewriteInterval(pInterval, wstrartIndex, &pMergeKeys);
if (TSDB_CODE_SUCCESS == code) {
code = smaOptCreateSmaScan(pScan, pIndex, pSmaCols, &pSmaScan);
}
if (TSDB_CODE_SUCCESS == code) {
code = smaOptCreateMerge(pScan->node.pParent, pMergeKeys, pMergeTargets, &pMerge);
}
if (TSDB_CODE_SUCCESS == code) {
code = smaOptRecombinationNode(pLogicSubplan, pScan->node.pParent, pMerge, pSmaScan);
}
return code;
}
static void smaOptDestroySmaIndex(void* p) { taosMemoryFree(((STableIndexInfo*)p)->expr); }
static int32_t smaOptimizeImpl(SOptimizeContext* pCxt, SLogicSubplan* pLogicSubplan, SScanLogicNode* pScan) {
int32_t code = TSDB_CODE_SUCCESS;
int32_t nindexes = taosArrayGetSize(pScan->pSmaIndexes);
for (int32_t i = 0; i < nindexes; ++i) {
STableIndexInfo* pIndex = taosArrayGet(pScan->pSmaIndexes, i);
SNodeList* pSmaCols = NULL;
int32_t wstrartIndex = -1;
code = smaOptCouldApplyIndex((SWindowLogicNode*)pScan->node.pParent, pIndex, &pSmaCols, &wstrartIndex);
if (TSDB_CODE_SUCCESS == code && NULL != pSmaCols) {
code = smaOptApplyIndex(pLogicSubplan, pScan, pIndex, pSmaCols, wstrartIndex);
taosArrayDestroyEx(pScan->pSmaIndexes, smaOptDestroySmaIndex);
pScan->pSmaIndexes = NULL;
break;
}
}
return code;
}
static int32_t smaOptimize(SOptimizeContext* pCxt, SLogicSubplan* pLogicSubplan) {
SScanLogicNode* pScan = (SScanLogicNode*)optFindPossibleNode(pLogicSubplan->pNode, smaOptMayBeOptimized);
if (NULL == pScan) {
return TSDB_CODE_SUCCESS;
}
return smaOptimizeImpl(pCxt, pLogicSubplan, pScan);
}
// clang-format off
static const SOptimizeRule optimizeRuleSet[] = {
{.pName = "OptimizeScanData", .optimizeFunc = osdOptimize},
{.pName = "ConditionPushDown", .optimizeFunc = cpdOptimize},
{.pName = "OrderByPrimaryKey", .optimizeFunc = opkOptimize}};
{.pName = "OrderByPrimaryKey", .optimizeFunc = opkOptimize},
{.pName = "SmaIndex", .optimizeFunc = smaOptimize}
};
// clang-format on
static const int32_t optimizeRuleNum = (sizeof(optimizeRuleSet) / sizeof(SOptimizeRule));
static int32_t applyOptimizeRule(SPlanContext* pCxt, SLogicNode* pLogicNode) {
static int32_t applyOptimizeRule(SPlanContext* pCxt, SLogicSubplan* pLogicSubplan) {
SOptimizeContext cxt = {.pPlanCxt = pCxt, .optimized = false};
do {
cxt.optimized = false;
for (int32_t i = 0; i < optimizeRuleNum; ++i) {
int32_t code = optimizeRuleSet[i].optimizeFunc(&cxt, pLogicNode);
int32_t code = optimizeRuleSet[i].optimizeFunc(&cxt, pLogicSubplan);
if (TSDB_CODE_SUCCESS != code) {
return code;
}
@ -773,4 +1037,6 @@ static int32_t applyOptimizeRule(SPlanContext* pCxt, SLogicNode* pLogicNode) {
return TSDB_CODE_SUCCESS;
}
int32_t optimizeLogicPlan(SPlanContext* pCxt, SLogicNode* pLogicNode) { return applyOptimizeRule(pCxt, pLogicNode); }
int32_t optimizeLogicPlan(SPlanContext* pCxt, SLogicSubplan* pLogicSubplan) {
return applyOptimizeRule(pCxt, pLogicSubplan);
}

118
source/libs/planner/src/planSpliter.c
View File

@ -80,29 +80,12 @@ static int32_t splCreateExchangeNode(SSplitContext* pCxt, SLogicNode* pChild, SE
return TSDB_CODE_SUCCESS;
}
static int32_t splReplaceLogicNode(SLogicSubplan* pSubplan, SLogicNode* pOld, SLogicNode* pNew) {
if (NULL == pOld->pParent) {
pSubplan->pNode = (SLogicNode*)pNew;
return TSDB_CODE_SUCCESS;
}
SNode* pNode;
FOREACH(pNode, pOld->pParent->pChildren) {
if (nodesEqualNode(pNode, pOld)) {
REPLACE_NODE(pNew);
pNew->pParent = pOld->pParent;
return TSDB_CODE_SUCCESS;
}
}
return TSDB_CODE_PLAN_INTERNAL_ERROR;
}
static int32_t splCreateExchangeNodeForSubplan(SSplitContext* pCxt, SLogicSubplan* pSubplan, SLogicNode* pSplitNode,
ESubplanType subplanType) {
SExchangeLogicNode* pExchange = NULL;
int32_t code = splCreateExchangeNode(pCxt, pSplitNode, &pExchange);
if (TSDB_CODE_SUCCESS == code) {
code = splReplaceLogicNode(pSubplan, pSplitNode, (SLogicNode*)pExchange);
code = replaceLogicNode(pSubplan, pSplitNode, (SLogicNode*)pExchange);
}
if (TSDB_CODE_SUCCESS == code) {
pSubplan->subplanType = subplanType;
@ -282,7 +265,7 @@ static int32_t stbSplCreatePartWindowNode(SWindowLogicNode* pMergeWindow, SLogic
code = stbSplAppendWStart(pPartWin->pFuncs, &index);
}
if (TSDB_CODE_SUCCESS == code) {
code = createColumnByRewriteExps(pPartWin->pFuncs, &pPartWin->node.pTargets);
code = createColumnByRewriteExprs(pPartWin->pFuncs, &pPartWin->node.pTargets);
}
if (TSDB_CODE_SUCCESS == code) {
nodesDestroyNode(pMergeWindow->pTspk);
@ -328,7 +311,7 @@ static int32_t stbSplCreateMergeNode(SSplitContext* pCxt, SLogicSubplan* pSubpla
if (NULL == pSubplan) {
code = nodesListMakeAppend(&pSplitNode->pChildren, pMerge);
} else {
code = splReplaceLogicNode(pSubplan, pSplitNode, (SLogicNode*)pMerge);
code = replaceLogicNode(pSubplan, pSplitNode, (SLogicNode*)pMerge);
}
}
if (TSDB_CODE_SUCCESS != code) {
@ -442,7 +425,7 @@ static int32_t stbSplCreatePartAggNode(SAggLogicNode* pMergeAgg, SLogicNode** pO
if (TSDB_CODE_SUCCESS == code && NULL != pGroupKeys) {
pPartAgg->pGroupKeys = pGroupKeys;
code = createColumnByRewriteExps(pPartAgg->pGroupKeys, &pPartAgg->node.pTargets);
code = createColumnByRewriteExprs(pPartAgg->pGroupKeys, &pPartAgg->node.pTargets);
}
if (TSDB_CODE_SUCCESS == code && NULL != pGroupKeys) {
pMergeAgg->pGroupKeys = nodesCloneList(pPartAgg->node.pTargets);
@ -457,7 +440,7 @@ static int32_t stbSplCreatePartAggNode(SAggLogicNode* pMergeAgg, SLogicNode** pO
code = stbSplRewriteFuns(pFunc, &pPartAgg->pAggFuncs, &pMergeAgg->pAggFuncs);
}
if (TSDB_CODE_SUCCESS == code) {
code = createColumnByRewriteExps(pPartAgg->pAggFuncs, &pPartAgg->node.pTargets);
code = createColumnByRewriteExprs(pPartAgg->pAggFuncs, &pPartAgg->node.pTargets);
}
nodesDestroyList(pFunc);
@ -897,12 +880,56 @@ static int32_t unionDistinctSplit(SSplitContext* pCxt, SLogicSubplan* pSubplan)
return code;
}
typedef struct SSmaIndexSplitInfo {
SMergeLogicNode* pMerge;
SLogicSubplan* pSubplan;
} SSmaIndexSplitInfo;
static SLogicNode* smaIdxSplMatchByNode(SLogicNode* pNode) {
if (QUERY_NODE_LOGIC_PLAN_MERGE == nodeType(pNode) && LIST_LENGTH(pNode->pChildren) > 1) {
return pNode;
}
SNode* pChild;
FOREACH(pChild, pNode->pChildren) {
SLogicNode* pSplitNode = smaIdxSplMatchByNode((SLogicNode*)pChild);
if (NULL != pSplitNode) {
return pSplitNode;
}
}
return NULL;
}
static bool smaIdxSplFindSplitNode(SSplitContext* pCxt, SLogicSubplan* pSubplan, SSmaIndexSplitInfo* pInfo) {
SLogicNode* pSplitNode = smaIdxSplMatchByNode(pSubplan->pNode);
if (NULL != pSplitNode) {
pInfo->pMerge = (SMergeLogicNode*)pSplitNode;
pInfo->pSubplan = pSubplan;
}
return NULL != pSplitNode;
}
static int32_t smaIndexSplit(SSplitContext* pCxt, SLogicSubplan* pSubplan) {
SSmaIndexSplitInfo info = {0};
if (!splMatch(pCxt, pSubplan, 0, (FSplFindSplitNode)smaIdxSplFindSplitNode, &info)) {
return TSDB_CODE_SUCCESS;
}
int32_t code = unionSplitSubplan(pCxt, info.pSubplan, (SLogicNode*)info.pMerge);
if (TSDB_CODE_SUCCESS == code) {
info.pMerge->srcGroupId = pCxt->groupId;
}
++(pCxt->groupId);
pCxt->split = true;
return code;
}
// clang-format off
static const SSplitRule splitRuleSet[] = {
{.pName = "SuperTableSplit", .splitFunc = stableSplit},
{.pName = "SingleTableJoinSplit", .splitFunc = singleTableJoinSplit},
{.pName = "UnionAllSplit", .splitFunc = unionAllSplit},
{.pName = "UnionDistinctSplit", .splitFunc = unionDistinctSplit}
{.pName = "UnionDistinctSplit", .splitFunc = unionDistinctSplit},
{.pName = "SmaIndexSplit", .splitFunc = smaIndexSplit}
};
// clang-format on
@ -936,14 +963,6 @@ static int32_t applySplitRule(SPlanContext* pCxt, SLogicSubplan* pSubplan) {
return TSDB_CODE_SUCCESS;
}
static void doSetLogicNodeParent(SLogicNode* pNode, SLogicNode* pParent) {
pNode->pParent = pParent;
SNode* pChild;
FOREACH(pChild, pNode->pChildren) { doSetLogicNodeParent((SLogicNode*)pChild, pNode); }
}
static void setLogicNodeParent(SLogicNode* pNode) { doSetLogicNodeParent(pNode, NULL); }
static void setVgroupsInfo(SLogicNode* pNode, SLogicSubplan* pSubplan) {
if (QUERY_NODE_LOGIC_PLAN_SCAN == nodeType(pNode)) {
TSWAP(((SScanLogicNode*)pNode)->pVgroupList, pSubplan->pVgroupList);
@ -954,37 +973,10 @@ static void setVgroupsInfo(SLogicNode* pNode, SLogicSubplan* pSubplan) {
FOREACH(pChild, pNode->pChildren) { setVgroupsInfo((SLogicNode*)pChild, pSubplan); }
}
int32_t splitLogicPlan(SPlanContext* pCxt, SLogicNode* pLogicNode, SLogicSubplan** pLogicSubplan) {
SLogicSubplan* pSubplan = (SLogicSubplan*)nodesMakeNode(QUERY_NODE_LOGIC_SUBPLAN);
if (NULL == pSubplan) {
return TSDB_CODE_OUT_OF_MEMORY;
int32_t splitLogicPlan(SPlanContext* pCxt, SLogicSubplan* pLogicSubplan) {
if (QUERY_NODE_LOGIC_PLAN_VNODE_MODIFY == nodeType(pLogicSubplan->pNode)) {
setVgroupsInfo(pLogicSubplan->pNode, pLogicSubplan);
return TSDB_CODE_SUCCESS;
}
pSubplan->pNode = nodesCloneNode(pLogicNode);
if (NULL == pSubplan->pNode) {
nodesDestroyNode(pSubplan);
return TSDB_CODE_OUT_OF_MEMORY;
}
pSubplan->id.queryId = pCxt->queryId;
pSubplan->id.groupId = 1;
setLogicNodeParent(pSubplan->pNode);
int32_t code = TSDB_CODE_SUCCESS;
if (QUERY_NODE_LOGIC_PLAN_VNODE_MODIFY == nodeType(pLogicNode)) {
pSubplan->subplanType = SUBPLAN_TYPE_MODIFY;
TSWAP(((SVnodeModifyLogicNode*)pLogicNode)->pDataBlocks, ((SVnodeModifyLogicNode*)pSubplan->pNode)->pDataBlocks);
setVgroupsInfo(pSubplan->pNode, pSubplan);
} else {
pSubplan->subplanType = SUBPLAN_TYPE_SCAN;
code = applySplitRule(pCxt, pSubplan);
}
if (TSDB_CODE_SUCCESS == code) {
*pLogicSubplan = pSubplan;
} else {
nodesDestroyNode(pSubplan);
}
return code;
return applySplitRule(pCxt, pLogicSubplan);
}

