From 8a9a2759a5ea73087a0dd84f0aadaa5ce5bbaa92 Mon Sep 17 00:00:00 2001 From: tomchon Date: Thu, 30 Jun 2022 21:28:47 +0800 Subject: [PATCH 1/7] test:modify testcase of stmt --- .../1-insert/test_stmt_insert_query_ex.py | 48 +++++++++---------- .../1-insert/test_stmt_muti_insert_query.py | 12 ++--- 2 files changed, 30 insertions(+), 30 deletions(-) diff --git a/tests/system-test/1-insert/test_stmt_insert_query_ex.py b/tests/system-test/1-insert/test_stmt_insert_query_ex.py index 6d49f6065a..dee7ee7451 100644 --- a/tests/system-test/1-insert/test_stmt_insert_query_ex.py +++ b/tests/system-test/1-insert/test_stmt_insert_query_ex.py @@ -82,7 +82,7 @@ class TDTestCase: return con def test_stmt_set_tbname_tag(self,conn): - dbname = "stmt_set_tbname_tag" + dbname = "stmt_tag" try: conn.execute("drop database if exists %s" % dbname) @@ -196,31 +196,31 @@ class TDTestCase: assert rows3[0][0] == 12, 'fourth case is failed' assert rows3[1][0] == 12, 'fourth case is failed' - # #query: conversion Functions + #query: conversion Functions - # querystmt4=conn.statement("select cast( ? as bigint) from log ") - # queryparam4=new_bind_params(1) - # print(type(queryparam4)) - # queryparam4[0].binary('1232a') - # querystmt4.bind_param(queryparam4) - # querystmt4.execute() - # result4=querystmt4.use_result() - # rows4=result4.fetch_all() - # print("5",rows4) - # assert rows4[0][0] == 1232 - # assert rows4[1][0] == 1232 + querystmt4=conn.statement("select cast( ? as bigint) from log ") + queryparam4=new_bind_params(1) + print(type(queryparam4)) + queryparam4[0].binary('1232a') + querystmt4.bind_param(queryparam4) + querystmt4.execute() + result4=querystmt4.use_result() + rows4=result4.fetch_all() + print("5",rows4) + assert rows4[0][0] == 1232 + assert rows4[1][0] == 1232 - # querystmt4=conn.statement("select cast( ? as binary(10)) from log ") - # queryparam4=new_bind_params(1) - # print(type(queryparam4)) - # queryparam4[0].int(123) - # querystmt4.bind_param(queryparam4) - # querystmt4.execute() - # result4=querystmt4.use_result() - # rows4=result4.fetch_all() - # print("6",rows4) - # assert rows4[0][0] == '123' - # assert rows4[1][0] == '123' + querystmt4=conn.statement("select cast( ? as binary(10)) from log ") + queryparam4=new_bind_params(1) + print(type(queryparam4)) + queryparam4[0].int(123) + querystmt4.bind_param(queryparam4) + querystmt4.execute() + result4=querystmt4.use_result() + rows4=result4.fetch_all() + print("6",rows4) + assert rows4[0][0] == '123' + assert rows4[1][0] == '123' # #query: datatime Functions diff --git a/tests/system-test/1-insert/test_stmt_muti_insert_query.py b/tests/system-test/1-insert/test_stmt_muti_insert_query.py index 971be6c85f..de10b2f6b9 100644 --- a/tests/system-test/1-insert/test_stmt_muti_insert_query.py +++ b/tests/system-test/1-insert/test_stmt_muti_insert_query.py @@ -84,21 +84,21 @@ class TDTestCase: def test_stmt_insert_multi(self,conn): # type: (TaosConnection) -> None - dbname = "pytest_taos_stmt_multi" + dbname = "db_stmt" try: conn.execute("drop database if exists %s" % dbname) conn.execute("create database if not exists %s" % dbname) conn.select_db(dbname) conn.execute( - "create table if not exists log(ts timestamp, bo bool, nil tinyint, ti tinyint, si smallint, ii int,\ + "create table if not exists stb1(ts timestamp, bo bool, nil tinyint, ti tinyint, si smallint, ii int,\ bi bigint, tu tinyint unsigned, su smallint unsigned, iu int unsigned, bu bigint unsigned, \ ff float, dd double, bb binary(100), nn nchar(100), tt timestamp)", ) # conn.load_table_info("log") start = datetime.now() - stmt = conn.statement("insert into log values(?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)") + stmt = conn.statement("insert into stb1 values(?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)") params = new_multi_binds(16) params[0].timestamp((1626861392589, 1626861392590, 1626861392591)) @@ -125,7 +125,7 @@ class TDTestCase: assert stmt.affected_rows == 3 #query 1 - querystmt=conn.statement("select ?,bu from log") + querystmt=conn.statement("select ?,bu from stb1") queryparam=new_bind_params(1) print(type(queryparam)) queryparam[0].binary("ts") @@ -135,7 +135,7 @@ class TDTestCase: # rows=result.fetch_all() # print( querystmt.use_result()) - # result = conn.query("select * from log") + # result = conn.query("select * from stb1") rows=result.fetch_all() # rows=result.fetch_all() print(rows) @@ -144,7 +144,7 @@ class TDTestCase: assert rows[2][1] == None #query 2 - querystmt1=conn.statement("select * from log where bu < ?") + querystmt1=conn.statement("select * from stb1 where bu < ?") queryparam1=new_bind_params(1) print(type(queryparam1)) queryparam1[0].int(4) From 1a8325021598527c35e50138e7db54d5115dc5a6 Mon Sep 17 00:00:00 2001 From: dapan1121 Date: Fri, 1 Jul 2022 15:28:45 +0800 Subject: [PATCH 2/7] fix: fix redirect crash issue --- include/libs/qcom/query.h | 7 ------- source/client/src/clientImpl.c | 3 +++ source/libs/scheduler/inc/schedulerInt.h | 7 ++++--- source/libs/scheduler/src/schJob.c | 2 +- source/libs/scheduler/src/schRemote.c | 16 ++++++++-------- 5 files changed, 16 insertions(+), 19 deletions(-) diff --git a/include/libs/qcom/query.h b/include/libs/qcom/query.h index 9881c8cb44..7f7fe76139 100644 --- a/include/libs/qcom/query.h +++ b/include/libs/qcom/query.h @@ -249,13 +249,6 @@ extern int32_t (*queryProcessMsgRsp[TDMT_MAX])(void* output, char* msg, int32_t ((_code) == TSDB_CODE_RPC_REDIRECT || (_code) == TSDB_CODE_NODE_NOT_DEPLOYED || \ (_code) == TSDB_CODE_SYN_NOT_LEADER || (_code) == TSDB_CODE_APP_NOT_READY) -#define NEED_SCHEDULER_RETRY_ERROR(_code) \ - (NEED_SCHEDULER_REDIRECT_ERROR(_code) || (_code) == TSDB_CODE_RPC_NETWORK_UNAVAIL || \ - (_code) == TSDB_CODE_SCH_TIMEOUT_ERROR || (_code) == TSDB_CODE_RPC_BROKEN_LINK) - - - - #define REQUEST_TOTAL_EXEC_TIMES 2 #define qFatal(...) \ diff --git a/source/client/src/clientImpl.c b/source/client/src/clientImpl.c index b6009ebbda..7d42bbdad9 100644 --- a/source/client/src/clientImpl.c +++ b/source/client/src/clientImpl.c @@ -786,6 +786,7 @@ int32_t handleQueryExecRsp(SRequestObj* pRequest) { void schedulerExecCb(SQueryResult* pResult, void* param, int32_t code) { SRequestObj* pRequest = (SRequestObj*)param; pRequest->code = code; + pRequest->body.resInfo.execRes = pResult->res; if (TDMT_VND_SUBMIT == pRequest->type || TDMT_VND_DELETE == pRequest->type || TDMT_VND_CREATE_TABLE == pRequest->type) { @@ -797,6 +798,8 @@ void schedulerExecCb(SQueryResult* pResult, void* param, int32_t code) { } } + taosMemoryFree(pResult); + tscDebug("0x%" PRIx64 " enter scheduler exec cb, code:%d - %s, reqId:0x%" PRIx64, pRequest->self, code, tstrerror(code), pRequest->requestId); diff --git a/source/libs/scheduler/inc/schedulerInt.h b/source/libs/scheduler/inc/schedulerInt.h index 94a1108710..aaa8274ce8 100644 --- a/source/libs/scheduler/inc/schedulerInt.h +++ b/source/libs/scheduler/inc/schedulerInt.h @@ -276,8 +276,6 @@ extern SSchedulerMgmt schMgmt; #define SCH_TASK_ID(_task) ((_task) ? (_task)->taskId : -1) #define SCH_TASK_EID(_task) ((_task) ? (_task)->execId : -1) -#define SCH_SET_TASK_LASTMSG_TYPE(_task, _type) do { if(_task) { atomic_store_32(&(_task)->lastMsgType, _type); } } while (0) -#define SCH_GET_TASK_LASTMSG_TYPE(_task) ((_task) ? atomic_load_32(&(_task)->lastMsgType) : -1) #define SCH_IS_DATA_SRC_QRY_TASK(task) ((task)->plan->subplanType == SUBPLAN_TYPE_SCAN) #define SCH_IS_DATA_SRC_TASK(task) (((task)->plan->subplanType == SUBPLAN_TYPE_SCAN) || ((task)->plan->subplanType == SUBPLAN_TYPE_MODIFY)) @@ -309,7 +307,10 @@ extern SSchedulerMgmt schMgmt; #define SCH_IS_NEED_DROP_JOB(_job) (SCH_IS_QUERY_JOB(_job)) #define SCH_IS_EXPLAIN_JOB(_job) (EXPLAIN_MODE_ANALYZE == (_job)->attr.explainMode) #define SCH_NETWORK_ERR(_code) ((_code) == TSDB_CODE_RPC_BROKEN_LINK || (_code) == TSDB_CODE_RPC_NETWORK_UNAVAIL) -#define SCH_SUB_TASK_NETWORK_ERR(_code, _len) (((_code) == TSDB_CODE_RPC_NETWORK_UNAVAIL || (_code) == TSDB_CODE_RPC_BROKEN_LINK) && ((_len) > 0)) +#define SCH_SUB_TASK_NETWORK_ERR(_code, _len) (SCH_NETWORK_ERR(_code) && ((_len) > 0)) +#define SCH_NEED_REDIRECT_MSGTYPE(_msgType) ((_msgType) == TDMT_SCH_QUERY || (_msgType) == TDMT_SCH_MERGE_QUERY || (_msgType) == TDMT_SCH_FETCH) +#define SCH_NEED_REDIRECT(_msgType, _code, _rspLen) (SCH_NEED_REDIRECT_MSGTYPE(_msgType) && (NEED_SCHEDULER_REDIRECT_ERROR(_code) || SCH_SUB_TASK_NETWORK_ERR(_code, _rspLen))) +#define SCH_NEED_RETRY(_msgType, _code) ((SCH_NETWORK_ERR(_code) && SCH_NEED_REDIRECT_MSGTYPE(_msgType)) || (_code) == TSDB_CODE_SCH_TIMEOUT_ERROR) #define SCH_IS_LEVEL_UNFINISHED(_level) ((_level)->taskLaunchedNum < (_level)->taskNum) #define SCH_GET_CUR_EP(_addr) (&(_addr)->epSet.eps[(_addr)->epSet.inUse]) diff --git a/source/libs/scheduler/src/schJob.c b/source/libs/scheduler/src/schJob.c index 425ea242fd..6e4838e50c 100644 --- a/source/libs/scheduler/src/schJob.c +++ b/source/libs/scheduler/src/schJob.c @@ -835,7 +835,7 @@ int32_t schTaskCheckSetRetry(SSchJob *pJob, SSchTask *pTask, int32_t errCode, bo return TSDB_CODE_SUCCESS; } - if (!NEED_SCHEDULER_RETRY_ERROR(errCode)) { + if (!SCH_NEED_RETRY(pTask->lastMsgType, errCode)) { *needRetry = false; SCH_TASK_DLOG("task no more retry cause of errCode, errCode:%x - %s", errCode, tstrerror(errCode)); return TSDB_CODE_SUCCESS; diff --git a/source/libs/scheduler/src/schRemote.c b/source/libs/scheduler/src/schRemote.c index 69e41d3111..f8a627600e 100644 --- a/source/libs/scheduler/src/schRemote.c +++ b/source/libs/scheduler/src/schRemote.c @@ -23,7 +23,7 @@ int32_t schValidateReceivedMsgType(SSchJob *pJob, SSchTask *pTask, int32_t msgType) { - int32_t lastMsgType = SCH_GET_TASK_LASTMSG_TYPE(pTask); + int32_t lastMsgType = pTask->lastMsgType; int32_t taskStatus = SCH_GET_TASK_STATUS(pTask); int32_t reqMsgType = msgType - 1; switch (msgType) { @@ -42,7 +42,7 @@ int32_t schValidateReceivedMsgType(SSchJob *pJob, SSchTask *pTask, int32_t msgTy TMSG_INFO(msgType)); } - SCH_SET_TASK_LASTMSG_TYPE(pTask, -1); + //SCH_SET_TASK_LASTMSG_TYPE(pTask, -1); return TSDB_CODE_SUCCESS; case TDMT_SCH_FETCH_RSP: if (lastMsgType != reqMsgType && -1 != lastMsgType) { @@ -57,7 +57,7 @@ int32_t schValidateReceivedMsgType(SSchJob *pJob, SSchTask *pTask, int32_t msgTy SCH_ERR_RET(TSDB_CODE_SCH_STATUS_ERROR); } - SCH_SET_TASK_LASTMSG_TYPE(pTask, -1); + //SCH_SET_TASK_LASTMSG_TYPE(pTask, -1); return TSDB_CODE_SUCCESS; case TDMT_VND_CREATE_TABLE_RSP: case TDMT_VND_DROP_TABLE_RSP: @@ -82,7 +82,7 @@ int32_t schValidateReceivedMsgType(SSchJob *pJob, SSchTask *pTask, int32_t msgTy SCH_ERR_RET(TSDB_CODE_SCH_STATUS_ERROR); } - SCH_SET_TASK_LASTMSG_TYPE(pTask, -1); + //SCH_SET_TASK_LASTMSG_TYPE(pTask, -1); return TSDB_CODE_SUCCESS; } @@ -396,7 +396,8 @@ int32_t schHandleCallback(void *param, SDataBuf *pMsg, int32_t rspCode) { SCH_ERR_JRET(schValidateReceivedMsgType(pJob, pTask, msgType)); - if (NEED_SCHEDULER_REDIRECT_ERROR(rspCode) || SCH_SUB_TASK_NETWORK_ERR(rspCode, pMsg->len > 0)) { + int32_t reqType = IsReq(pMsg) ? pMsg->msgType : (pMsg->msgType - 1); + if (SCH_NEED_REDIRECT(reqType, rspCode, pMsg->len)) { code = schHandleRedirect(pJob, pTask, (SDataBuf *)pMsg, rspCode); goto _return; } @@ -855,6 +856,7 @@ int32_t schAsyncSendMsg(SSchJob *pJob, SSchTask *pTask, SSchTrans *trans, SQuery addr->nodeId, epSet->eps[epSet->inUse].fqdn, epSet->eps[epSet->inUse].port, trans->pTrans, trans->pHandle); + pTask->lastMsgType = msgType; int64_t transporterId = 0; code = asyncSendMsgToServerExt(trans->pTrans, epSet, &transporterId, pMsgSendInfo, persistHandle, ctx); @@ -1098,8 +1100,6 @@ int32_t schBuildAndSendMsg(SSchJob *pJob, SSchTask *pTask, SQueryNodeAddr *addr, break; } - SCH_SET_TASK_LASTMSG_TYPE(pTask, msgType); - SSchTrans trans = {.pTrans = pJob->conn.pTrans, .pHandle = SCH_GET_TASK_HANDLE(pTask)}; SCH_ERR_JRET(schAsyncSendMsg(pJob, pTask, &trans, addr, msgType, msg, msgSize, persistHandle, (rpcCtx.args ? &rpcCtx : NULL))); @@ -1112,7 +1112,7 @@ int32_t schBuildAndSendMsg(SSchJob *pJob, SSchTask *pTask, SQueryNodeAddr *addr, _return: - SCH_SET_TASK_LASTMSG_TYPE(pTask, -1); + pTask->lastMsgType = -1; schFreeRpcCtx(&rpcCtx); taosMemoryFreeClear(msg); From 1aa45f73f4371c8784cdffcd63f17d4d7baf8f2c Mon Sep 17 00:00:00 2001 From: Cary Xu Date: Fri, 1 Jul 2022 20:54:57 +0800 Subject: [PATCH 3/7] feat: rsma recovery --- include/util/taoserror.h | 1 + source/dnode/vnode/CMakeLists.txt | 2 + source/dnode/vnode/src/inc/sma.h | 24 +-- source/dnode/vnode/src/inc/vnodeInt.h | 3 + source/dnode/vnode/src/sma/smaCommit.c | 150 +++++++++++++++++++ source/dnode/vnode/src/sma/smaRollup.c | 182 ++++++++++++++--------- source/dnode/vnode/src/sma/smaSnapshot.c | 16 ++ source/dnode/vnode/src/sma/smaUtil.c | 81 +++++----- source/dnode/vnode/src/vnd/vnodeCfg.c | 8 +- source/dnode/vnode/src/vnd/vnodeCommit.c | 6 + source/dnode/vnode/src/vnd/vnodeSvr.c | 1 + source/util/src/terror.c | 2 + 12 files changed, 352 insertions(+), 124 deletions(-) create mode 100644 source/dnode/vnode/src/sma/smaCommit.c create mode 100644 source/dnode/vnode/src/sma/smaSnapshot.c diff --git a/include/util/taoserror.h b/include/util/taoserror.h index db6f1c0e1e..116ad45757 100644 --- a/include/util/taoserror.h +++ b/include/util/taoserror.h @@ -617,6 +617,7 @@ int32_t* taosGetErrno(); #define TSDB_CODE_RSMA_INVALID_ENV TAOS_DEF_ERROR_CODE(0, 0x3150) #define TSDB_CODE_RSMA_INVALID_STAT TAOS_DEF_ERROR_CODE(0, 0x3151) #define TSDB_CODE_RSMA_QTASKINFO_CREATE TAOS_DEF_ERROR_CODE(0, 0x3152) +#define TSDB_CODE_RSMA_FILE_CORRUPTED TAOS_DEF_ERROR_CODE(0, 0x3153) //index #define TSDB_CODE_INDEX_REBUILDING TAOS_DEF_ERROR_CODE(0, 0x3200) diff --git a/source/dnode/vnode/CMakeLists.txt b/source/dnode/vnode/CMakeLists.txt index b09ab34224..174fe6dab5 100644 --- a/source/dnode/vnode/CMakeLists.txt +++ b/source/dnode/vnode/CMakeLists.txt @@ -31,6 +31,8 @@ target_sources( "src/sma/smaEnv.c" "src/sma/smaUtil.c" "src/sma/smaOpen.c" + "src/sma/smaCommit.c" + "src/sma/smaSnapshot.c" "src/sma/smaRollup.c" "src/sma/smaTimeRange.c" diff --git a/source/dnode/vnode/src/inc/sma.h b/source/dnode/vnode/src/inc/sma.h index b455d779c1..a95eb395c2 100644 --- a/source/dnode/vnode/src/inc/sma.h +++ b/source/dnode/vnode/src/inc/sma.h @@ -92,8 +92,9 @@ enum { TASK_TRIGGER_STAT_INIT = 0, TASK_TRIGGER_STAT_ACTIVE = 1, TASK_TRIGGER_STAT_INACTIVE = 2, - TASK_TRIGGER_STAT_CANCELLED = 3, - TASK_TRIGGER_STAT_FINISHED = 4, + TASK_TRIGGER_STAT_PAUSED = 3, + TASK_TRIGGER_STAT_CANCELLED = 4, + TASK_TRIGGER_STAT_FINISHED = 5, }; void tdDestroySmaEnv(SSmaEnv *pSmaEnv); void *tdFreeSmaEnv(SSmaEnv *pSmaEnv); @@ -214,25 +215,22 @@ struct STFInfo { }; struct STFile { - STFInfo info; - STfsFile f; - TdFilePtr pFile; uint8_t state; + STFInfo info; + char *fname; + TdFilePtr pFile; }; -#define TD_TFILE_F(tf) (&((tf)->f)) #define TD_TFILE_PFILE(tf) ((tf)->pFile) #define TD_TFILE_OPENED(tf) (TD_TFILE_PFILE(tf) != NULL) -#define TD_TFILE_FULL_NAME(tf) (TD_TFILE_F(tf)->aname) -#define TD_TFILE_REL_NAME(tf) (TD_TFILE_F(tf)->rname) +#define TD_TFILE_FULL_NAME(tf) ((tf)->fname) #define TD_TFILE_OPENED(tf) (TD_TFILE_PFILE(tf) != NULL) #define TD_TFILE_CLOSED(tf) (!TD_TFILE_OPENED(tf)) #define TD_TFILE_SET_CLOSED(f) (TD_TFILE_PFILE(f) = NULL) #define TD_TFILE_SET_STATE(tf, s) ((tf)->state = (s)) -#define TD_TFILE_DID(tf) (TD_TFILE_F(tf)->did) -int32_t tdInitTFile(STFile *pTFile, STfs *pTfs, const char *fname); -int32_t tdCreateTFile(STFile *pTFile, STfs *pTfs, bool updateHeader, int8_t fType); +int32_t tdInitTFile(STFile *pTFile, const char *dname, const char *fname); +int32_t tdCreateTFile(STFile *pTFile, bool updateHeader, int8_t fType); int32_t tdOpenTFile(STFile *pTFile, int flags); int64_t tdReadTFile(STFile *pTFile, void *buf, int64_t nbyte); int64_t tdSeekTFile(STFile *pTFile, int64_t offset, int whence); @@ -244,8 +242,10 @@ int32_t tdLoadTFileHeader(STFile *pTFile, STFInfo *pInfo); int32_t tdUpdateTFileHeader(STFile *pTFile); void tdUpdateTFileMagic(STFile *pTFile, void *pCksm); void tdCloseTFile(STFile *pTFile); +void tdDestroyTFile(STFile *pTFile); -void tdGetVndFileName(int32_t vgId, const char *dname, const char *fname, char *outputName); +void tdGetVndFileName(int32_t vgId, const char *dname, const char *fname, int64_t version, char *outputName); +void tdGetVndDirName(int32_t vgId, const char *dname, char *outputName); #ifdef __cplusplus } diff --git a/source/dnode/vnode/src/inc/vnodeInt.h b/source/dnode/vnode/src/inc/vnodeInt.h index 6bd7d4edd1..c52884d6ad 100644 --- a/source/dnode/vnode/src/inc/vnodeInt.h +++ b/source/dnode/vnode/src/inc/vnodeInt.h @@ -165,6 +165,9 @@ SSubmitReq* tdBlockToSubmit(const SArray* pBlocks, const STSchema* pSchema, bool int32_t smaOpen(SVnode* pVnode); int32_t smaCloseEnv(SSma* pSma); int32_t smaCloseEx(SSma* pSma); +int32_t smaPreCommit(SSma* pSma); +int32_t smaCommit(SSma* pSma); +int32_t smaPostCommit(SSma* pSma); int32_t tdProcessTSmaCreate(SSma* pSma, int64_t version, const char* msg); int32_t tdProcessTSmaInsert(SSma* pSma, int64_t indexUid, const char* msg); diff --git a/source/dnode/vnode/src/sma/smaCommit.c b/source/dnode/vnode/src/sma/smaCommit.c new file mode 100644 index 0000000000..b98f9bc866 --- /dev/null +++ b/source/dnode/vnode/src/sma/smaCommit.c @@ -0,0 +1,150 @@ +/* + * Copyright (c) 2019 TAOS Data, Inc. + * + * This program is free software: you can use, redistribute, and/or modify + * it under the terms of the GNU Affero General Public License, version 3 + * or later ("AGPL"), as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, but WITHOUT + * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or + * FITNESS FOR A PARTICULAR PURPOSE. + * + * You should have received a copy of the GNU Affero General Public License + * along with this program. If not, see . + */ + +#include "sma.h" + +static int32_t tdProcessRSmaPreCommitImpl(SSma *pSma); +static int32_t tdProcessRSmaCommitImpl(SSma *pSma); +static int32_t tdProcessRSmaPostCommitImpl(SSma *pSma); + +/** + * @brief Only applicable to Rollup SMA + * + * @param pSma + * @return int32_t + */ +int32_t smaPreCommit(SSma *pSma) { return tdProcessRSmaPreCommitImpl(pSma); } + +/** + * @brief Only applicable to Rollup SMA + * + * @param pSma + * @return int32_t + */ +int32_t smaCommit(SSma *pSma) { return tdProcessRSmaCommitImpl(pSma); } + +/** + * @brief Only applicable to Rollup SMA + * + * @param pSma + * @return int32_t + */ +int32_t smaPostCommit(SSma *pSma) { return tdProcessRSmaPostCommitImpl(pSma); } + +/** + * @brief pre-commit for rollup sma. + * 1) set trigger stat of rsma timer TASK_TRIGGER_STAT_PAUSED. + * 2) perform persist task for qTaskInfo + * 3) wait all triggered fetch tasks finished + * 4) set trigger stat of rsma timer TASK_TRIGGER_STAT_ACTIVE. + * 5) finish + * + * @param pSma + * @return int32_t + */ +static int32_t tdProcessRSmaPreCommitImpl(SSma *pSma) { + SSmaEnv *pSmaEnv = SMA_RSMA_ENV(pSma); + if (!pSmaEnv) { + return TSDB_CODE_SUCCESS; + } + + SSmaStat *pStat = SMA_ENV_STAT(pSmaEnv); + SRSmaStat *pRSmaStat = SMA_RSMA_STAT(pStat); + + // step 1 + atomic_store_8(RSMA_TRIGGER_STAT(pRSmaStat), TASK_TRIGGER_STAT_PAUSED); + + // step 2 + + return TSDB_CODE_SUCCESS; +} + +/** + * @brief commit for rollup sma + * + * @param pSma + * @return int32_t + */ +static int32_t tdProcessRSmaCommitImpl(SSma *pSma) { + SSmaEnv *pSmaEnv = SMA_RSMA_ENV(pSma); + if (!pSmaEnv) { + return TSDB_CODE_SUCCESS; + } + return TSDB_CODE_SUCCESS; +} + +/** + * @brief post-commit for rollup sma + * 1) clean up the outdated qtaskinfo files + * + * @param pSma + * @return int32_t + */ +static int32_t tdProcessRSmaPostCommitImpl(SSma *pSma) { + SVnode *pVnode = pSma->pVnode; + + if (!VND_IS_RSMA(pVnode)) { + return TSDB_CODE_SUCCESS; + } + + int64_t committed = pVnode->state.committed; + TdDirPtr pDir = NULL; + TdDirEntryPtr pDirEntry = NULL; + char dir[TSDB_FILENAME_LEN]; + char bname[TSDB_FILENAME_LEN]; + const char *pattern = "^v[0-9]+qtaskinfo\\.ver([0-9]+)?