36
source/libs/planner/src/planUtil.c
View File

@ -69,7 +69,7 @@ static EDealRes doCreateColumn(SNode* pNode, void* pContext) {
return DEAL_RES_CONTINUE;
}
int32_t createColumnByRewriteExps(SNodeList* pExprs, SNodeList** pList) {
int32_t createColumnByRewriteExprs(SNodeList* pExprs, SNodeList** pList) {
SCreateColumnCxt cxt = {.errCode = TSDB_CODE_SUCCESS, .pList = (NULL == *pList ? nodesMakeList() : *pList)};
if (NULL == cxt.pList) {
return TSDB_CODE_OUT_OF_MEMORY;
@ -85,3 +85,37 @@ int32_t createColumnByRewriteExps(SNodeList* pExprs, SNodeList** pList) {
}
return cxt.errCode;
}
int32_t createColumnByRewriteExpr(SNode* pExpr, SNodeList** pList) {
SCreateColumnCxt cxt = {.errCode = TSDB_CODE_SUCCESS, .pList = (NULL == *pList ? nodesMakeList() : *pList)};
if (NULL == cxt.pList) {
return TSDB_CODE_OUT_OF_MEMORY;
}
nodesWalkExpr(pExpr, doCreateColumn, &cxt);
if (TSDB_CODE_SUCCESS != cxt.errCode) {
nodesDestroyList(cxt.pList);
return cxt.errCode;
}
if (NULL == *pList) {
*pList = cxt.pList;
}
return cxt.errCode;
}
int32_t replaceLogicNode(SLogicSubplan* pSubplan, SLogicNode* pOld, SLogicNode* pNew) {
if (NULL == pOld->pParent) {
pSubplan->pNode = (SLogicNode*)pNew;
return TSDB_CODE_SUCCESS;
}
SNode* pNode;
FOREACH(pNode, pOld->pParent->pChildren) {
if (nodesEqualNode(pNode, pOld)) {
REPLACE_NODE(pNew);
pNew->pParent = pOld->pParent;
return TSDB_CODE_SUCCESS;
}
}
return TSDB_CODE_PLAN_INTERNAL_ERROR;
}

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

@ -26,16 +26,15 @@ static void dumpQueryPlan(SQueryPlan* pPlan) {
}
int32_t qCreateQueryPlan(SPlanContext* pCxt, SQueryPlan** pPlan, SArray* pExecNodeList) {
SLogicNode* pLogicNode = NULL;
SLogicSubplan* pLogicSubplan = NULL;
SQueryLogicPlan* pLogicPlan = NULL;
int32_t code = createLogicPlan(pCxt, &pLogicNode);
int32_t code = createLogicPlan(pCxt, &pLogicSubplan);
if (TSDB_CODE_SUCCESS == code) {
code = optimizeLogicPlan(pCxt, pLogicNode);
code = optimizeLogicPlan(pCxt, pLogicSubplan);
}
if (TSDB_CODE_SUCCESS == code) {
code = splitLogicPlan(pCxt, pLogicNode, &pLogicSubplan);
code = splitLogicPlan(pCxt, pLogicSubplan);
}
if (TSDB_CODE_SUCCESS == code) {
code = scaleOutLogicPlan(pCxt, pLogicSubplan, &pLogicPlan);
@ -47,7 +46,6 @@ int32_t qCreateQueryPlan(SPlanContext* pCxt, SQueryPlan** pPlan, SArray* pExecNo
dumpQueryPlan(*pPlan);
}
nodesDestroyNode(pLogicNode);
nodesDestroyNode(pLogicSubplan);
nodesDestroyNode(pLogicPlan);
terrno = code;

4
source/libs/planner/test/planOtherTest.cpp
View File

@ -42,7 +42,9 @@ TEST_F(PlanOtherTest, createStreamUseSTable) {
TEST_F(PlanOtherTest, createSmaIndex) {
useDb("root", "test");
run("create sma index index1 on t1 function(max(c1), min(c3 + 10), sum(c4)) interval(10s)");
run("CREATE SMA INDEX idx1 ON t1 FUNCTION(MAX(c1), MIN(c3 + 10), SUM(c4)) INTERVAL(10s)");
run("SELECT SUM(c4) FROM t1 INTERVAL(10s)");
}
TEST_F(PlanOtherTest, explain) {

43
source/libs/planner/test/planTestUtil.cpp
View File

@ -14,12 +14,14 @@
*/
#include "planTestUtil.h"
#include <getopt.h>
#include <algorithm>
#include <array>
#include "cmdnodes.h"
#include "mockCatalogService.h"
#include "parser.h"
#include "planInt.h"
@ -104,13 +106,12 @@ class PlannerTestBaseImpl {
SPlanContext cxt = {0};
setPlanContext(pQuery, &cxt);
SLogicNode* pLogicNode = nullptr;
doCreateLogicPlan(&cxt, &pLogicNode);
doOptimizeLogicPlan(&cxt, pLogicNode);
SLogicSubplan* pLogicSubplan = nullptr;
doSplitLogicPlan(&cxt, pLogicNode, &pLogicSubplan);
doCreateLogicPlan(&cxt, &pLogicSubplan);
doOptimizeLogicPlan(&cxt, pLogicSubplan);
doSplitLogicPlan(&cxt, pLogicSubplan);
SQueryLogicPlan* pLogicPlan = nullptr;
doScaleOutLogicPlan(&cxt, pLogicSubplan, &pLogicPlan);
@ -164,13 +165,12 @@ class PlannerTestBaseImpl {
SPlanContext cxt = {0};
setPlanContext(stmtEnv_.pQuery_, &cxt);
SLogicNode* pLogicNode = nullptr;
doCreateLogicPlan(&cxt, &pLogicNode);
doOptimizeLogicPlan(&cxt, pLogicNode);
SLogicSubplan* pLogicSubplan = nullptr;
doSplitLogicPlan(&cxt, pLogicNode, &pLogicSubplan);
doCreateLogicPlan(&cxt, &pLogicSubplan);
doOptimizeLogicPlan(&cxt, pLogicSubplan);
doSplitLogicPlan(&cxt, pLogicSubplan);
SQueryLogicPlan* pLogicPlan = nullptr;
doScaleOutLogicPlan(&cxt, pLogicSubplan, &pLogicPlan);
@ -324,19 +324,19 @@ class PlannerTestBaseImpl {
res_.ast_ = toString(pQuery->pRoot);
}
void doCreateLogicPlan(SPlanContext* pCxt, SLogicNode** pLogicNode) {
DO_WITH_THROW(createLogicPlan, pCxt, pLogicNode);
res_.rawLogicPlan_ = toString((SNode*)(*pLogicNode));
void doCreateLogicPlan(SPlanContext* pCxt, SLogicSubplan** pLogicSubplan) {
DO_WITH_THROW(createLogicPlan, pCxt, pLogicSubplan);
res_.rawLogicPlan_ = toString((SNode*)(*pLogicSubplan));
}
void doOptimizeLogicPlan(SPlanContext* pCxt, SLogicNode* pLogicNode) {
DO_WITH_THROW(optimizeLogicPlan, pCxt, pLogicNode);
res_.optimizedLogicPlan_ = toString((SNode*)pLogicNode);
void doOptimizeLogicPlan(SPlanContext* pCxt, SLogicSubplan* pLogicSubplan) {
DO_WITH_THROW(optimizeLogicPlan, pCxt, pLogicSubplan);
res_.optimizedLogicPlan_ = toString((SNode*)pLogicSubplan);
}
void doSplitLogicPlan(SPlanContext* pCxt, SLogicNode* pLogicNode, SLogicSubplan** pLogicSubplan) {
DO_WITH_THROW(splitLogicPlan, pCxt, pLogicNode, pLogicSubplan);
res_.splitLogicPlan_ = toString((SNode*)(*pLogicSubplan));
void doSplitLogicPlan(SPlanContext* pCxt, SLogicSubplan* pLogicSubplan) {
DO_WITH_THROW(splitLogicPlan, pCxt, pLogicSubplan);
res_.splitLogicPlan_ = toString((SNode*)(pLogicSubplan));
}
void doScaleOutLogicPlan(SPlanContext* pCxt, SLogicSubplan* pLogicSubplan, SQueryLogicPlan** pLogicPlan) {
@ -363,6 +363,7 @@ class PlannerTestBaseImpl {
} else if (QUERY_NODE_CREATE_INDEX_STMT == nodeType(pQuery->pRoot)) {
SMCreateSmaReq req = {0};
tDeserializeSMCreateSmaReq(pQuery->pCmdMsg->pMsg, pQuery->pCmdMsg->msgLen, &req);
g_mockCatalogService->createSmaIndex(&req);
nodesStringToNode(req.ast, &pCxt->pAstRoot);
pCxt->streamQuery = true;
} else if (QUERY_NODE_CREATE_STREAM_STMT == nodeType(pQuery->pRoot)) {

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

@ -201,8 +201,8 @@ void syncNodeRelease(SSyncNode* pNode);
// raft state change --------------
void syncNodeUpdateTerm(SSyncNode* pSyncNode, SyncTerm term);
void syncNodeBecomeFollower(SSyncNode* pSyncNode);
void syncNodeBecomeLeader(SSyncNode* pSyncNode);
void syncNodeBecomeFollower(SSyncNode* pSyncNode, const char* debugStr);
void syncNodeBecomeLeader(SSyncNode* pSyncNode, const char* debugStr);
void syncNodeCandidate2Leader(SSyncNode* pSyncNode);
void syncNodeFollower2Candidate(SSyncNode* pSyncNode);

7
source/libs/sync/inc/syncSnapshot.h
View File

@ -70,12 +70,13 @@ typedef struct SSyncSnapshotReceiver {
SyncTerm privateTerm;
SSyncNode *pSyncNode;
int32_t replicaIndex;
SRaftId fromId;
} SSyncSnapshotReceiver;
SSyncSnapshotReceiver *snapshotReceiverCreate(SSyncNode *pSyncNode, int32_t replicaIndex);
SSyncSnapshotReceiver *snapshotReceiverCreate(SSyncNode *pSyncNode, SRaftId fromId);
void snapshotReceiverDestroy(SSyncSnapshotReceiver *pReceiver);
void snapshotReceiverStart(SSyncSnapshotReceiver *pReceiver, SyncTerm privateTerm);
void snapshotReceiverStart(SSyncSnapshotReceiver *pReceiver, SyncTerm privateTerm, SRaftId fromId);
bool snapshotReceiverIsStart(SSyncSnapshotReceiver *pReceiver);
void snapshotReceiverStop(SSyncSnapshotReceiver *pReceiver, bool apply);
cJSON * snapshotReceiver2Json(SSyncSnapshotReceiver *pReceiver);

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

@ -150,7 +150,7 @@ int32_t syncNodeOnAppendEntriesCb(SSyncNode* ths, SyncAppendEntries* pMsg) {
"ths->state:%d, logOK:%d",
pMsg->term, ths->pRaftStore->currentTerm, ths->state, logOK);
syncNodeBecomeFollower(ths);
syncNodeBecomeFollower(ths, "from candidate by append entries");
// ret or reply?
return ret;
@ -380,18 +380,20 @@ int32_t syncNodeOnAppendEntriesCb(SSyncNode* ths, SyncAppendEntries* pMsg) {
// change isStandBy to normal
if (!isDrop) {
if (ths->state == TAOS_SYNC_STATE_LEADER) {
syncNodeBecomeLeader(ths);
syncNodeBecomeLeader(ths, "config change");
} else {
syncNodeBecomeFollower(ths);
syncNodeBecomeFollower(ths, "config change");
}
}
if (gRaftDetailLog) {
char* sOld = syncCfg2Str(&oldSyncCfg);
char* sNew = syncCfg2Str(&newSyncCfg);
sInfo("==config change== 0x11 old:%s new:%s isDrop:%d \n", sOld, sNew, isDrop);
taosMemoryFree(sOld);
taosMemoryFree(sNew);
}
}
// always call FpReConfigCb
if (ths->pFsm->FpReConfigCb != NULL) {
@ -399,10 +401,12 @@ int32_t syncNodeOnAppendEntriesCb(SSyncNode* ths, SyncAppendEntries* pMsg) {
cbMeta.currentTerm = ths->pRaftStore->currentTerm;
cbMeta.index = 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, newSyncCfg, cbMeta);
ths->pFsm->FpReConfigCb(ths->pFsm, &rpcMsg, cbMeta);
}
}
@ -469,7 +473,7 @@ static int32_t syncNodeMakeLogSame(SSyncNode* ths, SyncAppendEntries* pMsg) {
// delete confict entries
code = ths->pLogStore->syncLogTruncate(ths->pLogStore, delBegin);
ASSERT(code == 0);
sInfo("sync event log truncate, from %ld to %ld", delBegin, delEnd);
sInfo("sync event vgId:%d log truncate, from %ld to %ld", ths->vgId, delBegin, delEnd);
logStoreSimpleLog2("after syncNodeMakeLogSame", ths->pLogStore);
return code;
@ -571,7 +575,7 @@ int32_t syncNodeOnAppendEntriesSnapshotCb(SSyncNode* ths, SyncAppendEntries* pMs
if (condition) {
sTrace("recv SyncAppendEntries, candidate to follower");
syncNodeBecomeFollower(ths);
syncNodeBecomeFollower(ths, "from candidate by append entries");
// do not reply?
return ret;
}
@ -742,6 +746,18 @@ int32_t syncNodeOnAppendEntriesSnapshotCb(SSyncNode* ths, SyncAppendEntries* pMs
if (pMsg->commitIndex > ths->commitIndex) {
// has commit entry in local
if (pMsg->commitIndex <= ths->pLogStore->syncLogLastIndex(ths->pLogStore)) {
// advance commit index to sanpshot first
SSnapshot snapshot;
ths->pFsm->FpGetSnapshot(ths->pFsm, &snapshot);
if (snapshot.lastApplyIndex >= 0 && snapshot.lastApplyIndex > ths->commitIndex) {
SyncIndex commitBegin = ths->commitIndex;
SyncIndex commitEnd = snapshot.lastApplyIndex;
ths->commitIndex = snapshot.lastApplyIndex;
sInfo("sync event vgId:%d commit by snapshot from index:%ld to index:%ld, %s", ths->vgId, commitBegin,
commitEnd, syncUtilState2String(ths->state));
}
SyncIndex beginIndex = ths->commitIndex + 1;
SyncIndex endIndex = pMsg->commitIndex;

31
source/libs/sync/src/syncAppendEntriesReply.c
View File

@ -121,7 +121,7 @@ int32_t syncNodeOnAppendEntriesReplySnapshotCb(SSyncNode* ths, SyncAppendEntries
syncIndexMgrLog2("recv SyncAppendEntriesReply, before pNextIndex:", ths->pNextIndex);
syncIndexMgrLog2("recv SyncAppendEntriesReply, before pMatchIndex:", ths->pMatchIndex);
{
if (gRaftDetailLog) {
SSnapshot snapshot;
ths->pFsm->FpGetSnapshot(ths->pFsm, &snapshot);
sTrace("recv SyncAppendEntriesReply, before snapshot.lastApplyIndex:%ld, snapshot.lastApplyTerm:%lu",
@ -147,7 +147,10 @@ int32_t syncNodeOnAppendEntriesReplySnapshotCb(SSyncNode* ths, SyncAppendEntries
if (pMsg->success) {
// nextIndex' = [nextIndex EXCEPT ![i][j] = m.mmatchIndex + 1]
syncIndexMgrSetIndex(ths->pNextIndex, &(pMsg->srcId), pMsg->matchIndex + 1);
if (gRaftDetailLog) {
sTrace("update next match, index:%ld, success:%d", pMsg->matchIndex + 1, pMsg->success);
}
// matchIndex' = [matchIndex EXCEPT ![i][j] = m.mmatchIndex]
syncIndexMgrSetIndex(ths->pMatchIndex, &(pMsg->srcId), pMsg->matchIndex);
@ -159,7 +162,9 @@ int32_t syncNodeOnAppendEntriesReplySnapshotCb(SSyncNode* ths, SyncAppendEntries
} else {
SyncIndex nextIndex = syncIndexMgrGetIndex(ths->pNextIndex, &(pMsg->srcId));
sTrace("update next not match, begin, index:%ld, success:%d", nextIndex, pMsg->success);
if (gRaftDetailLog) {
sTrace("update next index not match, begin, index:%ld, success:%d", nextIndex, pMsg->success);
}
// notice! int64, uint64
if (nextIndex > SYNC_INDEX_BEGIN) {
@ -178,9 +183,23 @@ int32_t syncNodeOnAppendEntriesReplySnapshotCb(SSyncNode* ths, SyncAppendEntries
pMsg->privateTerm < pSender->privateTerm) {
snapshotSenderStart(pSender);
char host[128];
uint16_t port;
syncUtilU642Addr(pSender->pSyncNode->replicasId[pSender->replicaIndex].addr, host, sizeof(host), &port);
if (gRaftDetailLog) {
char* s = snapshotSender2Str(pSender);
sInfo("sync event snapshot send start sender first time, sender:%s", s);
sInfo(
"sync event vgId:%d snapshot send to %s:%d start sender first time, lastApplyIndex:%ld lastApplyTerm:%lu "
"sender:%s",
ths->vgId, host, port, pSender->snapshot.lastApplyIndex, pSender->snapshot.lastApplyTerm, s);
taosMemoryFree(s);
} else {
sInfo(
"sync event vgId:%d snapshot send to %s:%d start sender first time, lastApplyIndex:%ld "
"lastApplyTerm:%lu",
ths->vgId, host, port, pSender->snapshot.lastApplyIndex, pSender->snapshot.lastApplyTerm);
}
}
SyncIndex sentryIndex = pSender->snapshot.lastApplyIndex + 1;
@ -195,12 +214,14 @@ int32_t syncNodeOnAppendEntriesReplySnapshotCb(SSyncNode* ths, SyncAppendEntries
}
syncIndexMgrSetIndex(ths->pNextIndex, &(pMsg->srcId), nextIndex);
sTrace("update next not match, end, index:%ld, success:%d", nextIndex, pMsg->success);
if (gRaftDetailLog) {
sTrace("update next index not match, end, index:%ld, success:%d", nextIndex, pMsg->success);
}
}
syncIndexMgrLog2("recv SyncAppendEntriesReply, after pNextIndex:", ths->pNextIndex);
syncIndexMgrLog2("recv SyncAppendEntriesReply, after pMatchIndex:", ths->pMatchIndex);
{
if (gRaftDetailLog) {
SSnapshot snapshot;
ths->pFsm->FpGetSnapshot(ths->pFsm, &snapshot);
sTrace("recv SyncAppendEntriesReply, after snapshot.lastApplyIndex:%ld, snapshot.lastApplyTerm:%lu",

26
source/libs/sync/src/syncCommit.c
View File

@ -48,13 +48,28 @@ void syncMaybeAdvanceCommitIndex(SSyncNode* pSyncNode) {
syncIndexMgrLog2("==syncNodeMaybeAdvanceCommitIndex== pNextIndex", pSyncNode->pNextIndex);
syncIndexMgrLog2("==syncNodeMaybeAdvanceCommitIndex== pMatchIndex", pSyncNode->pMatchIndex);
// advance commit index to sanpshot first
SSnapshot snapshot;
pSyncNode->pFsm->FpGetSnapshot(pSyncNode->pFsm, &snapshot);
if (snapshot.lastApplyIndex > 0 && snapshot.lastApplyIndex > pSyncNode->commitIndex) {
SyncIndex commitBegin = pSyncNode->commitIndex;
SyncIndex commitEnd = snapshot.lastApplyIndex;
pSyncNode->commitIndex = snapshot.lastApplyIndex;
sInfo("sync event vgId:%d commit by snapshot from index:%ld to index:%ld, %s", pSyncNode->vgId,
pSyncNode->commitIndex, snapshot.lastApplyIndex, syncUtilState2String(pSyncNode->state));
}
// update commit index
SyncIndex newCommitIndex = pSyncNode->commitIndex;
for (SyncIndex index = pSyncNode->pLogStore->getLastIndex(pSyncNode->pLogStore); index > pSyncNode->commitIndex;
--index) {
for (SyncIndex index = syncNodeGetLastIndex(pSyncNode); index > pSyncNode->commitIndex; --index) {
bool agree = syncAgree(pSyncNode, index);
if (gRaftDetailLog) {
sTrace("syncMaybeAdvanceCommitIndex syncAgree:%d, index:%ld, pSyncNode->commitIndex:%ld", agree, index,
pSyncNode->commitIndex);
}
if (agree) {
// term
SSyncRaftEntry* pEntry = pSyncNode->pLogStore->getEntry(pSyncNode->pLogStore, index);
@ -64,17 +79,22 @@ void syncMaybeAdvanceCommitIndex(SSyncNode* pSyncNode) {
if (pEntry->term == pSyncNode->pRaftStore->currentTerm) {
// update commit index
newCommitIndex = index;
if (gRaftDetailLog) {
sTrace("syncMaybeAdvanceCommitIndex maybe to update, newCommitIndex:%ld commit, pSyncNode->commitIndex:%ld",
newCommitIndex, pSyncNode->commitIndex);
}
syncEntryDestory(pEntry);
break;
} else {
if (gRaftDetailLog) {
sTrace(
"syncMaybeAdvanceCommitIndex can not commit due to term not equal, pEntry->term:%lu, "
"pSyncNode->pRaftStore->currentTerm:%lu",
pEntry->term, pSyncNode->pRaftStore->currentTerm);
}
}
syncEntryDestory(pEntry);
}
@ -84,7 +104,9 @@ void syncMaybeAdvanceCommitIndex(SSyncNode* pSyncNode) {
SyncIndex beginIndex = pSyncNode->commitIndex + 1;
SyncIndex endIndex = newCommitIndex;
if (gRaftDetailLog) {
sTrace("syncMaybeAdvanceCommitIndex sync commit %ld", newCommitIndex);
}
// update commit index
pSyncNode->commitIndex = newCommitIndex;