$"; + regex_t regex; + + tdGetVndDirName(TD_VID(pVnode), VNODE_RSMA_DIR, dir); + + // Resource allocation and init + regcomp(®ex, pattern, REG_EXTENDED); + + if ((pDir = taosOpenDir(dir)) == NULL) { + terrno = TAOS_SYSTEM_ERROR(errno); + smaWarn("rsma post-commit open dir %s failed since %s", dir, terrstr()); + return TSDB_CODE_FAILED; + } + + regmatch_t regMatch[2]; + while ((pDirEntry = taosReadDir(pDir)) != NULL) { + char *entryName = taosGetDirEntryName(pDirEntry); + if (!entryName) { + continue; + } + char *fileName = taosDirEntryBaseName(entryName); + int code = regexec(®ex, bname, 2, regMatch, 0); + + if (code == 0) { + // match + printf("match 0 = %s\n", (char *)POINTER_SHIFT(fileName, regMatch[0].rm_so)); + printf("match 1 = %s\n", (char *)POINTER_SHIFT(fileName, regMatch[1].rm_so)); + } else if (code == REG_NOMATCH) { + // not match + smaInfo("rsma post-commit not match %s", fileName); + continue; + } else { + // has other error + terrno = TAOS_SYSTEM_ERROR(code); + smaWarn("rsma post-commit regexec failed since %s", terrstr()); + + taosCloseDir(&pDir); + regfree(®ex); + return TSDB_CODE_FAILED; + } + } + taosCloseDir(&pDir); + return TSDB_CODE_SUCCESS; +} diff --git a/source/dnode/vnode/src/sma/smaRollup.c b/source/dnode/vnode/src/sma/smaRollup.c index 41af641e9e..ff9b17bf6e 100644 --- a/source/dnode/vnode/src/sma/smaRollup.c +++ b/source/dnode/vnode/src/sma/smaRollup.c @@ -23,8 +23,7 @@ SSmaMgmt smaMgmt = { .smaRef = -1, }; -typedef enum { TD_QTASK_TMP_F = 0, TD_QTASK_CUR_F } TD_QTASK_FILE_T; -static const char *tdQTaskInfoFname[] = {"qtaskinfo.t", "qtaskinfo"}; +#define TD_QTASKINFO_FNAME_PREFIX "qtaskinfo.ver" typedef struct SRSmaQTaskInfoItem SRSmaQTaskInfoItem; typedef struct SRSmaQTaskInfoIter SRSmaQTaskInfoIter; @@ -37,7 +36,7 @@ static int32_t tdExecuteRSmaImpl(SSma *pSma, const void *pMsg, int32_t inputType static void tdRSmaFetchTrigger(void *param, void *tmrId); static void tdRSmaPersistTrigger(void *param, void *tmrId); static void *tdRSmaPersistExec(void *param); -static void tdRSmaQTaskInfoGetFName(int32_t vid, int8_t ftype, char *outputName); +static void tdRSmaQTaskInfoGetFName(int32_t vid, int64_t version, char *outputName); static int32_t tdRSmaQTaskInfoIterInit(SRSmaQTaskInfoIter *pIter, STFile *pTFile); static int32_t tdRSmaQTaskInfoIterNextBlock(SRSmaQTaskInfoIter *pIter, bool *isFinish); @@ -88,8 +87,8 @@ struct SRSmaQTaskInfoIter { int32_t nBufPos; }; -static void tdRSmaQTaskInfoGetFName(int32_t vgId, int8_t ftype, char *outputName) { - tdGetVndFileName(vgId, VNODE_RSMA_DIR, tdQTaskInfoFname[ftype], outputName); +static void tdRSmaQTaskInfoGetFName(int32_t vgId, int64_t version, char *outputName) { + tdGetVndFileName(vgId, VNODE_RSMA_DIR, TD_QTASKINFO_FNAME_PREFIX, version, outputName); } static FORCE_INLINE int32_t tdRSmaQTaskInfoContLen(int32_t lenWithHead) { @@ -493,7 +492,6 @@ static int32_t tdProcessSubmitReq(STsdb *pTsdb, int64_t version, void *pReq) { } static int32_t tdFetchSubmitReqSuids(SSubmitReq *pMsg, STbUidStore *pStore) { - ASSERT(pMsg != NULL); SSubmitMsgIter msgIter = {0}; SSubmitBlk *pBlock = NULL; SSubmitBlkIter blkIter = {0}; @@ -501,19 +499,26 @@ static int32_t tdFetchSubmitReqSuids(SSubmitReq *pMsg, STbUidStore *pStore) { terrno = TSDB_CODE_SUCCESS; - if (tInitSubmitMsgIter(pMsg, &msgIter) < 0) return -1; + if (tInitSubmitMsgIter(pMsg, &msgIter) < 0) { + return -1; + } while (true) { - if (tGetSubmitMsgNext(&msgIter, &pBlock) < 0) return -1; + if (tGetSubmitMsgNext(&msgIter, &pBlock) < 0) { + return -1; + } if (!pBlock) break; tdUidStorePut(pStore, msgIter.suid, NULL); } - if (terrno != TSDB_CODE_SUCCESS) return -1; + if (terrno != TSDB_CODE_SUCCESS) { + return -1; + } return 0; } static void tdDestroySDataBlockArray(SArray *pArray) { + // TODO #if 0 for (int32_t i = 0; i < taosArrayGetSize(pArray); ++i) { SSDataBlock *pDataBlock = taosArrayGet(pArray, i); @@ -598,33 +603,54 @@ static void tdRSmaFetchTrigger(void *param, void *tmrId) { pSma = RSMA_INFO_SMA(pItem->pRsmaInfo); - // if rsma trigger stat in cancelled or finished, not start fetch task anymore + // if rsma trigger stat in paused, cancelled or finished, not start fetch task int8_t rsmaTriggerStat = atomic_load_8(RSMA_TRIGGER_STAT(pStat)); - if (rsmaTriggerStat == TASK_TRIGGER_STAT_CANCELLED || rsmaTriggerStat == TASK_TRIGGER_STAT_FINISHED) { - taosReleaseRef(smaMgmt.smaRef, pItem->refId); - smaDebug("vgId:%d, not fetch rsma level %" PRIi8 " data for table:%" PRIi64 " since stat is cancelled", - SMA_VID(pSma), pItem->level, pItem->pRsmaInfo->suid); - return; + switch (rsmaTriggerStat) { + case TASK_TRIGGER_STAT_PAUSED: + case TASK_TRIGGER_STAT_CANCELLED: + case TASK_TRIGGER_STAT_FINISHED: { + taosReleaseRef(smaMgmt.smaRef, pItem->refId); + smaDebug("vgId:%d, not fetch rsma level %" PRIi8 " data for table:%" PRIi64 " since stat is cancelled", + SMA_VID(pSma), pItem->level, pItem->pRsmaInfo->suid); + return; + } + default: + break; } int8_t fetchTriggerStat = atomic_val_compare_exchange_8(&pItem->triggerStat, TASK_TRIGGER_STAT_ACTIVE, TASK_TRIGGER_STAT_INACTIVE); - if (fetchTriggerStat == TASK_TRIGGER_STAT_ACTIVE) { - smaDebug("vgId:%d, fetch rsma level %" PRIi8 " data for table:%" PRIi64 " since stat is active", SMA_VID(pSma), - pItem->level, pItem->pRsmaInfo->suid); + switch (fetchTriggerStat) { + case TASK_TRIGGER_STAT_ACTIVE: { + smaDebug("vgId:%d, fetch rsma level %" PRIi8 " data for table:%" PRIi64 " since stat is active", SMA_VID(pSma), + pItem->level, pItem->pRsmaInfo->suid); - tdRefSmaStat(pSma, (SSmaStat *)pStat); + tdRefSmaStat(pSma, (SSmaStat *)pStat); - SSDataBlock dataBlock = {.info.type = STREAM_GET_ALL}; - qSetStreamInput(pItem->taskInfo, &dataBlock, STREAM_INPUT__DATA_BLOCK, false); - tdFetchAndSubmitRSmaResult(pItem, STREAM_INPUT__DATA_BLOCK); + SSDataBlock dataBlock = {.info.type = STREAM_GET_ALL}; + qSetStreamInput(pItem->taskInfo, &dataBlock, STREAM_INPUT__DATA_BLOCK, false); + tdFetchAndSubmitRSmaResult(pItem, STREAM_INPUT__DATA_BLOCK); - tdUnRefSmaStat(pSma, (SSmaStat *)pStat); - - } else { - smaDebug("vgId:%d, not fetch rsma level %" PRIi8 " data for table:%" PRIi64 " since stat is inactive", - SMA_VID(pSma), pItem->level, pItem->pRsmaInfo->suid); + tdUnRefSmaStat(pSma, (SSmaStat *)pStat); + } break; + case TASK_TRIGGER_STAT_PAUSED: { + smaDebug("vgId:%d, not fetch rsma level %" PRIi8 " data for table:%" PRIi64 " since stat is paused", + SMA_VID(pSma), pItem->level, pItem->pRsmaInfo->suid); + } break; + case TASK_TRIGGER_STAT_INACTIVE: { + smaDebug("vgId:%d, not fetch rsma level %" PRIi8 " data for table:%" PRIi64 " since stat is inactive", + SMA_VID(pSma), pItem->level, pItem->pRsmaInfo->suid); + } break; + case TASK_TRIGGER_STAT_INIT: { + smaDebug("vgId:%d, not fetch rsma level %" PRIi8 " data for table:%" PRIi64 " since stat is init", SMA_VID(pSma), + pItem->level, pItem->pRsmaInfo->suid); + } break; + default: { + smaWarn("vgId:%d, not fetch rsma level %" PRIi8 " data for table:%" PRIi64 " since stat is unknown", + SMA_VID(pSma), pItem->level, pItem->pRsmaInfo->suid); + } break; } + _end: taosReleaseRef(smaMgmt.smaRef, pItem->refId); } @@ -780,10 +806,10 @@ _err: static int32_t tdRSmaRestoreQTaskInfoReload(SSma *pSma) { SVnode *pVnode = pSma->pVnode; STFile tFile = {0}; - char qTaskInfoFName[TSDB_FILENAME_LEN]; + char qTaskInfoFName[TSDB_FILENAME_LEN] = {0}; - tdRSmaQTaskInfoGetFName(TD_VID(pVnode), TD_QTASK_TMP_F, qTaskInfoFName); - if (tdInitTFile(&tFile, pVnode->pTfs, qTaskInfoFName) < 0) { + tdRSmaQTaskInfoGetFName(TD_VID(pVnode), pVnode->state.committed, qTaskInfoFName); + if (tdInitTFile(&tFile, tfsGetPrimaryPath(pVnode->pTfs), qTaskInfoFName) < 0) { goto _err; } @@ -799,17 +825,20 @@ static int32_t tdRSmaRestoreQTaskInfoReload(SSma *pSma) { if (tdRSmaQTaskInfoIterInit(&fIter, &tFile) < 0) { tdRSmaQTaskInfoIterDestroy(&fIter); tdCloseTFile(&tFile); + tdDestroyTFile(&tFile); goto _err; } if (tdRSmaQTaskInfoRestore(pSma, &fIter) < 0) { tdRSmaQTaskInfoIterDestroy(&fIter); tdCloseTFile(&tFile); + tdDestroyTFile(&tFile); goto _err; } tdRSmaQTaskInfoIterDestroy(&fIter); tdCloseTFile(&tFile); + tdDestroyTFile(&tFile); return TSDB_CODE_SUCCESS; _err: smaError("rsma restore, qtaskinfo reload failed since %s", terrstr()); @@ -931,19 +960,21 @@ static int32_t tdRSmaQTaskInfoIterNextBlock(SRSmaQTaskInfoIter *pIter, bool *isF } if (tdSeekTFile(pTFile, pIter->offset, SEEK_SET) < 0) { - ASSERT(0); return TSDB_CODE_FAILED; } if (tdReadTFile(pTFile, pIter->qBuf, nBytes) != nBytes) { - ASSERT(0); return TSDB_CODE_FAILED; } int32_t infoLen = 0; taosDecodeFixedI32(pIter->qBuf, &infoLen); if (infoLen > nBytes) { - ASSERT(infoLen > RSMA_QTASKINFO_BUFSIZE); + if (infoLen <= RSMA_QTASKINFO_BUFSIZE) { + terrno = TSDB_CODE_RSMA_FILE_CORRUPTED; + smaError("iterate rsma qtaskinfo file %s failed since %s", TD_TFILE_FULL_NAME(pIter->pTFile), terrstr()); + return TSDB_CODE_FAILED; + } pIter->nAlloc = infoLen; void *pBuf = taosMemoryRealloc(pIter->pBuf, infoLen); if (!pBuf) { @@ -955,12 +986,10 @@ static int32_t tdRSmaQTaskInfoIterNextBlock(SRSmaQTaskInfoIter *pIter, bool *isF nBytes = infoLen; if (tdSeekTFile(pTFile, pIter->offset, SEEK_SET)) { - ASSERT(0); return TSDB_CODE_FAILED; } if (tdReadTFile(pTFile, pIter->pBuf, nBytes) != nBytes) { - ASSERT(0); return TSDB_CODE_FAILED; } } @@ -977,7 +1006,6 @@ static int32_t tdRSmaQTaskInfoRestore(SSma *pSma, SRSmaQTaskInfoIter *pIter) { // block iter bool isFinish = false; if (tdRSmaQTaskInfoIterNextBlock(pIter, &isFinish) < 0) { - ASSERT(0); return TSDB_CODE_FAILED; } if (isFinish) { @@ -989,6 +1017,7 @@ static int32_t tdRSmaQTaskInfoRestore(SSma *pSma, SRSmaQTaskInfoIter *pIter) { pIter->qBuf = taosDecodeFixedI32(pIter->qBuf, &qTaskInfoLenWithHead); if (qTaskInfoLenWithHead < RSMA_QTASKINFO_HEAD_LEN) { terrno = TSDB_CODE_TDB_FILE_CORRUPTED; + smaError("restore rsma qtaskinfo file %s failed since %s", TD_TFILE_FULL_NAME(pIter->pTFile), terrstr()); return TSDB_CODE_FAILED; } @@ -1025,22 +1054,16 @@ static int32_t tdRSmaQTaskInfoRestore(SSma *pSma, SRSmaQTaskInfoIter *pIter) { return TSDB_CODE_SUCCESS; } -static void *tdRSmaPersistExec(void *param) { - setThreadName("rsma-task-persist"); - SRSmaStat *pRSmaStat = param; - SSma *pSma = pRSmaStat->pSma; - STfs *pTfs = pSma->pVnode->pTfs; - int32_t vid = SMA_VID(pSma); - int64_t toffset = 0; - bool isFileCreated = false; - - if (TASK_TRIGGER_STAT_CANCELLED == atomic_load_8(RSMA_TRIGGER_STAT(pRSmaStat))) { - goto _end; - } +static int32_t tdRSmaPersistExecImpl(SRSmaStat *pRSmaStat) { + SSma *pSma = pRSmaStat->pSma; + SVnode *pVnode = pSma->pVnode; + int32_t vid = SMA_VID(pSma); + int64_t toffset = 0; + bool isFileCreated = false; void *infoHash = taosHashIterate(RSMA_INFO_HASH(pRSmaStat), NULL); if (!infoHash) { - goto _end; + return TSDB_CODE_SUCCESS; } STFile tFile = {0}; @@ -1074,9 +1097,13 @@ static void *tdRSmaPersistExec(void *param) { if (!isFileCreated) { char qTaskInfoFName[TSDB_FILENAME_LEN]; - tdRSmaQTaskInfoGetFName(vid, TD_QTASK_TMP_F, qTaskInfoFName); - tdInitTFile(&tFile, pTfs, qTaskInfoFName); - tdCreateTFile(&tFile, pTfs, true, -1); + tdRSmaQTaskInfoGetFName(vid, pSma->pVnode->state.applied, qTaskInfoFName); + if (tdInitTFile(&tFile, tfsGetPrimaryPath(pVnode->pTfs), qTaskInfoFName) < 0) { + goto _err; + } + if (tdCreateTFile(&tFile, true, -1) < 0) { + goto _err; + } isFileCreated = true; } @@ -1101,49 +1128,55 @@ static void *tdRSmaPersistExec(void *param) { infoHash = taosHashIterate(RSMA_INFO_HASH(pRSmaStat), infoHash); } -_normal: + if (isFileCreated) { if (tdUpdateTFileHeader(&tFile) < 0) { smaError("vgId:%d, rsma, failed to update tfile %s header since %s", vid, TD_TFILE_FULL_NAME(&tFile), tstrerror(terrno)); - tdCloseTFile(&tFile); - tdRemoveTFile(&tFile); goto _err; } else { smaDebug("vgId:%d, rsma, succeed to update tfile %s header", vid, TD_TFILE_FULL_NAME(&tFile)); } tdCloseTFile(&tFile); - - char newFName[TSDB_FILENAME_LEN]; - strncpy(newFName, TD_TFILE_FULL_NAME(&tFile), TSDB_FILENAME_LEN); - char *pos = strstr(newFName, tdQTaskInfoFname[TD_QTASK_TMP_F]); - strncpy(pos, tdQTaskInfoFname[TD_QTASK_TMP_F], TSDB_FILENAME_LEN - POINTER_DISTANCE(pos, newFName)); - if (taosRenameFile(TD_TFILE_FULL_NAME(&tFile), newFName) != 0) { - smaError("vgId:%d, rsma, failed to rename %s to %s", vid, TD_TFILE_FULL_NAME(&tFile), newFName); - goto _err; - } else { - smaDebug("vgId:%d, rsma, succeed to rename %s to %s", vid, TD_TFILE_FULL_NAME(&tFile), newFName); - } + tdDestroyTFile(&tFile); } - goto _end; + return TSDB_CODE_SUCCESS; _err: if (isFileCreated) { tdRemoveTFile(&tFile); + tdDestroyTFile(&tFile); } + return TSDB_CODE_FAILED; +} + +static void *tdRSmaPersistExec(void *param) { + setThreadName("rsma-task-persist"); + SRSmaStat *pRSmaStat = param; + SSma *pSma = pRSmaStat->pSma; + + int8_t triggerStat = atomic_load_8(RSMA_TRIGGER_STAT(pRSmaStat)); + + if (TASK_TRIGGER_STAT_CANCELLED == triggerStat || TASK_TRIGGER_STAT_PAUSED == triggerStat) { + goto _end; + } + + // execution + tdRSmaPersistExecImpl(pRSmaStat); + _end: if (TASK_TRIGGER_STAT_INACTIVE == atomic_val_compare_exchange_8(RSMA_TRIGGER_STAT(pRSmaStat), TASK_TRIGGER_STAT_INACTIVE, TASK_TRIGGER_STAT_ACTIVE)) { - smaDebug("vgId:%d, rsma persist task is active again", vid); + smaDebug("vgId:%d, rsma persist task is active again", SMA_VID(pSma)); } else if (TASK_TRIGGER_STAT_CANCELLED == atomic_val_compare_exchange_8(RSMA_TRIGGER_STAT(pRSmaStat), TASK_TRIGGER_STAT_CANCELLED, TASK_TRIGGER_STAT_FINISHED)) { - smaDebug("vgId:%d, rsma persist task is cancelled", vid); + smaDebug("vgId:%d, rsma persist task is cancelled", SMA_VID(pSma)); } else { - smaWarn("vgId:%d, rsma persist task in abnormal stat %" PRIi8, vid, atomic_load_8(RSMA_TRIGGER_STAT(pRSmaStat))); - ASSERT(0); + smaWarn("vgId:%d, rsma persist task in stat %" PRIi8, SMA_VID(pSma), atomic_load_8(RSMA_TRIGGER_STAT(pRSmaStat))); } + atomic_store_8(RSMA_RUNNING_STAT(pRSmaStat), 0); taosReleaseRef(smaMgmt.smaRef, pRSmaStat->refId); taosThreadExit(NULL); @@ -1166,8 +1199,8 @@ static void tdRSmaPersistTask(SRSmaStat *pRSmaStat) { TASK_TRIGGER_STAT_FINISHED)) { smaDebug("vgId:%d, persist task is cancelled and set finished", SMA_VID(pRSmaStat->pSma)); } else { - smaWarn("vgId:%d, persist task in abnormal stat %" PRIi8, atomic_load_8(RSMA_TRIGGER_STAT(pRSmaStat)), - SMA_VID(pRSmaStat->pSma)); + smaWarn("vgId:%d, persist task in abnormal stat %" PRIi8, SMA_VID(pRSmaStat->pSma), + atomic_load_8(RSMA_TRIGGER_STAT(pRSmaStat))); ASSERT(0); } atomic_store_8(RSMA_RUNNING_STAT(pRSmaStat), 0); @@ -1216,6 +1249,9 @@ static void tdRSmaPersistTrigger(void *param, void *tmrId) { atomic_store_8(RSMA_TRIGGER_STAT(pRSmaStat), TASK_TRIGGER_STAT_FINISHED); smaDebug("rsma persistence not start since cancelled and finished"); } break; + case TASK_TRIGGER_STAT_PAUSED: { + smaDebug("rsma persistence not start since paused"); + } break; case TASK_TRIGGER_STAT_INACTIVE: { smaDebug("rsma persistence not start since inactive"); } break; diff --git a/source/dnode/vnode/src/sma/smaSnapshot.c b/source/dnode/vnode/src/sma/smaSnapshot.c new file mode 100644 index 0000000000..b2c85642b9 --- /dev/null +++ b/source/dnode/vnode/src/sma/smaSnapshot.c @@ -0,0 +1,16 @@ +/* + * Copyright (c) 2019 TAOS Data, Inc. + * + * This program is free software: you can use, redistribute, and/or modify + * it under the terms of the GNU Affero General Public License, version 3 + * or later ("AGPL"), as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, but WITHOUT + * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or + * FITNESS FOR A PARTICULAR PURPOSE. + * + * You should have received a copy of the GNU Affero General Public License + * along with this program. If not, see . + */ + +#include "sma.h" \ No newline at end of file diff --git a/source/dnode/vnode/src/sma/smaUtil.c b/source/dnode/vnode/src/sma/smaUtil.c index 17bc2cdaca..9bea3ff468 100644 --- a/source/dnode/vnode/src/sma/smaUtil.c +++ b/source/dnode/vnode/src/sma/smaUtil.c @@ -179,72 +179,83 @@ void tdCloseTFile(STFile *pTFile) { } } -void tdGetVndFileName(int32_t vgId, const char *dname, const char *fname, char *outputName) { - snprintf(outputName, TSDB_FILENAME_LEN, "vnode/vnode%d/%s/%s", vgId, dname, fname); +void tdDestroyTFile(STFile *pTFile) { taosMemoryFreeClear(TD_TFILE_FULL_NAME(pTFile)); } + +void tdGetVndFileName(int32_t vgId, const char *dname, const char *fname, int64_t version, char *outputName) { + if (version < 0) { + snprintf(outputName, TSDB_FILENAME_LEN, "vnode/vnode%d/%s/v%d%s", vgId, dname, vgId, fname); + } else { + snprintf(outputName, TSDB_FILENAME_LEN, "vnode/vnode%d/%s/v%d%s%" PRIi64, vgId, dname, vgId, fname, version); + } } -int32_t tdInitTFile(STFile *pTFile, STfs *pTfs, const char *fname) { - char fullname[TSDB_FILENAME_LEN]; - SDiskID did = {0}; +void tdGetVndDirName(int32_t vgId, const char *dname, char *outputName) { + snprintf(outputName, TSDB_FILENAME_LEN, "vnode/vnode%d/%s", vgId, dname); +} +int32_t tdInitTFile(STFile *pTFile, const char *dname, const char *fname) { TD_TFILE_SET_STATE(pTFile, TD_FILE_STATE_OK); TD_TFILE_SET_CLOSED(pTFile); memset(&(pTFile->info), 0, sizeof(pTFile->info)); pTFile->info.magic = TD_FILE_INIT_MAGIC; - if (tfsAllocDisk(pTfs, 0, &did) < 0) { - terrno = TSDB_CODE_NO_AVAIL_DISK; + char tmpName[TSDB_FILENAME_LEN * 2 + 32] = {0}; + snprintf(tmpName, TSDB_FILENAME_LEN * 2 + 32, "%s%s%s", dname, TD_DIRSEP, fname); + int32_t tmpNameLen = strlen(tmpName) + 1; + pTFile->fname = taosMemoryMalloc(tmpNameLen); + if (!pTFile->fname) { + terrno = TSDB_CODE_OUT_OF_MEMORY; return -1; } - - tfsInitFile(pTfs, &(pTFile->f), did, fname); + tstrncpy(pTFile->fname, tmpName, tmpNameLen); return 0; } -int32_t tdCreateTFile(STFile *pTFile, STfs *pTfs, bool updateHeader, int8_t fType) { +int32_t tdCreateTFile(STFile *pTFile, bool updateHeader, int8_t fType) { ASSERT(pTFile->info.fsize == 0 && pTFile->info.magic == TD_FILE_INIT_MAGIC); pTFile->pFile = taosOpenFile(TD_TFILE_FULL_NAME(pTFile), TD_FILE_CREATE | TD_FILE_WRITE | TD_FILE_TRUNC); if (pTFile->pFile == NULL) { if (errno == ENOENT) { // Try to create directory recursively - char *s = strdup(TD_TFILE_REL_NAME(pTFile)); - if (tfsMkdirRecurAt(pTfs, taosDirName(s), TD_TFILE_DID(pTFile)) < 0) { - taosMemoryFreeClear(s); - return -1; - } - taosMemoryFreeClear(s); - - pTFile->pFile = taosOpenFile(TD_TFILE_FULL_NAME(pTFile), TD_FILE_CREATE | TD_FILE_WRITE | TD_FILE_TRUNC); - if (pTFile->pFile == NULL) { + if (taosMulMkDir(taosDirName(TD_TFILE_FULL_NAME(pTFile))) != 0) { terrno = TAOS_SYSTEM_ERROR(errno); return -1; + } else { + pTFile->pFile = taosOpenFile(TD_TFILE_FULL_NAME(pTFile), TD_FILE_CREATE | TD_FILE_WRITE | TD_FILE_TRUNC); + if (pTFile->pFile == NULL) { + terrno = TAOS_SYSTEM_ERROR(errno); + return -1; + } } - } else { - terrno = TAOS_SYSTEM_ERROR(errno); + } + + if (!updateHeader) { + return 0; + } + + pTFile->info.fsize += TD_FILE_HEAD_SIZE; + pTFile->info.fver = 0; + + if (tdUpdateTFileHeader(pTFile) < 0) { + tdCloseTFile(pTFile); + tdRemoveTFile(pTFile); return -1; } } - if (!updateHeader) { - return 0; - } - - pTFile->info.fsize += TD_FILE_HEAD_SIZE; - pTFile->info.fver = 0; - - if (tdUpdateTFileHeader(pTFile) < 0) { - tdCloseTFile(pTFile); - tdRemoveTFile(pTFile); - return -1; - } - return 0; } -int32_t tdRemoveTFile(STFile *pTFile) { return tfsRemoveFile(TD_TFILE_F(pTFile)); } +int32_t tdRemoveTFile(STFile *pTFile) { + if (taosRemoveFile(TD_TFILE_FULL_NAME(pTFile)) != 0) { + terrno = TAOS_SYSTEM_ERROR(errno); + return -1; + }; + return 0; +} // smaXXXUtil ================ // ... \ No newline at end of file diff --git a/source/dnode/vnode/src/vnd/vnodeCfg.c b/source/dnode/vnode/src/vnd/vnodeCfg.c index 79e63d5abc..0b28d6bf10 100644 --- a/source/dnode/vnode/src/vnd/vnodeCfg.c +++ b/source/dnode/vnode/src/vnd/vnodeCfg.c @@ -28,12 +28,12 @@ const SVnodeCfg vnodeCfgDefault = { .update = 1, .compression = 2, .slLevel = 5, - .days = 10, + .days = 14400, .minRows = 100, .maxRows = 4096, - .keep2 = 3650, - .keep0 = 3650, - .keep1 = 3650}, + .keep2 = 5256000, + .keep0 = 5256000, + .keep1 = 5256000}, .walCfg = {.vgId = -1, .fsyncPeriod = 0, .retentionPeriod = 0, .rollPeriod = 0, .segSize = 0, .level = TAOS_WAL_WRITE}, .hashBegin = 0, diff --git a/source/dnode/vnode/src/vnd/vnodeCommit.c b/source/dnode/vnode/src/vnd/vnodeCommit.c index 02a0783927..a12e11b398 100644 --- a/source/dnode/vnode/src/vnd/vnodeCommit.c +++ b/source/dnode/vnode/src/vnd/vnodeCommit.c @@ -229,6 +229,9 @@ int vnodeCommit(SVnode *pVnode) { return -1; } + // preCommit + // TODO + // commit each sub-system if (metaCommit(pVnode->pMeta) < 0) { ASSERT(0); @@ -269,6 +272,9 @@ int vnodeCommit(SVnode *pVnode) { pVnode->state.committed = info.state.committed; + // postCommit + smaPostCommit(pVnode->pSma); + // apply the commit (TODO) vnodeBufPoolReset(pVnode->onCommit); pVnode->onCommit->next = pVnode->pPool; diff --git a/source/dnode/vnode/src/vnd/vnodeSvr.c