2
source/libs/sync/src/syncEnv.c
View File

@ -40,7 +40,7 @@ int32_t syncEnvStart() {
// gSyncEnv = doSyncEnvStart(gSyncEnv);
gSyncEnv = doSyncEnvStart();
assert(gSyncEnv != NULL);
sTrace("syncEnvStart ok!");
sTrace("sync env start ok");
return ret;
}

124
source/libs/sync/src/syncMain.c
View File

@ -87,7 +87,9 @@ int64_t syncOpen(const SSyncInfo* pSyncInfo) {
SSyncNode* pSyncNode = syncNodeOpen(pSyncInfo);
assert(pSyncNode != NULL);
if (gRaftDetailLog) {
syncNodeLog2("syncNodeOpen open success", pSyncNode);
}
pSyncNode->rid = taosAddRef(tsNodeRefId, pSyncNode);
if (pSyncNode->rid < 0) {
@ -173,20 +175,37 @@ int32_t syncSetStandby(int64_t rid) {
int32_t syncReconfig(int64_t rid, const SSyncCfg* pSyncCfg) {
int32_t ret = 0;
char* configChange = syncCfg2Str((SSyncCfg*)pSyncCfg);
sInfo("==syncReconfig== newconfig:%s", configChange);
char* newconfig = syncCfg2Str((SSyncCfg*)pSyncCfg);
if (gRaftDetailLog) {
sInfo("==syncReconfig== newconfig:%s", newconfig);
}
SRpcMsg rpcMsg = {0};
rpcMsg.msgType = TDMT_SYNC_CONFIG_CHANGE;
rpcMsg.info.noResp = 1;
rpcMsg.contLen = strlen(configChange) + 1;
rpcMsg.contLen = strlen(newconfig) + 1;
rpcMsg.pCont = rpcMallocCont(rpcMsg.contLen);
snprintf(rpcMsg.pCont, rpcMsg.contLen, "%s", configChange);
taosMemoryFree(configChange);
snprintf(rpcMsg.pCont, rpcMsg.contLen, "%s", newconfig);
taosMemoryFree(newconfig);
ret = syncPropose(rid, &rpcMsg, false);
return ret;
}
int32_t syncReconfigRaw(int64_t rid, const SSyncCfg* pNewCfg, SRpcMsg* pRpcMsg) {
int32_t ret = 0;
char* newconfig = syncCfg2Str((SSyncCfg*)pNewCfg);
pRpcMsg->msgType = TDMT_SYNC_CONFIG_CHANGE;
pRpcMsg->info.noResp = 1;
pRpcMsg->contLen = strlen(newconfig) + 1;
pRpcMsg->pCont = rpcMallocCont(pRpcMsg->contLen);
snprintf(pRpcMsg->pCont, pRpcMsg->contLen, "%s", newconfig);
taosMemoryFree(newconfig);
return ret;
}
int32_t syncForwardToPeer(int64_t rid, const SRpcMsg* pMsg, bool isWeak) {
int32_t ret = syncPropose(rid, pMsg, isWeak);
return ret;
@ -374,13 +393,14 @@ void setHeartbeatTimerMS(int64_t rid, int32_t hbTimerMS) {
}
int32_t syncPropose(int64_t rid, const SRpcMsg* pMsg, bool isWeak) {
sTrace("syncPropose msgType:%d ", pMsg->msgType);
int32_t ret = TAOS_SYNC_PROPOSE_SUCCESS;
SSyncNode* pSyncNode = taosAcquireRef(tsNodeRefId, rid);
if (pSyncNode == NULL) return TAOS_SYNC_PROPOSE_OTHER_ERROR;
SSyncNode* pSyncNode = taosAcquireRef(tsNodeRefId, rid);
if (pSyncNode == NULL) {
return TAOS_SYNC_PROPOSE_OTHER_ERROR;
}
assert(rid == pSyncNode->rid);
sTrace("sync event vgId:%d propose msgType:%s", pSyncNode->vgId, TMSG_INFO(pMsg->msgType));
if (pSyncNode->state == TAOS_SYNC_STATE_LEADER) {
SRespStub stub;
@ -411,6 +431,8 @@ int32_t syncPropose(int64_t rid, const SRpcMsg* pMsg, bool isWeak) {
SSyncNode* syncNodeOpen(const SSyncInfo* pOldSyncInfo) {
SSyncInfo* pSyncInfo = (SSyncInfo*)pOldSyncInfo;
sInfo("sync event vgId:%d sync open", pSyncInfo->vgId);
SSyncNode* pSyncNode = (SSyncNode*)taosMemoryMalloc(sizeof(SSyncNode));
assert(pSyncNode != NULL);
memset(pSyncNode, 0, sizeof(SSyncNode));
@ -439,9 +461,11 @@ SSyncNode* syncNodeOpen(const SSyncInfo* pOldSyncInfo) {
assert(pSyncNode->pRaftCfg != NULL);
pSyncInfo->syncCfg = pSyncNode->pRaftCfg->cfg;
if (gRaftDetailLog) {
char* seralized = raftCfg2Str(pSyncNode->pRaftCfg);
sInfo("syncNodeOpen update config :%s", seralized);
taosMemoryFree(seralized);
}
raftCfgClose(pSyncNode->pRaftCfg);
}
@ -612,7 +636,7 @@ SSyncNode* syncNodeOpen(const SSyncInfo* pOldSyncInfo) {
}
// snapshot receivers
pSyncNode->pNewNodeReceiver = snapshotReceiverCreate(pSyncNode, 100);
pSyncNode->pNewNodeReceiver = snapshotReceiverCreate(pSyncNode, EMPTY_RAFT_ID);
// start in syncNodeStart
// start raft
@ -628,9 +652,7 @@ void syncNodeStart(SSyncNode* pSyncNode) {
// start raft
if (pSyncNode->replicaNum == 1) {
raftStoreNextTerm(pSyncNode->pRaftStore);
syncNodeBecomeLeader(pSyncNode);
syncNodeLog2("==state change become leader immediately==", pSyncNode);
syncNodeBecomeLeader(pSyncNode, "one replica start");
// Raft 3.6.2 Committing entries from previous terms
@ -638,41 +660,22 @@ void syncNodeStart(SSyncNode* pSyncNode) {
syncNodeAppendNoop(pSyncNode);
syncMaybeAdvanceCommitIndex(pSyncNode); // maybe only one replica
/*
sInfo("==syncNodeStart== RestoreFinish begin 1 replica tsem_wait %p", pSyncNode);
tsem_wait(&pSyncNode->restoreSem);
sInfo("==syncNodeStart== RestoreFinish end 1 replica tsem_wait %p", pSyncNode);
*/
/*
while (pSyncNode->restoreFinish != true) {
taosMsleep(10);
if (gRaftDetailLog) {
syncNodeLog2("==state change become leader immediately==", pSyncNode);
}
*/
sInfo("==syncNodeStart== restoreFinish ok 1 replica %p vgId:%d", pSyncNode, pSyncNode->vgId);
return;
}
syncNodeBecomeFollower(pSyncNode);
syncNodeBecomeFollower(pSyncNode, "first start");
// for test
int32_t ret = 0;
// int32_t ret = 0;
// ret = syncNodeStartPingTimer(pSyncNode);
assert(ret == 0);
// assert(ret == 0);
/*
sInfo("==syncNodeStart== RestoreFinish begin multi replica tsem_wait %p", pSyncNode);
tsem_wait(&pSyncNode->restoreSem);
sInfo("==syncNodeStart== RestoreFinish end multi replica tsem_wait %p", pSyncNode);
*/
/*
while (pSyncNode->restoreFinish != true) {
taosMsleep(10);
if (gRaftDetailLog) {
syncNodeLog2("==state change become leader immediately==", pSyncNode);
}
*/
sInfo("==syncNodeStart== restoreFinish ok multi replica %p vgId:%d", pSyncNode, pSyncNode->vgId);
}
void syncNodeStartStandBy(SSyncNode* pSyncNode) {
@ -687,6 +690,8 @@ void syncNodeStartStandBy(SSyncNode* pSyncNode) {
}
void syncNodeClose(SSyncNode* pSyncNode) {
sInfo("sync event vgId:%d sync close", pSyncNode->vgId);
int32_t ret;
assert(pSyncNode != NULL);
@ -1061,7 +1066,7 @@ char* syncNode2SimpleStr(const SSyncNode* pSyncNode) {
int len = 256;
char* s = (char*)taosMemoryMalloc(len);
snprintf(s, len,
"syncNode2SimpleStr vgId:%d currentTerm:%lu, commitIndex:%ld, state:%d %s, isStandBy:%d, "
"syncNode: vgId:%d currentTerm:%lu, commitIndex:%ld, state:%d %s, isStandBy:%d, "
"electTimerLogicClock:%lu, "
"electTimerLogicClockUser:%lu, "
"electTimerMS:%d, replicaNum:%d",
@ -1131,8 +1136,11 @@ void syncNodeUpdateConfig(SSyncNode* pSyncNode, SSyncCfg* newConfig, bool* isDro
}
raftCfgPersist(pSyncNode->pRaftCfg);
if (gRaftDetailLog) {
syncNodeLog2("==syncNodeUpdateConfig==", pSyncNode);
}
}
SSyncNode* syncNodeAcquire(int64_t rid) {
SSyncNode* pNode = taosAcquireRef(tsNodeRefId, rid);
@ -1149,12 +1157,14 @@ void syncNodeRelease(SSyncNode* pNode) { taosReleaseRef(tsNodeRefId, pNode->rid)
void syncNodeUpdateTerm(SSyncNode* pSyncNode, SyncTerm term) {
if (term > pSyncNode->pRaftStore->currentTerm) {
raftStoreSetTerm(pSyncNode->pRaftStore, term);
syncNodeBecomeFollower(pSyncNode);
syncNodeBecomeFollower(pSyncNode, "update term");
raftStoreClearVote(pSyncNode->pRaftStore);
}
}
void syncNodeBecomeFollower(SSyncNode* pSyncNode) {
void syncNodeBecomeFollower(SSyncNode* pSyncNode, const char* debugStr) {
sInfo("sync event vgId:%d become follower, %s", pSyncNode->vgId, debugStr);
// maybe clear leader cache
if (pSyncNode->state == TAOS_SYNC_STATE_LEADER) {
pSyncNode->leaderCache = EMPTY_RAFT_ID;
@ -1186,7 +1196,9 @@ void syncNodeBecomeFollower(SSyncNode* pSyncNode) {
// evoterLog |-> voterLog[i]]}
// /\ UNCHANGED <<messages, currentTerm, votedFor, candidateVars, logVars>>
//
void syncNodeBecomeLeader(SSyncNode* pSyncNode) {
void syncNodeBecomeLeader(SSyncNode* pSyncNode, const char* debugStr) {
sInfo("sync event vgId:%d become leader, %s", pSyncNode->vgId, debugStr);
// state change
pSyncNode->state = TAOS_SYNC_STATE_LEADER;
@ -1237,7 +1249,7 @@ void syncNodeBecomeLeader(SSyncNode* pSyncNode) {
void syncNodeCandidate2Leader(SSyncNode* pSyncNode) {
assert(pSyncNode->state == TAOS_SYNC_STATE_CANDIDATE);
assert(voteGrantedMajority(pSyncNode->pVotesGranted));
syncNodeBecomeLeader(pSyncNode);
syncNodeBecomeLeader(pSyncNode, "candidate to leader");
syncNodeLog2("==state change syncNodeCandidate2Leader==", pSyncNode);
@ -1260,14 +1272,14 @@ void syncNodeFollower2Candidate(SSyncNode* pSyncNode) {
void syncNodeLeader2Follower(SSyncNode* pSyncNode) {
assert(pSyncNode->state == TAOS_SYNC_STATE_LEADER);
syncNodeBecomeFollower(pSyncNode);
syncNodeBecomeFollower(pSyncNode, "leader to follower");
syncNodeLog2("==state change syncNodeLeader2Follower==", pSyncNode);
}
void syncNodeCandidate2Follower(SSyncNode* pSyncNode) {
assert(pSyncNode->state == TAOS_SYNC_STATE_CANDIDATE);
syncNodeBecomeFollower(pSyncNode);
syncNodeBecomeFollower(pSyncNode, "candidate to follower");
syncNodeLog2("==state change syncNodeCandidate2Follower==", pSyncNode);
}
@ -1467,10 +1479,12 @@ void syncNodeLog(SSyncNode* pObj) {
}
void syncNodeLog2(char* s, SSyncNode* pObj) {
if (gRaftDetailLog) {
char* serialized = syncNode2Str(pObj);
sTraceLong("syncNodeLog2 | len:%lu | %s | %s", strlen(serialized), s, serialized);
taosMemoryFree(serialized);
}
}
// ------ local funciton ---------
// enqueue message ----
@ -1724,9 +1738,10 @@ const char* syncStr(ESyncState state) {
int32_t syncNodeCommit(SSyncNode* ths, SyncIndex beginIndex, SyncIndex endIndex, uint64_t flag) {
int32_t code = 0;
ESyncState state = flag;
sInfo("sync event commit from index:%" PRId64 " to index:%" PRId64 ", %s", beginIndex, endIndex,
syncUtilState2String(state));
sInfo("sync event vgId:%d commit by wal from index:%" PRId64 " to index:%" PRId64 ", %s", ths->vgId, beginIndex,
endIndex, syncUtilState2String(state));
/*
// maybe execute by leader, skip snapshot
SSnapshot snapshot = {.data = NULL, .lastApplyIndex = -1, .lastApplyTerm = 0};
if (ths->pFsm->FpGetSnapshot != NULL) {
@ -1735,6 +1750,7 @@ int32_t syncNodeCommit(SSyncNode* ths, SyncIndex beginIndex, SyncIndex endIndex,
if (beginIndex <= snapshot.lastApplyIndex) {
beginIndex = snapshot.lastApplyIndex + 1;
}
*/
// execute fsm
if (ths->pFsm != NULL) {
@ -1791,18 +1807,20 @@ int32_t syncNodeCommit(SSyncNode* ths, SyncIndex beginIndex, SyncIndex endIndex,
// change isStandBy to normal
if (!isDrop) {
if (ths->state == TAOS_SYNC_STATE_LEADER) {
syncNodeBecomeLeader(ths);
syncNodeBecomeLeader(ths, "config change");
} else {
syncNodeBecomeFollower(ths);
syncNodeBecomeFollower(ths, "config change");
}
}
if (gRaftDetailLog) {
char* sOld = syncCfg2Str(&oldSyncCfg);
char* sNew = syncCfg2Str(&newSyncCfg);
sInfo("==config change== 0x11 old:%s new:%s isDrop:%d \n", sOld, sNew, isDrop);
taosMemoryFree(sOld);
taosMemoryFree(sNew);
}
}
// always call FpReConfigCb
if (ths->pFsm->FpReConfigCb != NULL) {
@ -1810,10 +1828,12 @@ int32_t syncNodeCommit(SSyncNode* ths, SyncIndex beginIndex, SyncIndex endIndex,
cbMeta.currentTerm = ths->pRaftStore->currentTerm;
cbMeta.index = 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, newSyncCfg, cbMeta);
ths->pFsm->FpReConfigCb(ths->pFsm, &rpcMsg, cbMeta);
}
}
@ -1824,7 +1844,7 @@ int32_t syncNodeCommit(SSyncNode* ths, SyncIndex beginIndex, SyncIndex endIndex,
ths->pFsm->FpRestoreFinishCb(ths->pFsm);
}
ths->restoreFinish = true;
sInfo("restore finish %p vgId:%d", ths, ths->vgId);
sInfo("sync event vgId:%d restore finish", ths->vgId);
}
}

5
source/libs/sync/src/syncRaftLog.c
View File

@ -14,6 +14,7 @@
*/
#include "syncRaftLog.h"
#include "syncRaftCfg.h"
#include "wal.h"
// refactor, log[0 .. n] ==> log[m .. n]
@ -161,7 +162,9 @@ static int32_t raftLogAppendEntry(struct SSyncLogStore* pLogStore, SSyncRaftEntr
walFsync(pWal, true);
sTrace("sync event write index:%" PRId64, pEntry->index);
sTrace("sync event vgId:%d write index:%ld, %s, isStandBy:%d, msgType:%s, originalRpcType:%s", pData->pSyncNode->vgId,
pEntry->index, syncUtilState2String(pData->pSyncNode->state), pData->pSyncNode->pRaftCfg->isStandBy,
TMSG_INFO(pEntry->msgType), TMSG_INFO(pEntry->originalRpcType));
return code;
}

3
source/libs/sync/src/syncReplication.c
View File

@ -122,7 +122,7 @@ int32_t syncNodeAppendEntriesPeersSnapshot(SSyncNode* pSyncNode) {
syncIndexMgrLog2("begin append entries peers pNextIndex:", pSyncNode->pNextIndex);
syncIndexMgrLog2("begin append entries peers pMatchIndex:", pSyncNode->pMatchIndex);
logStoreSimpleLog2("begin append entries peers LogStore:", pSyncNode->pLogStore);
{
if (gRaftDetailLog) {
SSnapshot snapshot;
pSyncNode->pFsm->FpGetSnapshot(pSyncNode->pFsm, &snapshot);
sTrace("begin append entries peers, snapshot.lastApplyIndex:%ld, snapshot.lastApplyTerm:%lu",
@ -201,7 +201,6 @@ int32_t syncNodeReplicate(SSyncNode* pSyncNode) {
}
int32_t syncNodeAppendEntries(SSyncNode* pSyncNode, const SRaftId* destRaftId, const SyncAppendEntries* pMsg) {
sTrace("syncNodeAppendEntries pSyncNode:%p ", pSyncNode);
int32_t ret = 0;
SRpcMsg rpcMsg;