b/source/dnode/vnode/src/vnd/vnodeSvr.c index b73eaa4288..649e8299f4 100644 --- a/source/dnode/vnode/src/vnd/vnodeSvr.c +++ b/source/dnode/vnode/src/vnd/vnodeSvr.c @@ -779,6 +779,7 @@ _exit: taosArrayDestroy(submitRsp.pArray); // TODO: the partial success scenario and the error case + // => If partial success, extract the success submitted rows and reconstruct a new submit msg, and push to level 1/level 2. // TODO: refactor if ((terrno == TSDB_CODE_SUCCESS) && (pRsp->code == TSDB_CODE_SUCCESS)) { tdProcessRSmaSubmit(pVnode->pSma, pReq, STREAM_INPUT__DATA_SUBMIT); diff --git a/source/util/src/terror.c b/source/util/src/terror.c index 3ab307dc3f..7909e322f1 100644 --- a/source/util/src/terror.c +++ b/source/util/src/terror.c @@ -586,6 +586,8 @@ TAOS_DEFINE_ERROR(TSDB_CODE_TSMA_NO_INDEX_IN_CACHE, "No tsma index in ca TAOS_DEFINE_ERROR(TSDB_CODE_RSMA_INVALID_ENV, "Invalid rsma env") TAOS_DEFINE_ERROR(TSDB_CODE_RSMA_INVALID_STAT, "Invalid rsma state") TAOS_DEFINE_ERROR(TSDB_CODE_RSMA_QTASKINFO_CREATE, "Rsma qtaskinfo creation error") +TAOS_DEFINE_ERROR(TSDB_CODE_RSMA_FILE_CORRUPTED, "Rsma file corrupted") + //tq TAOS_DEFINE_ERROR(TSDB_CODE_TQ_NO_COMMITTED_OFFSET, "No committed offset") From 410e15c5e7ded9b5661dbf71885728bbac30d987 Mon Sep 17 00:00:00 2001 From: tomchon Date: Fri, 1 Jul 2022 21:11:57 +0800 Subject: [PATCH 4/7] test: modify testcases of muti-mnodes --- .../1-insert/insertWithMoreVgroup.py | 34 ---- tests/system-test/2-query/queryQnode.py | 161 +++++++++++++++++- .../5dnode3mnodeSep1VnodeStopVnodeCreateDb.py | 28 +-- .../6-cluster/clusterCommonCheck.py | 3 +- .../6-cluster/clusterCommonCreate.py | 1 - tests/system-test/fulltest.sh | 2 +- 6 files changed, 158 insertions(+), 71 deletions(-) diff --git a/tests/system-test/1-insert/insertWithMoreVgroup.py b/tests/system-test/1-insert/insertWithMoreVgroup.py index e3a20cebf8..80468509ee 100644 --- a/tests/system-test/1-insert/insertWithMoreVgroup.py +++ b/tests/system-test/1-insert/insertWithMoreVgroup.py @@ -355,40 +355,6 @@ class TDTestCase: return - def test_case4(self): - self.taosBenchCreate("127.0.0.1","no","db1", "stb1", 1, 2, 1*10) - tdSql.execute("use db1;") - tdSql.query("show dnodes;") - dnodeId=tdSql.getData(0,0) - print(dnodeId) - tdSql.execute("create qnode on dnode %s"%dnodeId) - tdSql.query("select max(c1) from stb10;") - maxQnode=tdSql.getData(0,0) - tdSql.query("select min(c1) from stb11;") - minQnode=tdSql.getData(0,0) - tdSql.query("select c0,c1 from stb11_1 where (c0>1000) union select c0,c1 from stb11_1 where c0>2000;") - unionQnode=tdSql.queryResult - tdSql.query("select c0,c1 from stb11_1 where (c0>1000) union all select c0,c1 from stb11_1 where c0>2000;") - unionallQnode=tdSql.queryResult - - # tdSql.query("show qnodes;") - # qnodeId=tdSql.getData(0,0) - tdSql.execute("drop qnode on dnode %s"%dnodeId) - tdSql.execute("reset query cache") - tdSql.query("select max(c1) from stb10;") - tdSql.checkData(0, 0, "%s"%maxQnode) - tdSql.query("select min(c1) from stb11;") - tdSql.checkData(0, 0, "%s"%minQnode) - tdSql.query("select c0,c1 from stb11_1 where (c0>1000) union select c0,c1 from stb11_1 where c0>2000;") - unionVnode=tdSql.queryResult - assert unionQnode == unionVnode - tdSql.query("select c0,c1 from stb11_1 where (c0>1000) union all select c0,c1 from stb11_1 where c0>2000;") - unionallVnode=tdSql.queryResult - assert unionallQnode == unionallVnode - - - # tdSql.execute("create qnode on dnode %s"%dnodeId) - # run case def run(self): diff --git a/tests/system-test/2-query/queryQnode.py b/tests/system-test/2-query/queryQnode.py index 15b643f5d7..0011fe248f 100644 --- a/tests/system-test/2-query/queryQnode.py +++ b/tests/system-test/2-query/queryQnode.py @@ -17,6 +17,8 @@ import threading as thd import multiprocessing as mp from numpy.lib.function_base import insert import taos +from util.dnodes import TDDnode +from util.dnodes import * from util.log import * from util.cases import * from util.sql import * @@ -30,9 +32,9 @@ class TDTestCase: # # --------------- main frame ------------------- # - clientCfgDict = {'queryproxy': '1','debugFlag': 135} - clientCfgDict["queryproxy"] = '2' - clientCfgDict["debugFlag"] = 143 + clientCfgDict = {'queryPolicy': '1','debugFlag': 135} + clientCfgDict["queryPolicy"] = '1' + clientCfgDict["debugFlag"] = 131 updatecfgDict = {'clientCfg': {}} updatecfgDict = {'debugFlag': 143} @@ -62,7 +64,7 @@ class TDTestCase: return buildPath # init - def init(self, conn, logSql): + def init(self, conn, logSql=True): tdLog.debug("start to execute %s" % __file__) tdSql.init(conn.cursor()) # tdSql.prepare() @@ -292,12 +294,13 @@ class TDTestCase: tdLog.debug("-----create database and muti-thread create tables test------- ") - def test_case4(self): + def test_case1(self): self.taosBenchCreate("127.0.0.1","no","db1", "stb1", 1, 2, 1*10) tdSql.execute("use db1;") tdSql.query("show dnodes;") dnodeId=tdSql.getData(0,0) print(dnodeId) + tdLog.debug("create qnode on dnode %s"%dnodeId) tdSql.execute("create qnode on dnode %s"%dnodeId) tdSql.query("select max(c1) from stb10;") maxQnode=tdSql.getData(0,0) @@ -310,6 +313,7 @@ class TDTestCase: # tdSql.query("show qnodes;") # qnodeId=tdSql.getData(0,0) + tdLog.debug("drop qnode on dnode %s"%dnodeId) tdSql.execute("drop qnode on dnode %s"%dnodeId) tdSql.execute("reset query cache") tdSql.query("select max(c1) from stb10;") @@ -323,15 +327,156 @@ class TDTestCase: unionallVnode=tdSql.queryResult assert unionallQnode == unionallVnode + queryPolicy=2 + simClientCfg="%s/taos.cfg"%tdDnodes.getSimCfgPath() + cmd='sed -i "s/^queryPolicy.*/queryPolicy 2/g" %s'%simClientCfg + os.system(cmd) + # tdDnodes.stop(1) + # tdDnodes.start(1) + tdSql.execute("reset query cache") + tdSql.execute('alter local "queryPolicy" "%d"'%queryPolicy) + tdSql.query("show local variables;") + for i in range(tdSql.queryRows): + if tdSql.queryResult[i][0] == "queryPolicy" : + if int(tdSql.queryResult[i][1]) == int(queryPolicy): + tdLog.success('alter queryPolicy to %d successfully'%queryPolicy) + else : + tdLog.debug(tdSql.queryResult) + tdLog.exit("alter queryPolicy to %d failed"%queryPolicy) + tdSql.execute("reset query cache") + + tdSql.execute("use db1;") + tdSql.query("show dnodes;") + dnodeId=tdSql.getData(0,0) + tdLog.debug("create qnode on dnode %s"%dnodeId) + + tdSql.execute("create qnode on dnode %s"%dnodeId) + tdSql.query("select max(c1) from stb10;") + assert maxQnode==tdSql.getData(0,0) + tdSql.query("select min(c1) from stb11;") + assert minQnode==tdSql.getData(0,0) + tdSql.query("select c0,c1 from stb11_1 where (c0>1000) union select c0,c1 from stb11_1 where c0>2000;") + assert unionQnode==tdSql.queryResult + tdSql.query("select c0,c1 from stb11_1 where (c0>1000) union all select c0,c1 from stb11_1 where c0>2000;") + assert unionallQnode==tdSql.queryResult + + # tdSql.query("show qnodes;") + # qnodeId=tdSql.getData(0,0) + tdLog.debug("drop qnode on dnode %s"%dnodeId) + tdSql.execute("drop qnode on dnode %s"%dnodeId) + tdSql.execute("reset query cache") + tdSql.query("select max(c1) from stb10;") + assert maxQnode==tdSql.getData(0,0) + tdSql.query("select min(c1) from stb11;") + assert minQnode==tdSql.getData(0,0) + tdSql.error("select c0,c1 from stb11_1 where (c0>1000) union select c0,c1 from stb11_1 where c0>2000;") + tdSql.error("select c0,c1 from stb11_1 where (c0>1000) union all select c0,c1 from stb11_1 where c0>2000;") # tdSql.execute("create qnode on dnode %s"%dnodeId) + queryPolicy=3 + simClientCfg="%s/taos.cfg"%tdDnodes.getSimCfgPath() + cmd='sed -i "s/^queryPolicy.*/queryPolicy 2/g" %s'%simClientCfg + os.system(cmd) + # tdDnodes.stop(1) + # tdDnodes.start(1) + tdSql.execute("reset query cache") + tdSql.execute('alter local "queryPolicy" "%d"'%queryPolicy) + tdSql.query("show local variables;") + for i in range(tdSql.queryRows): + if tdSql.queryResult[i][0] == "queryPolicy" : + if int(tdSql.queryResult[i][1]) == int(queryPolicy): + tdLog.success('alter queryPolicy to %d successfully'%queryPolicy) + else : + tdLog.debug(tdSql.queryResult) + tdLog.exit("alter queryPolicy to %d failed"%queryPolicy) + tdSql.execute("reset query cache") + + tdSql.execute("use db1;") + tdSql.query("show dnodes;") + dnodeId=tdSql.getData(0,0) + tdLog.debug("create qnode on dnode %s"%dnodeId) + + tdSql.execute("create qnode on dnode %s"%dnodeId) + tdSql.query("select max(c1) from stb10;") + assert maxQnode==tdSql.getData(0,0) + tdSql.query("select min(c1) from stb11;") + assert minQnode==tdSql.getData(0,0) + tdSql.query("select c0,c1 from stb11_1 where (c0>1000) union select c0,c1 from stb11_1 where c0>2000;") + assert unionQnode==tdSql.queryResult + tdSql.query("select c0,c1 from stb11_1 where (c0>1000) union all select c0,c1 from stb11_1 where c0>2000;") + assert unionallQnode==tdSql.queryResult + + def test_case2(self): + self.taosBenchCreate("127.0.0.1","no","db1", "stb1", 10, 2, 1*10) + tdSql.query("show qnodes") + if tdSql.queryRows == 1 : + tdLog.debug("drop qnode on dnode 1") + tdSql.execute("drop qnode on dnode 1") + queryPolicy=2 + simClientCfg="%s/taos.cfg"%tdDnodes.getSimCfgPath() + cmd='sed -i "s/^queryPolicy.