164
source/libs/sync/src/syncSnapshot.c
View File

@ -20,7 +20,7 @@
#include "syncUtil.h"
#include "wal.h"
static void snapshotReceiverDoStart(SSyncSnapshotReceiver *pReceiver, SyncTerm privateTerm);
static void snapshotReceiverDoStart(SSyncSnapshotReceiver *pReceiver, SyncTerm privateTerm, SRaftId fromId);
//----------------------------------
SSyncSnapshotSender *snapshotSenderCreate(SSyncNode *pSyncNode, int32_t replicaIndex) {
@ -105,13 +105,23 @@ void snapshotSenderStart(SSyncSnapshotSender *pSender) {
syncSnapshotSend2RpcMsg(pMsg, &rpcMsg);
syncNodeSendMsgById(&(pMsg->destId), pSender->pSyncNode, &rpcMsg);
char *msgStr = syncSnapshotSend2Str(pMsg);
char host[128];
uint16_t port;
syncUtilU642Addr(pSender->pSyncNode->replicasId[pSender->replicaIndex].addr, host, sizeof(host), &port);
sTrace("sync event snapshot send to %s:%d begin seq:%d ack:%d lastApplyIndex:%ld lastApplyTerm:%lu send msg:%s", host,
port, pSender->seq, pSender->ack, pSender->snapshot.lastApplyIndex, pSender->snapshot.lastApplyTerm, msgStr);
if (gRaftDetailLog) {
char *msgStr = syncSnapshotSend2Str(pMsg);
sTrace(
"sync event vgId:%d snapshot send to %s:%d begin seq:%d ack:%d lastApplyIndex:%ld lastApplyTerm:%lu send "
"msg:%s",
pSender->pSyncNode->vgId, host, port, pSender->seq, pSender->ack, pSender->snapshot.lastApplyIndex,
pSender->snapshot.lastApplyTerm, msgStr);
taosMemoryFree(msgStr);
} else {
sTrace("sync event vgId:%d snapshot send to %s:%d begin seq:%d ack:%d lastApplyIndex:%ld lastApplyTerm:%lu",
pSender->pSyncNode->vgId, host, port, pSender->seq, pSender->ack, pSender->snapshot.lastApplyIndex,
pSender->snapshot.lastApplyTerm);
}
syncSnapshotSendDestroy(pMsg);
}
@ -183,10 +193,12 @@ void snapshotSenderStop(SSyncSnapshotSender *pSender) {
pSender->start = false;
if (gRaftDetailLog) {
char *s = snapshotSender2Str(pSender);
sInfo("snapshotSenderStop %s", s);
taosMemoryFree(s);
}
}
// when sender receiver ack, call this function to send msg from seq
// seq = ack + 1, already updated
@ -225,20 +237,29 @@ int32_t snapshotSend(SSyncSnapshotSender *pSender) {
syncSnapshotSend2RpcMsg(pMsg, &rpcMsg);
syncNodeSendMsgById(&(pMsg->destId), pSender->pSyncNode, &rpcMsg);
char *msgStr = syncSnapshotSend2Str(pMsg);
char host[128];
uint16_t port;
syncUtilU642Addr(pSender->pSyncNode->replicasId[pSender->replicaIndex].addr, host, sizeof(host), &port);
if (pSender->seq == SYNC_SNAPSHOT_SEQ_END) {
sTrace("sync event snapshot send to %s:%d finish seq:%d ack:%d lastApplyIndex:%ld lastApplyTerm:%lu send msg:%s",
host, port, pSender->seq, pSender->ack, pSender->snapshot.lastApplyIndex, pSender->snapshot.lastApplyTerm,
msgStr);
} else {
sTrace("sync event snapshot send to %s:%d sending seq:%d ack:%d lastApplyIndex:%ld lastApplyTerm:%lu send msg:%s",
host, port, pSender->seq, pSender->ack, pSender->snapshot.lastApplyIndex, pSender->snapshot.lastApplyTerm,
msgStr);
}
if (gRaftDetailLog) {
char *msgStr = syncSnapshotSend2Str(pMsg);
sTrace(
"sync event vgId:%d snapshot send to %s:%d finish seq:%d ack:%d lastApplyIndex:%ld lastApplyTerm:%lu send "
"msg:%s",
pSender->pSyncNode->vgId, host, port, pSender->seq, pSender->ack, pSender->snapshot.lastApplyIndex,
pSender->snapshot.lastApplyTerm, msgStr);
taosMemoryFree(msgStr);
} else {
sTrace("sync event vgId:%d snapshot send to %s:%d finish seq:%d ack:%d lastApplyIndex:%ld lastApplyTerm:%lu",
pSender->pSyncNode->vgId, host, port, pSender->seq, pSender->ack, pSender->snapshot.lastApplyIndex,
pSender->snapshot.lastApplyTerm);
}
} else {
sTrace("sync event vgId:%d snapshot send to %s:%d sending seq:%d ack:%d lastApplyIndex:%ld lastApplyTerm:%lu",
pSender->pSyncNode->vgId, host, port, pSender->seq, pSender->ack, pSender->snapshot.lastApplyIndex,
pSender->snapshot.lastApplyTerm);
}
syncSnapshotSendDestroy(pMsg);
return 0;
@ -260,13 +281,19 @@ int32_t snapshotReSend(SSyncSnapshotSender *pSender) {
syncSnapshotSend2RpcMsg(pMsg, &rpcMsg);
syncNodeSendMsgById(&(pMsg->destId), pSender->pSyncNode, &rpcMsg);
char *msgStr = syncSnapshotSend2Str(pMsg);
char host[128];
uint16_t port;
syncUtilU642Addr(pSender->pSyncNode->replicasId[pSender->replicaIndex].addr, host, sizeof(host), &port);
sTrace("sync event snapshot send to %s:%d resend seq:%d ack:%d send msg:%s", host, port, pSender->seq, pSender->ack,
msgStr);
if (gRaftDetailLog) {
char *msgStr = syncSnapshotSend2Str(pMsg);
sTrace("sync event vgId:%d snapshot send to %s:%d resend seq:%d ack:%d send msg:%s", pSender->pSyncNode->vgId,
host, port, pSender->seq, pSender->ack, msgStr);
taosMemoryFree(msgStr);
} else {
sTrace("sync event vgId:%d snapshot send to %s:%d resend seq:%d ack:%d", pSender->pSyncNode->vgId, host, port,
pSender->seq, pSender->ack);
}
syncSnapshotSendDestroy(pMsg);
}
@ -331,7 +358,7 @@ char *snapshotSender2Str(SSyncSnapshotSender *pSender) {
}
// -------------------------------------
SSyncSnapshotReceiver *snapshotReceiverCreate(SSyncNode *pSyncNode, int32_t replicaIndex) {
SSyncSnapshotReceiver *snapshotReceiverCreate(SSyncNode *pSyncNode, SRaftId fromId) {
bool condition = (pSyncNode->pFsm->FpSnapshotStartWrite != NULL) && (pSyncNode->pFsm->FpSnapshotStopWrite != NULL) &&
(pSyncNode->pFsm->FpSnapshotDoWrite != NULL);
@ -345,7 +372,7 @@ SSyncSnapshotReceiver *snapshotReceiverCreate(SSyncNode *pSyncNode, int32_t repl
pReceiver->ack = SYNC_SNAPSHOT_SEQ_BEGIN;
pReceiver->pWriter = NULL;
pReceiver->pSyncNode = pSyncNode;
pReceiver->replicaIndex = replicaIndex;
pReceiver->fromId = fromId;
pReceiver->term = pSyncNode->pRaftStore->currentTerm;
pReceiver->privateTerm = 0;
@ -365,10 +392,11 @@ 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) {
static void snapshotReceiverDoStart(SSyncSnapshotReceiver *pReceiver, SyncTerm privateTerm, SRaftId fromId) {
pReceiver->term = pReceiver->pSyncNode->pRaftStore->currentTerm;
pReceiver->privateTerm = privateTerm;
pReceiver->ack = SYNC_SNAPSHOT_SEQ_BEGIN;
pReceiver->fromId = fromId;
ASSERT(pReceiver->pWriter == NULL);
int32_t ret = pReceiver->pSyncNode->pFsm->FpSnapshotStartWrite(pReceiver->pSyncNode->pFsm, &(pReceiver->pWriter));
@ -377,14 +405,15 @@ 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) {
void snapshotReceiverStart(SSyncSnapshotReceiver *pReceiver, SyncTerm privateTerm, SRaftId fromId) {
if (!snapshotReceiverIsStart(pReceiver)) {
// start
snapshotReceiverDoStart(pReceiver, privateTerm);
snapshotReceiverDoStart(pReceiver, privateTerm, fromId);
pReceiver->start = true;
} else {
// already start
sInfo("snapshot recv, receiver already start");
// force close, abandon incomplete data
int32_t ret =
@ -393,16 +422,16 @@ void snapshotReceiverStart(SSyncSnapshotReceiver *pReceiver, SyncTerm privateTer
pReceiver->pWriter = NULL;
// start again
snapshotReceiverDoStart(pReceiver, privateTerm);
snapshotReceiverDoStart(pReceiver, privateTerm, fromId);
pReceiver->start = true;
ASSERT(0);
}
if (gRaftDetailLog) {
char *s = snapshotReceiver2Str(pReceiver);
sInfo("snapshotReceiverStart %s", s);
taosMemoryFree(s);
}
}
void snapshotReceiverStop(SSyncSnapshotReceiver *pReceiver, bool apply) {
if (pReceiver->pWriter != NULL) {
@ -418,10 +447,12 @@ void snapshotReceiverStop(SSyncSnapshotReceiver *pReceiver, bool apply) {
++(pReceiver->privateTerm);
}
if (gRaftDetailLog) {
char *s = snapshotReceiver2Str(pReceiver);
sInfo("snapshotReceiverStop %s", s);
taosMemoryFree(s);
}
}
cJSON *snapshotReceiver2Json(SSyncSnapshotReceiver *pReceiver) {
char u64buf[128];
@ -436,7 +467,22 @@ cJSON *snapshotReceiver2Json(SSyncSnapshotReceiver *pReceiver) {
snprintf(u64buf, sizeof(u64buf), "%p", pReceiver->pSyncNode);
cJSON_AddStringToObject(pRoot, "pSyncNode", u64buf);
cJSON_AddNumberToObject(pRoot, "replicaIndex", pReceiver->replicaIndex);
cJSON *pFromId = cJSON_CreateObject();
snprintf(u64buf, sizeof(u64buf), "%lu", pReceiver->fromId.addr);
cJSON_AddStringToObject(pFromId, "addr", u64buf);
{
uint64_t u64 = pReceiver->fromId.addr;
cJSON *pTmp = pFromId;
char host[128] = {0};
uint16_t port;
syncUtilU642Addr(u64, host, sizeof(host), &port);
cJSON_AddStringToObject(pTmp, "addr_host", host);
cJSON_AddNumberToObject(pTmp, "addr_port", port);
}
cJSON_AddNumberToObject(pFromId, "vgId", pReceiver->fromId.vgId);
cJSON_AddItemToObject(pRoot, "fromId", pFromId);
snprintf(u64buf, sizeof(u64buf), "%lu", pReceiver->term);
cJSON_AddStringToObject(pRoot, "term", u64buf);
@ -468,17 +514,23 @@ 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);
snapshotReceiverStart(pReceiver, pMsg->privateTerm, pMsg->srcId);
pReceiver->ack = pMsg->seq;
needRsp = true;
char *msgStr = syncSnapshotSend2Str(pMsg);
char host[128];
uint16_t port;
syncUtilU642Addr(pMsg->srcId.addr, host, sizeof(host), &port);
sTrace("sync event snapshot recv from %s:%d begin ack:%d, lastIndex:%ld, lastTerm:%lu, recv msg:%s", host, port,
pReceiver->ack, pMsg->lastIndex, pMsg->lastTerm, msgStr);
if (gRaftDetailLog) {
char *msgStr = syncSnapshotSend2Str(pMsg);
sTrace("sync event vgId:%d snapshot recv from %s:%d begin ack:%d, lastIndex:%ld, lastTerm:%lu, recv msg:%s",
pSyncNode->vgId, host, port, pReceiver->ack, pMsg->lastIndex, pMsg->lastTerm, msgStr);
taosMemoryFree(msgStr);
} else {
sTrace("sync event vgId:%d snapshot recv from %s:%d begin ack:%d, lastIndex:%ld, lastTerm:%lu",
pSyncNode->vgId, host, port, pReceiver->ack, pMsg->lastIndex, pMsg->lastTerm);
}
} else if (pMsg->seq == SYNC_SNAPSHOT_SEQ_END) {
// end, finish FSM
@ -486,29 +538,46 @@ int32_t syncNodeOnSnapshotSendCb(SSyncNode *pSyncNode, SyncSnapshotSend *pMsg) {
ASSERT(writeCode == 0);
pSyncNode->pFsm->FpSnapshotStopWrite(pSyncNode->pFsm, pReceiver->pWriter, true);
pSyncNode->pLogStore->syncLogSetBeginIndex(pSyncNode->pLogStore, pMsg->lastIndex + 1);
char *logSimpleStr = logStoreSimple2Str(pSyncNode->pLogStore);
SSnapshot snapshot;
pSyncNode->pFsm->FpGetSnapshot(pSyncNode->pFsm, &snapshot);
char host[128];
uint16_t port;
syncUtilU642Addr(pMsg->srcId.addr, host, sizeof(host), &port);
if (gRaftDetailLog) {
char *logSimpleStr = logStoreSimple2Str(pSyncNode->pLogStore);
sInfo(
"sync event snapshot recv from %s:%d finish, update log begin index:%ld, snapshot.lastApplyIndex:%ld, "
"sync event vgId:%d snapshot recv from %s:%d finish, update log begin index:%ld, "
"snapshot.lastApplyIndex:%ld, "
"snapshot.lastApplyTerm:%lu, raft log:%s",
host, port, pMsg->lastIndex + 1, snapshot.lastApplyIndex, snapshot.lastApplyTerm, logSimpleStr);
pSyncNode->vgId, host, port, pMsg->lastIndex + 1, snapshot.lastApplyIndex, snapshot.lastApplyTerm,
logSimpleStr);
taosMemoryFree(logSimpleStr);
} else {
sInfo(
"sync event vgId:%d snapshot recv from %s:%d finish, update log begin index:%ld, "
"snapshot.lastApplyIndex:%ld, "
"snapshot.lastApplyTerm:%lu",
pSyncNode->vgId, host, port, pMsg->lastIndex + 1, snapshot.lastApplyIndex, snapshot.lastApplyTerm);
}
pReceiver->pWriter = NULL;
snapshotReceiverStop(pReceiver, true);
pReceiver->ack = pMsg->seq;
needRsp = true;
if (gRaftDetailLog) {
char *msgStr = syncSnapshotSend2Str(pMsg);
sTrace("sync event snapshot recv from %s:%d end ack:%d, lastIndex:%ld, lastTerm:%lu, recv msg:%s", host, port,
pReceiver->ack, pMsg->lastIndex, pMsg->lastTerm, msgStr);
sTrace("sync event vgId:%d snapshot recv from %s:%d end ack:%d, lastIndex:%ld, lastTerm:%lu, recv msg:%s",
pReceiver->pSyncNode->vgId, host, port, pReceiver->ack, pMsg->lastIndex, pMsg->lastTerm, msgStr);
taosMemoryFree(msgStr);
} else {
sTrace("sync event vgId:%d snapshot recv from %s:%d end ack:%d, lastIndex:%ld, lastTerm:%lu",
pReceiver->pSyncNode->vgId, host, port, pReceiver->ack, pMsg->lastIndex, pMsg->lastTerm);
}
} else if (pMsg->seq == SYNC_SNAPSHOT_SEQ_FORCE_CLOSE) {
pSyncNode->pFsm->FpSnapshotStopWrite(pSyncNode->pFsm, pReceiver->pWriter, false);
@ -519,11 +588,17 @@ int32_t syncNodeOnSnapshotSendCb(SSyncNode *pSyncNode, SyncSnapshotSend *pMsg) {
uint16_t port;
syncUtilU642Addr(pMsg->srcId.addr, host, sizeof(host), &port);
if (gRaftDetailLog) {
char *msgStr = syncSnapshotSend2Str(pMsg);
sTrace("sync event snapshot recv from %s:%d force close ack:%d, lastIndex:%ld, lastTerm:%lu, recv msg:%s", host,
port, pReceiver->ack, pMsg->lastIndex, pMsg->lastTerm, msgStr);
sTrace(
"sync event vgId:%d snapshot recv from %s:%d force close ack:%d, lastIndex:%ld, lastTerm:%lu, recv "
"msg:%s",
pReceiver->pSyncNode->vgId, host, port, pReceiver->ack, pMsg->lastIndex, pMsg->lastTerm, msgStr);
taosMemoryFree(msgStr);
} else {
sTrace("sync event vgId:%d snapshot recv from %s:%d force close ack:%d, lastIndex:%ld, lastTerm:%lu",
pReceiver->pSyncNode->vgId, host, port, pReceiver->ack, pMsg->lastIndex, pMsg->lastTerm);
}
} else if (pMsg->seq > SYNC_SNAPSHOT_SEQ_BEGIN && pMsg->seq < SYNC_SNAPSHOT_SEQ_END) {
// transfering
@ -535,13 +610,20 @@ int32_t syncNodeOnSnapshotSendCb(SSyncNode *pSyncNode, SyncSnapshotSend *pMsg) {
}
needRsp = true;
char *msgStr = syncSnapshotSend2Str(pMsg);
char host[128];
uint16_t port;
syncUtilU642Addr(pMsg->srcId.addr, host, sizeof(host), &port);
sTrace("sync event snapshot recv from %s:%d receiving ack:%d, lastIndex:%ld, lastTerm:%lu, recv msg:%s", host,
port, pReceiver->ack, pMsg->lastIndex, pMsg->lastTerm, msgStr);
if (gRaftDetailLog) {
char *msgStr = syncSnapshotSend2Str(pMsg);
sTrace(
"sync event vgId:%d snapshot recv from %s:%d receiving ack:%d, lastIndex:%ld, lastTerm:%lu, recv msg:%s",
pSyncNode->vgId, host, port, pReceiver->ack, pMsg->lastIndex, pMsg->lastTerm, msgStr);
taosMemoryFree(msgStr);
} else {
sTrace("sync event vgId:%d snapshot recv from %s:%d receiving ack:%d, lastIndex:%ld, lastTerm:%lu",
pSyncNode->vgId, host, port, pReceiver->ack, pMsg->lastIndex, pMsg->lastTerm);
}
} else {
ASSERT(0);