*/queryPolicy 2/g" %s'%simClientCfg + os.system(cmd) + # tdDnodes.stop(1) + # tdDnodes.start(1) + tdSql.execute("reset query cache") + tdSql.execute('alter local "queryPolicy" "%d"'%queryPolicy) + tdSql.query("show local variables;") + for i in range(tdSql.queryRows): + if tdSql.queryResult[i][0] == "queryPolicy" : + if int(tdSql.queryResult[i][1]) == int(queryPolicy): + tdLog.success('alter queryPolicy to %d successfully'%queryPolicy) + else : + tdLog.debug(tdSql.queryResult) + tdLog.exit("alter queryPolicy to %d failed"%queryPolicy) + tdSql.execute("use db1;") + tdSql.error("select max(c1) from stb10;") + tdSql.error("select min(c1) from stb11;") + tdSql.error("select c0,c1 from stb11_1 where (c0>1000) union select c0,c1 from stb11_1 where c0>2000;") + tdSql.error("select c0,c1 from stb11_1 where (c0>1000) union all select c0,c1 from stb11_1 where c0>2000;") + + tdSql.query("select max(c1) from stb10_0;") + tdSql.query("select min(c1) from stb11_0;") + + def test_case3(self): + + tdSql.execute('alter local "queryPolicy" "3"') + tdLog.debug("create qnode on dnode 1") + tdSql.execute("create qnode on dnode 1") + tdSql.execute("use db1;") + tdSql.query("show dnodes;") + dnodeId=tdSql.getData(0,0) + print(dnodeId) + + tdSql.query("select max(c1) from stb10;") + maxQnode=tdSql.getData(0,0) + tdSql.query("select min(c1) from stb11;") + minQnode=tdSql.getData(0,0) + tdSql.query("select c0,c1 from stb11_1 where (c0>1000) union select c0,c1 from stb11_1 where c0>2000;") + unionQnode=tdSql.queryResult + tdSql.query("select c0,c1 from stb11_1 where (c0>1000) union all select c0,c1 from stb11_1 where c0>2000;") + unionallQnode=tdSql.queryResult + + # tdSql.query("show qnodes;") + # qnodeId=tdSql.getData(0,0) + tdLog.debug("drop qnode on dnode %s"%dnodeId) + + tdSql.execute("drop qnode on dnode %s"%dnodeId) + tdSql.execute("reset query cache") + + tdSql.error("select max(c1) from stb10;") + tdSql.error("select min(c1) from stb11;") + tdSql.error("select c0,c1 from stb11_1 where (c0>1000) union select c0,c1 from stb11_1 where c0>2000;") + tdSql.error("select c0,c1 from stb11_1 where (c0>1000) union all select c0,c1 from stb11_1 where c0>2000;") + # run case def run(self): - # test qnode - self.test_case4() - tdLog.debug(" LIMIT test_case3 ............ [OK]") + self.test_case1() + self.test_case2() + + self.test_case3() + # tdLog.debug(" LIMIT test_case3 ............ [OK]") return diff --git a/tests/system-test/6-cluster/5dnode3mnodeSep1VnodeStopVnodeCreateDb.py b/tests/system-test/6-cluster/5dnode3mnodeSep1VnodeStopVnodeCreateDb.py index e7353b0f13..80508979c5 100644 --- a/tests/system-test/6-cluster/5dnode3mnodeSep1VnodeStopVnodeCreateDb.py +++ b/tests/system-test/6-cluster/5dnode3mnodeSep1VnodeStopVnodeCreateDb.py @@ -65,31 +65,6 @@ class TDTestCase: self._async_raise(thread.ident, SystemExit) - def insertData(self,countstart,countstop): - # fisrt add data : db\stable\childtable\general table - - for couti in range(countstart,countstop): - tdLog.debug("drop database if exists db%d" %couti) - tdSql.execute("drop database if exists db%d" %couti) - print("create database if not exists db%d replica 1 duration 300" %couti) - tdSql.execute("create database if not exists db%d replica 1 duration 300" %couti) - tdSql.execute("use db%d" %couti) - tdSql.execute( - '''create table stb1 - (ts timestamp, c1 int, c2 bigint, c3 smallint, c4 tinyint, c5 float, c6 double, c7 bool, c8 binary(16),c9 nchar(32), c10 timestamp) - tags (t1 int) - ''' - ) - tdSql.execute( - ''' - create table t1 - (ts timestamp, c1 int, c2 bigint, c3 smallint, c4 tinyint, c5 float, c6 double, c7 bool, c8 binary(16),c9 nchar(32), c10 timestamp) - ''' - ) - for i in range(4): - tdSql.execute(f'create table ct{i+1} using stb1 tags ( {i+1} )') - - def fiveDnodeThreeMnode(self,dnodeNumbers,mnodeNums,restartNumbers,stopRole): tdLog.printNoPrefix("======== test case 1: ") paraDict = {'dbName': 'db', @@ -143,7 +118,8 @@ class TDTestCase: threads=[] for i in range(restartNumbers): dbNameIndex = '%s%d'%(paraDict["dbName"],i) - threads.append(threading.Thread(target=clusterComCreate.create_databases, args=(tdSql, dbNameIndex,paraDict["dbNumbers"],paraDict["dropFlag"], paraDict["vgroups"],paraDict['replica']))) + newTdSql=tdCom.newTdSql() + threads.append(threading.Thread(target=clusterComCreate.create_databases, args=(newTdSql, dbNameIndex,paraDict["dbNumbers"],paraDict["dropFlag"], paraDict["vgroups"],paraDict['replica']))) for tr in threads: tr.start() diff --git a/tests/system-test/6-cluster/clusterCommonCheck.py b/tests/system-test/6-cluster/clusterCommonCheck.py index 2d378e623d..cde09b4c58 100644 --- a/tests/system-test/6-cluster/clusterCommonCheck.py +++ b/tests/system-test/6-cluster/clusterCommonCheck.py @@ -39,6 +39,7 @@ class ClusterComCheck: def checkDnodes(self,dnodeNumbers): count=0 + # print(tdSql) while count < 5: tdSql.query("show dnodes") # tdLog.debug(tdSql.queryResult) @@ -85,7 +86,7 @@ class ClusterComCheck: tdLog.debug("check %s_%d that status is ready "%(dbNameIndex,j)) else: continue - print(query_status) + # print(query_status) count+=1 if query_status == dbNumbers: tdLog.success("we find cluster with %d dnode and check all databases are ready within 5s! " %dbNumbers) diff --git a/tests/system-test/6-cluster/clusterCommonCreate.py b/tests/system-test/6-cluster/clusterCommonCreate.py index 78aac309a0..9d3f74a2e3 100644 --- a/tests/system-test/6-cluster/clusterCommonCreate.py +++ b/tests/system-test/6-cluster/clusterCommonCreate.py @@ -125,7 +125,6 @@ class ClusterComCreate: def create_databases(self,tsql,dbNameIndex,dbNumbers,dropFlag=1,vgroups=4,replica=1): for i in range(dbNumbers): - print(dbNumbers) if dropFlag == 1: tsql.execute("drop database if exists %s_%d"%(dbNameIndex,i)) tsql.execute("create database if not exists %s_%d vgroups %d replica %d"%(dbNameIndex,i, vgroups, replica)) diff --git a/tests/system-test/fulltest.sh b/tests/system-test/fulltest.sh index 44a3afcf73..22c749743b 100755 --- a/tests/system-test/fulltest.sh +++ b/tests/system-test/fulltest.sh @@ -121,7 +121,7 @@ python3 ./test.py -f 6-cluster/5dnode3mnodeStop.py -N 5 -M 3 python3 ./test.py -f 6-cluster/5dnode3mnodeStopLoop.py -N 5 -M 3 # BUG python3 ./test.py -f 6-cluster/5dnode3mnodeSep1VnodeStopDnodeCreateDb.py -N 5 -M 3 # BUG python3 ./test.py -f 6-cluster/5dnode3mnodeSep1VnodeStopMnodeCreateDb.py -N 5 -M 3 -# python3 ./test.py -f 6-cluster/5dnode3mnodeSep1VnodeStopVnodeCreateDb.py -N 5 -M 3 +python3 ./test.py -f 6-cluster/5dnode3mnodeSep1VnodeStopVnodeCreateDb.py -N 5 -M 3 # BUG python3 ./test.py -f 6-cluster/5dnode3mnodeSep1VnodeStopDnodeCreateStb.py -N 5 -M 3 # BUG python3 ./test.py -f 6-cluster/5dnode3mnodeSep1VnodeStopMnodeCreateStb.py -N 5 -M 3 # python3 ./test.py -f 6-cluster/5dnode3mnodeSep1VnodeStopVnodeCreateStb.py -N 5 -M 3 From b9cc3e6ec62fe9562e2ba1b3f1841f8c6979df03 Mon Sep 17 00:00:00 2001 From: plum-lihui Date: Fri, 1 Jul 2022 22:15:20 +0800 Subject: [PATCH 5/7] test:add test case for tmq --- tests/pytest/util/common.py | 6 ++ tests/system-test/7-tmq/stbFilter.py | 6 +- tests/system-test/7-tmq/tmqConsFromTsdb.py | 101 ++++++++++++++++++++- 3 files changed, 108 insertions(+), 5 deletions(-) diff --git a/tests/pytest/util/common.py b/tests/pytest/util/common.py index 6425e7acfa..94043ed01a 100644 --- a/tests/pytest/util/common.py +++ b/tests/pytest/util/common.py @@ -662,6 +662,12 @@ class TDCom: return res_list else: tdLog.exit(f"getOneRow out of range: row_index={location} row_count={self.query_row}") + + def killProcessor(self, processorName): + if (platform.system().lower() == 'windows'): + os.system("TASKKILL /F /IM %s.exe"%processorName) + else: + os.system('pkill %s'%processorName) def is_json(msg): diff --git a/tests/system-test/7-tmq/stbFilter.py b/tests/system-test/7-tmq/stbFilter.py index 542894574b..7383d5b692 100644 --- a/tests/system-test/7-tmq/stbFilter.py +++ b/tests/system-test/7-tmq/stbFilter.py @@ -122,9 +122,9 @@ class TDTestCase: tdLog.info("wait the consume result") expectRows = 1 resultList = tmqCom.selectConsumeResult(expectRows) - # if expectRowsList[2] != resultList[0]: - # tdLog.info("expect consume rows: %d, act consume rows: %d"%(expectRowsList[2], resultList[0])) - # tdLog.exit("2 tmq consume rows error!") + if expectRowsList[2] != resultList[0]: + tdLog.info("expect consume rows: %d, act consume rows: %d"%(expectRowsList[2], resultList[0])) + tdLog.exit("2 tmq consume rows error!") time.sleep(10) for i in range(len(topicNameList)): diff --git a/tests/system-test/7-tmq/tmqConsFromTsdb.py b/tests/system-test/7-tmq/tmqConsFromTsdb.py index ae0e658344..64f09f4a17 100644 --- a/tests/system-test/7-tmq/tmqConsFromTsdb.py +++ b/tests/system-test/7-tmq/tmqConsFromTsdb.py @@ -16,7 +16,7 @@ sys.path.append("./7-tmq") from tmqCommon import * class TDTestCase: - def __int__(self): + def __init__(self): self.vgroups = 1 self.ctbNum = 10 self.rowsPerTbl = 10000 @@ -216,7 +216,7 @@ class TDTestCase: 'rowsPerTbl': 10000, 'batchNum': 10, 'startTs': 1640966400000, # 2022-01-01 00:00:00.000 - 'pollDelay': -1, + 'pollDelay': 10, 'showMsg': 1, 'showRow': 1, 'snapshot': 1} @@ -281,10 +281,107 @@ class TDTestCase: tdLog.printNoPrefix("======== test case 3 end ...... ") + def tmqCase4(self): + tdLog.printNoPrefix("======== test case 4: ") + paraDict = {'dbName': 'dbt', + 'dropFlag': 1, + 'event': '', + 'vgroups': 1, + 'stbName': 'stb', + 'colPrefix': 'c', + 'tagPrefix': 't', + 'colSchema': [{'type': 'INT', 'count':1}, {'type': 'binary', 'len':20, 'count':1}], + 'tagSchema': [{'type': 'INT', 'count':1}, {'type': 'binary', 'len':20, 'count':1}], + 'ctbPrefix': 'ctb', + 'ctbNum': 1, + 'rowsPerTbl': 10000, + 'batchNum': 10, + 'startTs': 1640966400000, # 2022-01-01 00:00:00.000 + 'pollDelay': 10, + 'showMsg': 1, + 'showRow': 1, + 'snapshot': 1} + + paraDict['vgroups'] = self.vgroups + paraDict['ctbNum'] = self.ctbNum + paraDict['rowsPerTbl'] = self.rowsPerTbl + + topicNameList = ['topic1'] + expectRowsList = [] + tmqCom.initConsumerTable() + # tdCom.create_database(tdSql, paraDict["dbName"],paraDict["dropFlag"], vgroups=paraDict["vgroups"],replica=1) + # tdLog.info("create stb") + # tdCom.create_stable(tdSql, dbname=paraDict["dbName"],stbname=paraDict["stbName"], column_elm_list=paraDict['colSchema'], tag_elm_list=paraDict['tagSchema']) + # tdLog.info("create ctb") + # tdCom.create_ctable(tdSql, dbname=paraDict["dbName"],stbname=paraDict["stbName"],tag_elm_list=paraDict['tagSchema'],count=paraDict["ctbNum"], default_ctbname_prefix=paraDict['ctbPrefix']) + # tdLog.info("insert data") + # tmqCom.insert_data(tdSql,paraDict["dbName"],paraDict["ctbPrefix"],paraDict["ctbNum"],paraDict["rowsPerTbl"],paraDict["batchNum"],paraDict["startTs"]) + + # tdDnodes.stop(1) + # tdDnodes.start(1) + + tdLog.info("create topics from stb with filter") + queryString = "select * from %s.%s"%(paraDict['dbName'], paraDict['stbName']) + # sqlString = "create topic %s as stable %s" %(topicNameList[0], paraDict['stbName']) + sqlString = "create topic %s as %s" %(topicNameList[0], queryString) + tdLog.info("create topic sql: %s"%sqlString) + tdSql.execute(sqlString) + tdSql.query(queryString) + expectRowsList.append(tdSql.getRows()) + totalRowsInserted = expectRowsList[0] + + # init consume info, and start tmq_sim, then check consume result + tdLog.info("insert consume info to consume processor") + consumerId = 5 + expectrowcnt = math.ceil(paraDict["rowsPerTbl"] * paraDict["ctbNum"]) + topicList = topicNameList[0] + ifcheckdata = 1 + ifManualCommit = 1 + keyList = 'group.id:cgrp1, enable.auto.commit:true, auto.commit.interval.ms:1000, auto.offset.reset:earliest' + tmqCom.insertConsumerInfo(consumerId, expectrowcnt,topicList,keyList,ifcheckdata,ifManualCommit) + + tdLog.info("start consume processor 0") + tmqCom.startTmqSimProcess(pollDelay=paraDict['pollDelay'],dbName=paraDict["dbName"],showMsg=paraDict['showMsg'], showRow=paraDict['showRow'],snapshot=paraDict['snapshot']) + + tdLog.info("wait commit notify") + tmqCom.getStartCommitNotifyFromTmqsim() + + tdLog.info("pkill consume processor") + tdCom.killProcessor("tmq_sim") + + # time.sleep(10) + + # reinit consume info, and start tmq_sim, then check consume result + tmqCom.initConsumerTable() + consumerId = 6 + tmqCom.insertConsumerInfo(consumerId, expectrowcnt,topicList,keyList,ifcheckdata,ifManualCommit) + + tdLog.info("start consume processor 1") + tmqCom.startTmqSimProcess(pollDelay=paraDict['pollDelay'],dbName=paraDict["dbName"],showMsg=paraDict['showMsg'], showRow=paraDict['showRow'],snapshot=paraDict['snapshot']) + tdLog.info("wait the consume result") + + expectRows = 1 + resultList = tmqCom.selectConsumeResult(expectRows) + + actConsumeTotalRows = resultList[0] + + if not (actConsumeTotalRows > 0 and actConsumeTotalRows < totalRowsInserted): + tdLog.info("act consume rows: %d"%(actConsumeTotalRows)) + tdLog.info("and second consume rows should be between 0 and %d"%(totalRowsInserted)) + tdLog.exit("%d tmq consume rows error!"%consumerId) + + time.sleep(10) + for i in range(len(topicNameList)): + tdSql.query("drop topic %s"%topicNameList[i]) + + tdLog.printNoPrefix("======== test case 4 end ...... ") + def run(self): tdSql.prepare() self.tmqCase1() self.tmqCase2() + self.tmqCase3() + self.tmqCase4() def stop(self): tdSql.close() From 4a2b2666125184f38f27590741735d0669f50d34 Mon Sep 17 00:00:00 2001 From: Liu Jicong Date: Sat, 2 Jul 2022 10:55:01 +0800 Subject: [PATCH 6/7] feat(stream): support ignore expired flag --- source/dnode/mnode/impl/inc/mndDef.h | 2 +- source/dnode/mnode/impl/src/mndDef.c | 4 +- source/dnode/mnode/impl/src/mndStream.c | 3 +- source/dnode/vnode/src/tq/tq.c | 17 ++++---- source/dnode/vnode/src/tq/tqRead.c | 2 + source/libs/executor/inc/executorimpl.h | 1 + source/libs/executor/src/executor.c | 4 +- source/libs/executor/src/executorimpl.c | 37 ++++++++-------- source/libs/executor/src/scanoperator.c | 1 + tests/script/tsim/tmq/snapshot1.sim | 2 +- tests/test/c/tmqSim.c | 56 +++++++++++++------------ 11 files changed, 71 insertions(+), 58 deletions(-) diff --git a/source/dnode/mnode/impl/inc/mndDef.h b/source/dnode/mnode/impl/inc/mndDef.h index 1fb7fa85b1..21002f5f9c 100644 --- a/source/dnode/mnode/impl/inc/mndDef.h +++ b/source/dnode/mnode/impl/inc/mndDef.h @@ -556,7 +556,7 @@ typedef struct { int64_t uid; int8_t status; // config - int8_t dropPolicy; + int8_t igExpired; int8_t trigger; int64_t triggerParam; int64_t watermark; diff --git a/source/dnode/mnode/impl/src/mndDef.c b/source/dnode/mnode/impl/src/mndDef.c index dc7f08ebc2..727b20ef8a 100644 --- a/source/dnode/mnode/impl/src/mndDef.c +++ b/source/dnode/mnode/impl/src/mndDef.c @@ -28,7 +28,7 @@ int32_t tEncodeSStreamObj(SEncoder *pEncoder, const SStreamObj *pObj) { if (tEncodeI64(pEncoder, pObj->uid) < 0) return -1; if (tEncodeI8(pEncoder, pObj->status) < 0) return -1; - if (tEncodeI8(pEncoder, pObj->dropPolicy) < 0) return -1; + if (tEncodeI8(pEncoder, pObj->igExpired) < 0) return -1; if (tEncodeI8(pEncoder, pObj->trigger) < 0) return -1; if (tEncodeI64(pEncoder, pObj->triggerParam) < 0) return -1; if (tEncodeI64(pEncoder, pObj->watermark) < 0) return -1; @@ -73,7 +73,7 @@ int32_t tDecodeSStreamObj(SDecoder *pDecoder, SStreamObj *pObj) { if (tDecodeI64(pDecoder, &pObj->uid) < 0) return -1; if (tDecodeI8(pDecoder, &pObj->status) < 0) return -1; - if (tDecodeI8(pDecoder, &pObj->dropPolicy) < 0) return -1; + if (tDecodeI8(pDecoder, &pObj->igExpired) < 0) return -1; if (tDecodeI8(pDecoder, &pObj->trigger) < 0) return -1; if (tDecodeI64(pDecoder, &pObj->triggerParam) < 0) return -1; if (tDecodeI64(pDecoder, &pObj->watermark) < 0) return -1; diff --git a/source/dnode/mnode/impl/src/mndStream.c b/source/dnode/mnode/impl/src/mndStream.c index cabbac14f1..c27d007fbd 100644 --- a/source/dnode/mnode/impl/src/mndStream.c +++ b/source/dnode/mnode/impl/src/mndStream.c @@ -248,7 +248,7 @@ static int32_t mndBuildStreamObjFromCreateReq(SMnode *pMnode, SStreamObj *pObj, pObj->status = 0; // TODO - pObj->dropPolicy = 0; + pObj->igExpired = pCreate->igExpired; pObj->trigger = pCreate->triggerType; pObj->triggerParam = pCreate->maxDelay; pObj->watermark = pCreate->watermark; @@ -301,6 +301,7 @@ static int32_t mndBuildStreamObjFromCreateReq(SMnode *pMnode, SStreamObj *pObj, .streamQuery = true, .triggerType = pObj->trigger == STREAM_TRIGGER_MAX_DELAY ? STREAM_TRIGGER_WINDOW_CLOSE : pObj->trigger, .watermark = pObj->watermark, + .igExpired = pObj->igExpired, }; // using ast and param to build physical plan diff --git a/source/dnode/vnode/src/tq/tq.c b/source/dnode/vnode/src/tq/tq.c index 35486a5267..7c4a7d520d 100644 --- a/source/dnode/vnode/src/tq/tq.c +++ b/source/dnode/vnode/src/tq/tq.c @@ -183,13 +183,15 @@ int32_t tqProcessOffsetCommitReq(STQ* pTq, char* msg, int32_t msgLen) { } else { ASSERT(0); } - STqOffset* pOffset = tqOffsetRead(pTq->pOffsetStore, offset.subKey); - if (pOffset == NULL || pOffset->val.version < offset.val.version) { - if (tqOffsetWrite(pTq->pOffsetStore, &offset) < 0) { - ASSERT(0); - return -1; - } + /*STqOffset* pOffset = tqOffsetRead(pTq->pOffsetStore, offset.subKey);*/ + /*if (pOffset != NULL) {*/ + /*if (pOffset->val.type == TMQ_OFFSET__LOG && pOffset->val.version < offset.val.version) {*/ + if (tqOffsetWrite(pTq->pOffsetStore, &offset) < 0) { + ASSERT(0); + return -1; } + /*}*/ + /*}*/ return 0; } @@ -375,8 +377,7 @@ int32_t tqProcessPollReq(STQ* pTq, SRpcMsg* pMsg, int32_t workerId) { taosMemoryFree(pCkHead); } else if (fetchOffsetNew.type == TMQ_OFFSET__SNAPSHOT_DATA) { - tqInfo("retrieve using snapshot req offset: uid %ld ts %ld, actual offset: uid %ld ts %ld", dataRsp.reqOffset.uid, - dataRsp.reqOffset.ts, fetchOffsetNew.uid, fetchOffsetNew.ts); + tqInfo("retrieve using snapshot actual offset: uid %ld ts %ld", fetchOffsetNew.uid, fetchOffsetNew.ts); if (tqScanSnapshot(pTq, &pHandle->execHandle, &dataRsp, fetchOffsetNew, workerId) < 0) { ASSERT(0); } diff --git a/source/dnode/vnode/src/tq/tqRead.c b/source/dnode/vnode/src/tq/tqRead.c index 9e5c67fed4..bc992b2211 100644 --- a/source/dnode/vnode/src/tq/tqRead.c +++ b/source/dnode/vnode/src/tq/tqRead.c @@ -120,7 +120,9 @@ bool tqNextDataBlock(SStreamReader* pHandle) { return true; } void* ret = taosHashGet(pHandle->tbIdHash, &pHandle->msgIter.uid, sizeof(int64_t)); + /*tqDebug("search uid %ld", pHandle->msgIter.uid);*/ if (ret != NULL) { + /*tqDebug("find uid %ld", pHandle->msgIter.uid);*/ return true; } } diff --git a/source/libs/executor/inc/executorimpl.h b/source/libs/executor/inc/executorimpl.h index d1d2c55acf..cfe144e7d7 100644 --- a/source/libs/executor/inc/executorimpl.h +++ b/source/libs/executor/inc/executorimpl.h @@ -356,6 +356,7 @@ typedef struct SStreamBlockScanInfo { SUpdateInfo* pUpdateInfo; EStreamScanMode scanMode; + SOperatorInfo* pStreamScanOp; SOperatorInfo* pSnapshotReadOp; SArray* childIds; SessionWindowSupporter sessionSup; diff --git a/source/libs/executor/src/executor.c b/source/libs/executor/src/executor.c index 31edc46b4d..0e76607c8f 100644 --- a/source/libs/executor/src/executor.c +++ b/source/libs/executor/src/executor.c @@ -145,10 +145,12 @@ static SArray* filterQualifiedChildTables(const SStreamBlockScanInfo* pScanInfo, continue; } + // TODO handle ntb case if (mr.me.type != TSDB_CHILD_TABLE || mr.me.ctbEntry.suid != pScanInfo->tableUid) { continue; } - // TODO handle ntb case + /*pScanInfo->pStreamScanOp->pTaskInfo->tableqinfoList.*/ + // handle multiple partition taosArrayPush(qa, id); } diff --git a/source/libs/executor/src/executorimpl.c b/source/libs/executor/src/executorimpl.c index e8d3686bfc..ed88bfc02c 100644 --- a/source/libs/executor/src/executorimpl.c +++ b/source/libs/executor/src/executorimpl.c @@ -2010,8 +2010,9 @@ static int32_t doSendFetchDataRequest(SExchangeInfo* pExchangeInfo, SExecTaskInf ASSERT(pDataInfo->status == EX_SOURCE_DATA_NOT_READY); - qDebug("%s build fetch msg and send to vgId:%d, ep:%s, taskId:0x%" PRIx64 ", execId:%d, %d/%" PRIzu, GET_TASKID(pTaskInfo), - pSource->addr.nodeId, pSource->addr.epSet.eps[0].fqdn, pSource->taskId, pSource->execId, sourceIndex, totalSources); + qDebug("%s build fetch msg and send to vgId:%d, ep:%s, taskId:0x%" PRIx64 ", execId:%d, %d/%" PRIzu, + GET_TASKID(pTaskInfo), pSource->addr.nodeId, pSource->addr.epSet.eps[0].fqdn, pSource->taskId, pSource->execId, + sourceIndex, totalSources); pMsg->header.vgId = htonl(pSource->addr.nodeId); pMsg->sId = htobe64(pSource->schedId); @@ -2146,8 +2147,8 @@ static SSDataBlock* concurrentlyLoadRemoteDataImpl(SOperatorInfo* pOperator, SEx SSDataBlock* pRes = pExchangeInfo->pResult; SLoadRemoteDataInfo* pLoadInfo = &pExchangeInfo->loadInfo; if (pRsp->numOfRows == 0) { - qDebug("%s vgId:%d, taskId:0x%" PRIx64 " execId:%d index:%d completed, rowsOfSource:%" PRIu64 ", totalRows:%" PRIu64 - ", completed:%d try next %d/%" PRIzu, + qDebug("%s vgId:%d, taskId:0x%" PRIx64 " execId:%d index:%d completed, rowsOfSource:%" PRIu64 + ", totalRows:%" PRIu64 ", completed:%d try next %d/%" PRIzu, GET_TASKID(pTaskInfo), pSource->addr.nodeId, pSource->taskId, pSource->execId, i, pDataInfo->totalRows, pExchangeInfo->loadInfo.totalRows, completed + 1, i + 1, totalSources); pDataInfo->status = EX_SOURCE_DATA_EXHAUSTED; @@ -2166,18 +2167,19 @@ static SSDataBlock* concurrentlyLoadRemoteDataImpl(SOperatorInfo* pOperator, SEx } if (pRsp->completed == 1) { - qDebug("%s fetch msg rsp from vgId:%d, taskId:0x%" PRIx64 " execId:%d" + qDebug("%s fetch msg rsp from vgId:%d, taskId:0x%" PRIx64 + " execId:%d" " index:%d completed, numOfRows:%d, rowsOfSource:%" PRIu64 ", totalRows:%" PRIu64 ", totalBytes:%" PRIu64 ", completed:%d try next %d/%" PRIzu, - GET_TASKID(pTaskInfo), pSource->addr.nodeId, pSource->taskId, pSource->execId, i, pRes->info.rows, pDataInfo->totalRows, - pLoadInfo->totalRows, pLoadInfo->totalSize, completed + 1, i + 1, totalSources); + GET_TASKID(pTaskInfo), pSource->addr.nodeId, pSource->taskId, pSource->execId, i, pRes->info.rows, + pDataInfo->totalRows, pLoadInfo->totalRows, pLoadInfo->totalSize, completed + 1, i + 1, totalSources); completed += 1; pDataInfo->status = EX_SOURCE_DATA_EXHAUSTED; } else { qDebug("%s fetch msg rsp from vgId:%d, taskId:0x%" PRIx64 " execId:%d numOfRows:%d, totalRows:%" PRIu64 ", totalBytes:%" PRIu64, - GET_TASKID(pTaskInfo), pSource->addr.nodeId, pSource->taskId, pSource->execId, pRes->info.rows, pLoadInfo->totalRows, - pLoadInfo->totalSize); + GET_TASKID(pTaskInfo), pSource->addr.nodeId, pSource->taskId, pSource->execId, pRes->info.rows, + pLoadInfo->totalRows, pLoadInfo->totalSize); } taosMemoryFreeClear(pDataInfo->pRsp); @@ -2250,8 +2252,8 @@ static SSDataBlock* seqLoadRemoteData(SOperatorInfo* pOperator) { SDownstreamSourceNode* pSource = taosArrayGet(pExchangeInfo->pSources, pExchangeInfo->current); if (pDataInfo->code != TSDB_CODE_SUCCESS) { - qError("%s vgId:%d, taskID:0x%" PRIx64 " execId:%d error happens, code:%s", GET_TASKID(pTaskInfo), pSource->addr.nodeId, - pSource->taskId, pSource->execId, tstrerror(pDataInfo->code)); + qError("%s vgId:%d, taskID:0x%" PRIx64 " execId:%d error happens, code:%s", GET_TASKID(pTaskInfo), + pSource->addr.nodeId, pSource->taskId, pSource->execId, tstrerror(pDataInfo->code)); pOperator->pTaskInfo->code = pDataInfo->code; return NULL; } @@ -2259,8 +2261,8 @@ static SSDataBlock* seqLoadRemoteData(SOperatorInfo* pOperator) { SRetrieveTableRsp* pRsp = pDataInfo->pRsp; SLoadRemoteDataInfo* pLoadInfo = &pExchangeInfo->loadInfo; if (pRsp->numOfRows == 0) { - qDebug("%s vgId:%d, taskID:0x%" PRIx64 " execId:%d %d of total completed, rowsOfSource:%" PRIu64 ", totalRows:%" PRIu64 - " try next", + qDebug("%s vgId:%d, taskID:0x%" PRIx64 " execId:%d %d of total completed, rowsOfSource:%" PRIu64 + ", totalRows:%" PRIu64 " try next", GET_TASKID(pTaskInfo), pSource->addr.nodeId, pSource->taskId, pSource->execId, pExchangeInfo->current + 1, pDataInfo->totalRows, pLoadInfo->totalRows); @@ -2279,16 +2281,17 @@ static SSDataBlock* seqLoadRemoteData(SOperatorInfo* pOperator) { if (pRsp->completed == 1) { qDebug("%s fetch msg rsp from vgId:%d, taskId:0x%" PRIx64 " execId:%d numOfRows:%d, rowsOfSource:%" PRIu64 ", totalRows:%" PRIu64 ", totalBytes:%" PRIu64 " try next %d/%" PRIzu, - GET_TASKID(pTaskInfo), pSource->addr.nodeId, pSource->taskId, pSource->execId, pRes->info.rows, pDataInfo->totalRows, - pLoadInfo->totalRows, pLoadInfo->totalSize, pExchangeInfo->current + 1, totalSources); + GET_TASKID(pTaskInfo), pSource->addr.nodeId, pSource->taskId, pSource->execId, pRes->info.rows, + pDataInfo->totalRows, pLoadInfo->totalRows, pLoadInfo->totalSize, pExchangeInfo->current + 1, + totalSources); pDataInfo->status = EX_SOURCE_DATA_EXHAUSTED; pExchangeInfo->current += 1; } else { qDebug("%s fetch msg rsp from vgId:%d, taskId:0x%" PRIx64 " execId:%d numOfRows:%d, totalRows:%" PRIu64 ", totalBytes:%" PRIu64, - GET_TASKID(pTaskInfo), pSource->addr.nodeId, pSource->taskId, pSource->execId, pRes->info.rows, pLoadInfo->totalRows, - pLoadInfo->totalSize); + GET_TASKID(pTaskInfo), pSource->addr.nodeId, pSource->taskId, pSource->execId, pRes->info.rows, + pLoadInfo->totalRows, pLoadInfo->totalSize); } pOperator->resultInfo.totalRows += pRes->info.rows; diff --git a/source/libs/executor/src/scanoperator.c b/source/libs/executor/src/scanoperator.c index 89a47f4e2c..00df72609e 100644 --- a/source/libs/executor/src/scanoperator.c +++ b/source/libs/executor/src/scanoperator.c @@ -1274,6 +1274,7 @@ SOperatorInfo* createStreamScanOperatorInfo(SReadHandle* pHandle, STableScanPhys pInfo->sessionSup = (SessionWindowSupporter){.pStreamAggSup = NULL, .gap = -1}; pInfo->groupId = 0; pInfo->pPullDataRes = createPullDataBlock(); + pInfo->pStreamScanOp = pOperator; pOperator->name = "StreamBlockScanOperator"; pOperator->operatorType = QUERY_NODE_PHYSICAL_PLAN_STREAM_SCAN; diff --git a/tests/script/tsim/tmq/snapshot1.sim b/tests/script/tsim/tmq/snapshot1.sim index d534bb68da..58541b725d 100644 --- a/tests/script/tsim/tmq/snapshot1.sim +++ b/tests/script/tsim/tmq/snapshot1.sim @@ -25,7 +25,7 @@ $rowsPerCtb = 10 $tstart = 1640966400000 # 2022-01-01 00:00:00.000 #---- global parameters end ----# -$pullDelay = 5 +$pullDelay = 2 $ifcheckdata = 1 $ifmanualcommit = 1 $showMsg = 1 diff --git a/tests/test/c/tmqSim.c b/tests/test/c/tmqSim.c index 3f2a81197b..a333e6ea5e 100644 --- a/tests/test/c/tmqSim.c +++ b/tests/test/c/tmqSim.c @@ -36,6 +36,7 @@ #define MAX_CONSUMER_THREAD_CNT (16) #define MAX_VGROUP_CNT (32) +int64_t now; typedef enum { NOTIFY_CMD_START_CONSUM, NOTIFY_CMD_START_COMMIT, @@ -525,15 +526,15 @@ int32_t notifyMainScript(SThreadInfo* pInfo, int32_t cmdId) { static int32_t g_once_commit_flag = 0; static void tmq_commit_cb_print(tmq_t* tmq, int32_t code, void* param) { - pError("tmq_commit_cb_print() commit %d\n", code); + pError("tmq_commit_cb_print() commit %d\n", code); - if (0 == g_once_commit_flag) { - g_once_commit_flag = 1; - notifyMainScript((SThreadInfo*)param, (int32_t)NOTIFY_CMD_START_COMMIT); + if (0 == g_once_commit_flag) { + g_once_commit_flag = 1; + notifyMainScript((SThreadInfo*)param, (int32_t)NOTIFY_CMD_START_COMMIT); } - char tmpString[128]; - taosFprintfFile(g_fp, "%s tmq_commit_cb_print() be called\n", getCurrentTimeString(tmpString)); + char tmpString[128]; + taosFprintfFile(g_fp, "%s tmq_commit_cb_print() be called\n", getCurrentTimeString(tmpString)); } void build_consumer(SThreadInfo* pInfo) { @@ -588,12 +589,10 @@ void build_topic_list(SThreadInfo* pInfo) { int32_t saveConsumeResult(SThreadInfo* pInfo) { char sqlStr[1024] = {0}; - - int64_t now = taosGetTimestampMs(); - // schema: ts timestamp, consumerid int, consummsgcnt bigint, checkresult int sprintf(sqlStr, "insert into %s.consumeresult values (%" PRId64 ", %d, %" PRId64 ", %" PRId64 ", %d)", - g_stConfInfo.cdbName, now, pInfo->consumerId, pInfo->consumeMsgCnt, pInfo->consumeRowCnt, pInfo->checkresult); + g_stConfInfo.cdbName, atomic_fetch_add_64(&now, 1), pInfo->consumerId, pInfo->consumeMsgCnt, + pInfo->consumeRowCnt, pInfo->checkresult); char tmpString[128]; taosFprintfFile(g_fp, "%s, consume id %d result: %s\n", getCurrentTimeString(tmpString), pInfo->consumerId, sqlStr); @@ -637,9 +636,9 @@ void loop_consume(SThreadInfo* pInfo) { } } - int64_t lastTotalMsgs = 0; - uint64_t lastPrintTime = taosGetTimestampMs(); - uint64_t startTs = taosGetTimestampMs(); + int64_t lastTotalMsgs = 0; + uint64_t lastPrintTime = taosGetTimestampMs(); + uint64_t startTs = taosGetTimestampMs(); int32_t consumeDelay = g_stConfInfo.consumeDelay == -1 ? -1 : (g_stConfInfo.consumeDelay * 1000); while (running) { @@ -652,16 +651,16 @@ void loop_consume(SThreadInfo* pInfo) { taos_free_result(tmqMsg); totalMsgs++; - - int64_t currentPrintTime = taosGetTimestampMs(); - if (currentPrintTime - lastPrintTime > 10 * 1000) { - taosFprintfFile(g_fp, - "consumer id %d has currently poll total msgs: %" PRId64 ", period rate: %.3f msgs/second\n", - pInfo->consumerId, totalMsgs, (totalMsgs - lastTotalMsgs) * 1000.0/(currentPrintTime - lastPrintTime)); - lastPrintTime = currentPrintTime; - lastTotalMsgs = totalMsgs; - } - + + int64_t currentPrintTime = taosGetTimestampMs(); + if (currentPrintTime - lastPrintTime > 10 * 1000) { + taosFprintfFile( + g_fp, "consumer id %d has currently poll total msgs: %" PRId64 ", period rate: %.3f msgs/second\n", + pInfo->consumerId, totalMsgs, (totalMsgs - lastTotalMsgs) * 1000.0 / (currentPrintTime - lastPrintTime)); + lastPrintTime = currentPrintTime; + lastTotalMsgs = totalMsgs; + } + if (0 == once_flag) { once_flag = 1; notifyMainScript(pInfo, NOTIFY_CMD_START_CONSUM); @@ -678,7 +677,7 @@ void loop_consume(SThreadInfo* pInfo) { break; } } - + if (0 == running) { taosFprintfFile(g_fp, "receive stop signal and not continue consume\n"); } @@ -696,6 +695,7 @@ void* consumeThreadFunc(void* param) { pInfo->taos = taos_connect(NULL, "root", "taosdata", NULL, 0); if (pInfo->taos == NULL) { taosFprintfFile(g_fp, "taos_connect() fail, can not notify and save consume result to main scripte\n"); + ASSERT(0); return NULL; } @@ -855,6 +855,8 @@ int32_t getConsumeInfo() { } int main(int32_t argc, char* argv[]) { + now = taosGetTimestampMs(); + parseArgument(argc, argv); getConsumeInfo(); saveConfigToLogFile(); @@ -888,11 +890,11 @@ int main(int32_t argc, char* argv[]) { int64_t t = end - start; if (0 == t) t = 1; - + double tInMs = (double)t / 1000000.0; taosFprintfFile(g_fp, - "Spent %.3f seconds to poll msgs: %" PRIu64 " with %d thread(s), throughput: %.3f msgs/second\n\n", - tInMs, totalMsgs, g_stConfInfo.numOfThread, (double)(totalMsgs / tInMs)); + "Spent %.3f seconds to poll msgs: %" PRIu64 " with %d thread(s), throughput: %.3f msgs/second\n\n", + tInMs, totalMsgs, g_stConfInfo.numOfThread, (double)(totalMsgs / tInMs)); taosFprintfFile(g_fp, "==== close tmqlog ====\n"); taosCloseFile(&g_fp); From 504d25a156889b645295c33d8c2dbfce380c623c Mon Sep 17 00:00:00 2001 From: dapan1121 Date: Sat, 2 Jul 2022 11:29:28 +0800 Subject: [PATCH 7/7] fix: fix task is NULL issue --- source/libs/scheduler/src/schRemote.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) diff --git a/source/libs/scheduler/src/schRemote.c b/source/libs/scheduler/src/schRemote.c index f8a627600e..02e7b085ae 100644 --- a/source/libs/scheduler/src/schRemote.c +++ b/source/libs/scheduler/src/schRemote.c @@ -856,7 +856,9 @@ int32_t schAsyncSendMsg(SSchJob *pJob, SSchTask *pTask, SSchTrans *trans, SQuery addr->nodeId, epSet->eps[epSet->inUse].fqdn, epSet->eps[epSet->inUse].port, trans->pTrans, trans->pHandle); - pTask->lastMsgType = msgType; + if (pTask) { + pTask->lastMsgType = msgType; + } int64_t transporterId = 0; code = asyncSendMsgToServerExt(trans->pTrans, epSet, &transporterId, pMsgSendInfo, persistHandle, ctx);