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

@ -146,7 +146,7 @@ int32_t SnapshotDoWrite(struct SSyncFSM* pFsm, void* pWriter, void* pBuf, int32_
void RestoreFinishCb(struct SSyncFSM* pFsm) { sTrace("==callback== ==RestoreFinishCb=="); }
void ReConfigCb(struct SSyncFSM* pFsm, SSyncCfg newCfg, SReConfigCbMeta cbMeta) {
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);
}

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

@ -77,7 +77,7 @@ int32_t GetSnapshotCb(struct SSyncFSM* pFsm, SSnapshot* pSnapshot) {
void RestoreFinishCb(struct SSyncFSM* pFsm) { sTrace("==callback== ==RestoreFinishCb=="); }
void ReConfigCb(struct SSyncFSM* pFsm, SSyncCfg newCfg, SReConfigCbMeta cbMeta) {
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);
}

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

@ -41,7 +41,11 @@ SSyncSnapshotReceiver* createReceiver() {
pSyncNode->pFsm->FpSnapshotStopWrite = SnapshotStopWrite;
pSyncNode->pFsm->FpSnapshotDoWrite = SnapshotDoWrite;
SSyncSnapshotReceiver* pReceiver = snapshotReceiverCreate(pSyncNode, 2);
SRaftId id;
id.addr = syncUtilAddr2U64("1.2.3.4", 99);
id.vgId = 100;
SSyncSnapshotReceiver* pReceiver = snapshotReceiverCreate(pSyncNode, id);
pReceiver->start = true;
pReceiver->ack = 20;
pReceiver->pWriter = (void*)0x11;

5
source/libs/sync/test/syncTestTool.cpp
View File

@ -146,8 +146,8 @@ int32_t SnapshotDoWrite(struct SSyncFSM* pFsm, void* pWriter, void* pBuf, int32_
void RestoreFinishCb(struct SSyncFSM* pFsm) { sTrace("==callback== ==RestoreFinishCb== pFsm:%p", pFsm); }
void ReConfigCb(struct SSyncFSM* pFsm, SSyncCfg newCfg, SReConfigCbMeta cbMeta) {
char* s = syncCfg2Str(&newCfg);
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);
taosMemoryFree(s);
@ -235,7 +235,6 @@ int64_t createSyncNode(int32_t replicaNum, int32_t myIndex, int32_t vgId, SWal*
}
}
int64_t rid = syncOpen(&syncInfo);
assert(rid > 0);

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

@ -90,7 +90,6 @@ TAOS_DEFINE_ERROR(TSDB_CODE_RPC_AUTH_FAILURE, "Authentication failur
TAOS_DEFINE_ERROR(TSDB_CODE_RPC_NETWORK_UNAVAIL, "Unable to establish connection")
TAOS_DEFINE_ERROR(TSDB_CODE_RPC_FQDN_ERROR, "Unable to resolve FQDN")
TAOS_DEFINE_ERROR(TSDB_CODE_RPC_PORT_EADDRINUSE, "Port already in use")
TAOS_DEFINE_ERROR(TSDB_CODE_RPC_INDIRECT_NETWORK_UNAVAIL, "Unable to establish connection")
//client
TAOS_DEFINE_ERROR(TSDB_CODE_TSC_INVALID_OPERATION, "Invalid operation")
@ -225,6 +224,7 @@ TAOS_DEFINE_ERROR(TSDB_CODE_MND_VGROUP_NOT_IN_DNODE, "Vgroup not in dnode")
TAOS_DEFINE_ERROR(TSDB_CODE_MND_VGROUP_NOT_EXIST, "Vgroup does not exist")
TAOS_DEFINE_ERROR(TSDB_CODE_MND_VGROUP_UN_CHANGED, "Vgroup distribution has not changed")
TAOS_DEFINE_ERROR(TSDB_CODE_MND_HAS_OFFLINE_DNODE, "Offline dnode exists")
TAOS_DEFINE_ERROR(TSDB_CODE_MND_INVALID_REPLICA, "Invalid vgroup replica")
// mnode-stable
TAOS_DEFINE_ERROR(TSDB_CODE_MND_STB_ALREADY_EXIST, "STable already exists")
@ -265,6 +265,7 @@ TAOS_DEFINE_ERROR(TSDB_CODE_MND_TRANS_INVALID_STAGE, "Invalid stage to kill
TAOS_DEFINE_ERROR(TSDB_CODE_MND_TRANS_CONFLICT, "Conflict transaction not completed")
TAOS_DEFINE_ERROR(TSDB_CODE_MND_TRANS_UNKNOW_ERROR, "Unknown transaction error")
TAOS_DEFINE_ERROR(TSDB_CODE_MND_TRANS_CLOG_IS_NULL, "Transaction commitlog is null")
TAOS_DEFINE_ERROR(TSDB_CODE_MND_TRANS_NETWORK_UNAVAILL, "Unable to establish connection While execute transaction")
// mnode-mq
TAOS_DEFINE_ERROR(TSDB_CODE_MND_TOPIC_ALREADY_EXIST, "Topic already exists")

37
source/util/src/tjson.c
View File

@ -14,6 +14,7 @@
*/
#define _DEFAULT_SOURCE
#include "tjson.h"
#include "cJSON.h"
#include "taoserror.h"
@ -138,6 +139,23 @@ int32_t tjsonAddArray(SJson* pJson, const char* pName, FToJson func, const void*
return TSDB_CODE_SUCCESS;
}
int32_t tjsonAddTArray(SJson* pJson, const char* pName, FToJson func, const SArray* pArray) {
int32_t num = taosArrayGetSize(pArray);
if (num > 0) {
SJson* pJsonArray = tjsonAddArrayToObject(pJson, pName);
if (NULL == pJsonArray) {
return TSDB_CODE_OUT_OF_MEMORY;
}
for (int32_t i = 0; i < num; ++i) {
int32_t code = tjsonAddItem(pJsonArray, func, taosArrayGet(pArray, i));
if (TSDB_CODE_SUCCESS != code) {
return code;
}
}
}
return TSDB_CODE_SUCCESS;
}
char* tjsonToString(const SJson* pJson) { return cJSON_Print((cJSON*)pJson); }
char* tjsonToUnformattedString(const SJson* pJson) { return cJSON_PrintUnformatted((cJSON*)pJson); }
@ -299,6 +317,25 @@ int32_t tjsonToArray(const SJson* pJson, const char* pName, FToObject func, void
return TSDB_CODE_SUCCESS;
}
int32_t tjsonToTArray(const SJson* pJson, const char* pName, FToObject func, SArray** pArray, int32_t itemSize) {
const cJSON* jArray = tjsonGetObjectItem(pJson, pName);
int32_t size = tjsonGetArraySize(jArray);
if (size > 0) {
*pArray = taosArrayInit(size, itemSize);
if (NULL == *pArray) {
return TSDB_CODE_OUT_OF_MEMORY;
}
taosArraySetSize(*pArray, size);
for (int32_t i = 0; i < size; ++i) {
int32_t code = func(tjsonGetArrayItem(jArray, i), taosArrayGet(*pArray, i));
if (TSDB_CODE_SUCCESS != code) {
return code;
}
}
}
return TSDB_CODE_SUCCESS;
}
SJson* tjsonParse(const char* pStr) { return cJSON_Parse(pStr); }
bool tjsonValidateJson(const char* jIn) {

202
tests/script/tsim/dnode/redistribute_vgroup_replica3_v1_leader.sim Normal file
View File

@ -0,0 +1,202 @@
system sh/stop_dnodes.sh
system sh/deploy.sh -n dnode1 -i 1
system sh/deploy.sh -n dnode2 -i 2
system sh/deploy.sh -n dnode3 -i 3
system sh/deploy.sh -n dnode4 -i 4
system sh/deploy.sh -n dnode5 -i 5
system sh/cfg.sh -n dnode1 -c supportVnodes -v 0
system sh/exec.sh -n dnode1 -s start
system sh/exec.sh -n dnode2 -s start
system sh/exec.sh -n dnode3 -s start
system sh/exec.sh -n dnode4 -s start
#system sh/exec.sh -n dnode5 -s start
sql connect
sql create user u1 pass 'taosdata'
print =============== step1 create dnode2
sql create dnode $hostname port 7200
sql create dnode $hostname port 7300
sql create dnode $hostname port 7400
sql create dnode $hostname port 7500
$x = 0
step1:
$x = $x + 1
sleep 1000
if $x == 10 then
print ====> dnode not ready!
return -1
endi
sql show dnodes
print ===> $data00 $data01 $data02 $data03 $data04 $data05
print ===> $data10 $data11 $data12 $data13 $data14 $data15
print ===> $data20 $data21 $data22 $data23 $data24 $data25
print ===> $data30 $data31 $data32 $data33 $data34 $data35
print ===> $data40 $data41 $data42 $data43 $data44 $data45
if $rows != 5 then
return -1
endi
if $data(1)[4] != ready then
goto step1
endi
if $data(2)[4] != ready then
goto step1
endi
if $data(3)[4] != ready then
goto step1
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
# Invalid vgroup
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:
$x = $x + 1
sleep 1000
if $x == 10 then
print ====> dnode not ready!
return -1
endi
sql show dnodes
print ===> $data00 $data01 $data02 $data03 $data04 $data05
print ===> $data10 $data11 $data12 $data13 $data14 $data15
print ===> $data20 $data21 $data22 $data23 $data24 $data25
print ===> $data30 $data31 $data32 $data33 $data34 $data35
print ===> $data40 $data41 $data42 $data43 $data44 $data45
if $rows != 5 then
return -1
endi
if $data(1)[4] != ready then
goto step2
endi
if $data(2)[4] != ready then
goto step2
endi
if $data(3)[4] != ready then
goto step2
endi
if $data(4)[4] != ready then
goto step2
endi
if $data(5)[4] != ready then
goto step2
endi
print =============== step31: move follower
$leaderExist = 0
$leaderVnode = 0
$follower1 = 0
$follower2 = 0
$x = 0
step3:
$x = $x + 1
sleep 1000
if $x == 60 then
print ====> db not ready!
return -1
endi
sql show d1.vgroups
print ===> $data00 $data01 $data02 $data03 $data04 $data05 $data06 $data07 $data08 $data09
if $rows != 1 then
return -1
endi
if $data(2)[4] == leader then
$leaderExist = 1
$leaderVnode = 4
$follower1 = 2
$follower2 = 3
endi
if $data(2)[6] == leader then
$leaderExist = 1
$leaderVnode = 3
$follower1 = 2
$follower2 = 4
endi
if $data(2)[8] == leader then
$leaderExist = 1
$leaderVnode = 2
$follower1 = 3
$follower2 = 4
endi
if $leaderExist != 1 then
goto step3
endi
print leader $leaderVnode
print follower1 $follower1
print follower2 $follower2
sql use d1
sql create table d1.st (ts timestamp, i int) tags (j int)
sql create table d1.c1 using st tags(1)
sql show d1.tables
if $rows != 1 then
return -1
endi
print =============== step32: move follower2
print redistribute vgroup 2 dnode $leaderVnode dnode $follower2 dnode 5
sql redistribute vgroup 2 dnode $leaderVnode dnode $follower2 dnode 5
sql show d1.tables
if $rows != 1 then
return -1
endi
print =============== step33: move follower1
print redistribute vgroup 2 dnode $leaderVnode dnode $follower1 dnode 5
sql redistribute vgroup 2 dnode $leaderVnode dnode $follower1 dnode 5
sql show d1.tables
if $rows != 1 then
return -1
endi
print =============== step34: move follower2
print redistribute vgroup 2 dnode $leaderVnode dnode 5 dnode $follower2
sql redistribute vgroup 2 dnode $leaderVnode dnode 5 dnode $follower2
sql show d1.tables
if $rows != 1 then
return -1
endi
print =============== step35: move follower1
print redistribute vgroup 2 dnode $leaderVnode dnode 5 dnode $follower1
sql redistribute vgroup 2 dnode $leaderVnode dnode 5 dnode $follower1
sql show d1.tables
if $rows != 1 then
return -1
endi
print =============== step4: move leader
system sh/exec.sh -n dnode1 -s stop -x SIGINT
system sh/exec.sh -n dnode2 -s stop -x SIGINT
system sh/exec.sh -n dnode3 -s stop -x SIGINT
system sh/exec.sh -n dnode4 -s stop -x SIGINT
system sh/exec.sh -n dnode5 -s stop -x SIGINT

6
tests/system-test/1-insert/alter_stable.py
View File

@ -122,10 +122,10 @@ class TDTestCase:
for i in ['bigint','unsigned int','float','double','binary(10)','nchar(10)']:
for j in [1,2,3]:
tdSql.error(f'alter table {stbname} modify tag t{j} {i}')
tdSql.error(f'alter stable {stbname} modify tag t{j} {i}')
for i in ['int','unsigned int','float','binary(10)','nchar(10)']:
tdSql.error(f'alter table {stbname} modify tag t8 {i}')
tdSql.error(f'alter table {stbname} modify tag t4 int')
tdSql.error(f'alter stable {stbname} modify tag t8 {i}')
tdSql.error(f'alter stable {stbname} modify tag t4 int')
tdSql.execute(f'drop database {dbname}')
def run(self):

109
tests/system-test/1-insert/alter_table.py
View File

@ -145,10 +145,17 @@ class TDTestCase:
tdSql.execute(f'alter table {dbname}.{tbname} rename column c1 c21')
tdSql.query(f'describe {dbname}.{tbname}')
tdSql.checkData(1,0,'c21')
# !bug TD-16423
# tdSql.error(f'select c1 from {dbname}.{tbname}')
# tdSql.query(f'select c21 from {dbname}.{tbname}')
# tdSql.checkData(0,1,1)
tdSql.execute(f'alter table {dbname}.{tbname} rename column `c21` c1')
tdSql.query(f'describe {dbname}.{tbname}')
tdSql.checkData(1,0,'c1')
# !bug TD-16423
# tdSql.error(f'select c1 from {dbname}.{tbname}')
# tdSql.query(f'select c1 from {dbname}.{tbname}')
# tdSql.checkData(0,1,1)
tdSql.error(f'alter table {dbname}.{tbname} modify column c1 bigint')
tdSql.error(f'alter table {dbname}.{tbname} modify column c1 double')
tdSql.error(f'alter table {dbname}.{tbname} modify column c4 int')
@ -158,9 +165,109 @@ class TDTestCase:
tdSql.error(f'alter table {dbname}.{tbname} modify column c1 bool')
tdSql.error(f'alter table {dbname}.{tbname} modify column c1 binary(10)')
tdSql.execute(f'drop database {dbname}')
def alter_stb_column_check(self):
dbname = self.get_long_name(length=10, mode="letters")
tdSql.execute(f'create database if not exists {dbname}')
stbname = self.get_long_name(length=3, mode="letters")
tbname = self.get_long_name(length=3, mode="letters")
tdSql.execute(f'create database if not exists {dbname}')
tdSql.execute(f'use {dbname}')
tdSql.execute(
f'create table {stbname} (ts timestamp, c1 tinyint, c2 smallint, c3 int, \
c4 bigint, c5 tinyint unsigned, c6 smallint unsigned, c7 int unsigned, c8 bigint unsigned, c9 float, c10 double, c11 bool,c12 binary(20),c13 nchar(20)) tags(t0 int) ')
tdSql.execute(f'create table {tbname} using {stbname} tags(1)')
tdSql.execute(f'insert into {tbname} values (now,1,2,3,4,5,6,7,8,9.9,10.1,true,"abcd","涛思数据")')
tdSql.execute(f'alter table {stbname} add column c14 int')
tdSql.query(f'select c14 from {stbname}')
tdSql.checkRows(1)
tdSql.execute(f'alter table {stbname} add column `c15` int')
tdSql.query(f'select c15 from {stbname}')
tdSql.checkRows(1)
tdSql.query(f'describe {stbname}')
tdSql.checkRows(17)
tdSql.execute(f'alter table {stbname} drop column c14')
tdSql.query(f'describe {stbname}')
tdSql.checkRows(16)
tdSql.execute(f'alter table {stbname} drop column `c15`')
tdSql.query(f'describe {stbname}')
tdSql.checkRows(15)
tdSql.execute(f'alter table {stbname} modify column c12 binary(30)')
tdSql.query(f'describe {stbname}')
tdSql.checkData(12,2,30)
tdSql.execute(f'alter table {stbname} modify column `c12` binary(35)')
tdSql.query(f'describe {stbname}')
tdSql.checkData(12,2,35)
tdSql.error(f'alter table {stbname} modify column `c12` binary(34)')
tdSql.execute(f'alter table {stbname} modify column c13 nchar(30)')
tdSql.query(f'describe {stbname}')
tdSql.checkData(13,2,30)
tdSql.error(f'alter table {stbname} modify column c13 nchar(29)')
tdSql.error(f'alter table {stbname} rename column c1 c21')
tdSql.error(f'alter table {stbname} modify column c1 int')
tdSql.error(f'alter table {stbname} modify column c4 int')
tdSql.error(f'alter table {stbname} modify column c8 int')
tdSql.error(f'alter table {stbname} modify column c1 unsigned int')
tdSql.error(f'alter table {stbname} modify column c9 double')
tdSql.error(f'alter table {stbname} modify column c10 float')
tdSql.error(f'alter table {stbname} modify column c11 int')
tdSql.execute(f'drop database {dbname}')
def alter_stb_tag_check(self):
dbname = self.get_long_name(length=10, mode="letters")
tdSql.execute(f'create database if not exists {dbname}')
stbname = self.get_long_name(length=3, mode="letters")
tbname = self.get_long_name(length=3, mode="letters")
tdSql.execute(f'create database if not exists {dbname}')
tdSql.execute(f'use {dbname}')
tdSql.execute(
f'create table {stbname} (ts timestamp, c1 int) tags(ts_tag timestamp, t1 tinyint, t2 smallint, t3 int, \
t4 bigint, t5 tinyint unsigned, t6 smallint unsigned, t7 int unsigned, t8 bigint unsigned, t9 float, t10 double, t11 bool,t12 binary(20),t13 nchar(20)) ')
tdSql.execute(f'create table {tbname} using {stbname} tags(now,1,2,3,4,5,6,7,8,9.9,10.1,true,"abcd","涛思数据")')
tdSql.execute(f'insert into {tbname} values(now,1)')
tdSql.execute(f'alter table {stbname} add tag t14 int')
tdSql.query(f'select t14 from {stbname}')
tdSql.checkRows(1)
tdSql.execute(f'alter table {stbname} add tag `t15` int')
tdSql.query(f'select t14 from {stbname}')
tdSql.checkRows(1)
tdSql.query(f'describe {stbname}')
tdSql.checkRows(18)
tdSql.execute(f'alter table {stbname} drop tag t14')
tdSql.query(f'describe {stbname}')
tdSql.checkRows(17)
tdSql.execute(f'alter table {stbname} drop tag `t15`')
tdSql.query(f'describe {stbname}')
tdSql.checkRows(16)
tdSql.execute(f'alter table {stbname} modify tag t12 binary(30)')
tdSql.query(f'describe {stbname}')
tdSql.checkData(14,2,30)
tdSql.execute(f'alter table {stbname} modify tag `t12` binary(35)')
tdSql.query(f'describe {stbname}')
tdSql.checkData(14,2,35)
tdSql.error(f'alter table {stbname} modify tag `t12` binary(34)')
tdSql.execute(f'alter table {stbname} modify tag t13 nchar(30)')
tdSql.query(f'describe {stbname}')
tdSql.checkData(15,2,30)
tdSql.error(f'alter table {stbname} modify tag t13 nchar(29)')
tdSql.execute(f'alter table {stbname} rename tag t1 t21')
tdSql.query(f'describe {stbname}')
tdSql.checkData(3,0,'t21')
tdSql.execute(f'alter table {stbname} rename tag `t21` t1')
tdSql.query(f'describe {stbname}')
tdSql.checkData(3,0,'t1')
for i in ['bigint','unsigned int','float','double','binary(10)','nchar(10)']:
for j in [1,2,3]:
tdSql.error(f'alter table {stbname} modify tag t{j} {i}')
for i in ['int','unsigned int','float','binary(10)','nchar(10)']:
tdSql.error(f'alter table {stbname} modify tag t8 {i}')
tdSql.error(f'alter table {stbname} modify tag t4 int')
tdSql.execute(f'drop database {dbname}')
def run(self):
self.alter_tb_tag_check()
self.alter_ntb_column_check()
self.alter_stb_column_check()
self.alter_stb_tag_check()
def stop(self):
tdSql.close()

4
tests/system-test/1-insert/influxdb_line_taosc_insert.py
View File

@ -23,6 +23,10 @@ from util.sql import *
import threading
from util.types import TDSmlProtocolType, TDSmlTimestampType
from util.common import tdCom
import platform
import io
if platform.system().lower() == 'windows':
sys.stdout = io.TextIOWrapper(sys.stdout.buffer,encoding='utf8')
class TDTestCase:
def init(self, conn, logSql):

4
tests/system-test/1-insert/opentsdb_telnet_line_taosc_insert.py
View File

@ -22,6 +22,10 @@ from util.sql import *
from util.common import tdCom
from util.types import TDSmlProtocolType, TDSmlTimestampType
import threading
import platform
import io
if platform.system().lower() == 'windows':
sys.stdout = io.TextIOWrapper(sys.stdout.buffer,encoding='utf8')
class TDTestCase:
def init(self, conn, logSql):

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