diff --git a/include/common/tdatablock.h b/include/common/tdatablock.h index af46535c94..709462a744 100644 --- a/include/common/tdatablock.h +++ b/include/common/tdatablock.h @@ -234,9 +234,6 @@ int32_t buildSubmitReqFromDataBlock(SSubmitReq** pReq, const SArray* pDataBlocks char* buildCtbNameByGroupId(const char* stbName, uint64_t groupId); -SSubmitReq* tdBlockToSubmit(const SArray* pBlocks, const STSchema* pSchema, bool createTb, int64_t suid, - const char* stbFullName, int32_t vgId); - static FORCE_INLINE int32_t blockGetEncodeSize(const SSDataBlock* pBlock) { return blockDataGetSerialMetaSize(pBlock->info.numOfCols) + blockDataGetSize(pBlock); } diff --git a/include/common/tmsg.h b/include/common/tmsg.h index 935e2f6626..fb32cb382a 100644 --- a/include/common/tmsg.h +++ b/include/common/tmsg.h @@ -1134,14 +1134,16 @@ void tFreeSMAlterStbRsp(SMAlterStbRsp* pRsp); int32_t tSerializeSTableMetaRsp(void* buf, int32_t bufLen, STableMetaRsp* pRsp); int32_t tDeserializeSTableMetaRsp(void* buf, int32_t bufLen, STableMetaRsp* pRsp); void tFreeSTableMetaRsp(STableMetaRsp* pRsp); +void tFreeSTableIndexRsp(void *info); typedef struct { - SArray* pArray; // Array of STableMetaRsp -} STableMetaBatchRsp; + SArray* pMetaRsp; // Array of STableMetaRsp + SArray* pIndexRsp; // Array of STableIndexRsp; +} SSTbHbRsp; -int32_t tSerializeSTableMetaBatchRsp(void* buf, int32_t bufLen, STableMetaBatchRsp* pRsp); -int32_t tDeserializeSTableMetaBatchRsp(void* buf, int32_t bufLen, STableMetaBatchRsp* pRsp); -void tFreeSTableMetaBatchRsp(STableMetaBatchRsp* pRsp); +int32_t tSerializeSSTbHbRsp(void* buf, int32_t bufLen, SSTbHbRsp* pRsp); +int32_t tDeserializeSSTbHbRsp(void* buf, int32_t bufLen, SSTbHbRsp* pRsp); +void tFreeSSTbHbRsp(SSTbHbRsp* pRsp); typedef struct { int32_t numOfTables; @@ -2306,6 +2308,29 @@ int32_t tDecodeSMqOffset(SDecoder* decoder, SMqOffset* pOffset); int32_t tEncodeSMqCMCommitOffsetReq(SEncoder* encoder, const SMqCMCommitOffsetReq* pReq); int32_t tDecodeSMqCMCommitOffsetReq(SDecoder* decoder, SMqCMCommitOffsetReq* pReq); +// tqOffset +enum { + TMQ_OFFSET__SNAPSHOT = 1, + TMQ_OFFSET__LOG, +}; + +typedef struct { + int8_t type; + union { + struct { + int64_t uid; + int64_t ts; + }; + struct { + int64_t version; + }; + }; + char subKey[TSDB_SUBSCRIBE_KEY_LEN]; +} STqOffset; + +int32_t tEncodeSTqOffset(SEncoder* pEncoder, const STqOffset* pOffset); +int32_t tDecodeSTqOffset(SDecoder* pDecoder, STqOffset* pOffset); + typedef struct { char name[TSDB_TABLE_FNAME_LEN]; char stb[TSDB_TABLE_FNAME_LEN]; @@ -2502,7 +2527,11 @@ typedef struct { } STableIndexInfo; typedef struct { - SArray* pIndex; + char tbName[TSDB_TABLE_NAME_LEN]; + char dbFName[TSDB_DB_FNAME_LEN]; + uint64_t suid; + int32_t version; + SArray* pIndex; } STableIndexRsp; int32_t tSerializeSTableIndexRsp(void* buf, int32_t bufLen, const STableIndexRsp* pRsp); diff --git a/include/common/tmsgdef.h b/include/common/tmsgdef.h index 0f272aef9c..6863e3c336 100644 --- a/include/common/tmsgdef.h +++ b/include/common/tmsgdef.h @@ -141,7 +141,7 @@ enum { TD_DEF_MSG_TYPE(TDMT_MND_MQ_CONSUMER_RECOVER, "consumer-recover", SMqConsumerRecoverMsg, NULL) TD_DEF_MSG_TYPE(TDMT_MND_MQ_DO_REBALANCE, "do-rebalance", SMqDoRebalanceMsg, NULL) TD_DEF_MSG_TYPE(TDMT_MND_MQ_DROP_CGROUP, "drop-cgroup", SMqDropCGroupReq, SMqDropCGroupRsp) - TD_DEF_MSG_TYPE(TDMT_MND_MQ_COMMIT_OFFSET, "commit-offset", SMqCMCommitOffsetReq, SMqCMCommitOffsetRsp) + TD_DEF_MSG_TYPE(TDMT_MND_MQ_COMMIT_OFFSET, "mnode-commit-offset", SMqCMCommitOffsetReq, SMqCMCommitOffsetRsp) TD_DEF_MSG_TYPE(TDMT_MND_MQ_TIMER, "mq-tmr", SMTimerReq, NULL) TD_DEF_MSG_TYPE(TDMT_MND_TELEM_TIMER, "telem-tmr", SMTimerReq, SMTimerReq) TD_DEF_MSG_TYPE(TDMT_MND_TRANS_TIMER, "trans-tmr", NULL, NULL) @@ -177,6 +177,7 @@ enum { TD_DEF_MSG_TYPE(TDMT_VND_DROP_STB, "vnode-drop-stb", SVDropStbReq, NULL) TD_DEF_MSG_TYPE(TDMT_VND_MQ_VG_CHANGE, "vnode-mq-vg-change", SMqRebVgReq, SMqRebVgRsp) TD_DEF_MSG_TYPE(TDMT_VND_MQ_VG_DELETE, "vnode-mq-vg-delete", SMqVDeleteReq, SMqVDeleteRsp) + TD_DEF_MSG_TYPE(TDMT_VND_MQ_COMMIT_OFFSET, "vnode-commit-offset", STqOffset, STqOffset) TD_DEF_MSG_TYPE(TDMT_VND_CANCEL_TASK, "vnode-cancel-task", NULL, NULL) TD_DEF_MSG_TYPE(TDMT_VND_DROP_TASK, "vnode-drop-task", NULL, NULL) TD_DEF_MSG_TYPE(TDMT_VND_CREATE_TOPIC, "vnode-create-topic", NULL, NULL) diff --git a/include/libs/catalog/catalog.h b/include/libs/catalog/catalog.h index 9e573b7ed1..f3600a509e 100644 --- a/include/libs/catalog/catalog.h +++ b/include/libs/catalog/catalog.h @@ -98,14 +98,15 @@ typedef struct SCatalogCfg { uint32_t stbRentSec; } SCatalogCfg; -typedef struct SSTableMetaVersion { +typedef struct SSTableVersion { char dbFName[TSDB_DB_FNAME_LEN]; char stbName[TSDB_TABLE_NAME_LEN]; uint64_t dbId; uint64_t suid; int16_t sversion; - int16_t tversion; -} SSTableMetaVersion; + int16_t tversion; + int32_t smaVer; +} SSTableVersion; typedef struct SDbVgVersion { char dbFName[TSDB_DB_FNAME_LEN]; @@ -267,7 +268,7 @@ int32_t catalogAsyncGetAllMeta(SCatalog* pCtg, SRequestConnInfo* pConn, uint64_t int32_t catalogGetQnodeList(SCatalog* pCatalog, SRequestConnInfo* pConn, SArray* pQnodeList); -int32_t catalogGetExpiredSTables(SCatalog* pCatalog, SSTableMetaVersion** stables, uint32_t* num); +int32_t catalogGetExpiredSTables(SCatalog* pCatalog, SSTableVersion **stables, uint32_t *num); int32_t catalogGetExpiredDBs(SCatalog* pCatalog, SDbVgVersion** dbs, uint32_t* num); @@ -279,6 +280,8 @@ int32_t catalogGetIndexMeta(SCatalog* pCtg, SRequestConnInfo* pConn, const char* int32_t catalogGetTableIndex(SCatalog* pCtg, SRequestConnInfo* pConn, const SName* pTableName, SArray** pRes); +int32_t catalogUpdateTableIndex(SCatalog* pCtg, STableIndexRsp *pRsp); + int32_t catalogGetUdfInfo(SCatalog* pCtg, SRequestConnInfo* pConn, const char* funcName, SFuncInfo* pInfo); int32_t catalogChkAuth(SCatalog* pCtg, SRequestConnInfo* pConn, const char* user, const char* dbFName, AUTH_TYPE type, bool *pass); diff --git a/include/libs/function/functionMgt.h b/include/libs/function/functionMgt.h index b5641f2d07..f324d8fed1 100644 --- a/include/libs/function/functionMgt.h +++ b/include/libs/function/functionMgt.h @@ -142,6 +142,8 @@ typedef enum EFunctionType { FUNCTION_TYPE_FIRST_MERGE, FUNCTION_TYPE_LAST_PARTIAL, FUNCTION_TYPE_LAST_MERGE, + FUNCTION_TYPE_AVG_PARTIAL, + FUNCTION_TYPE_AVG_MERGE, // user defined funcion FUNCTION_TYPE_UDF = 10000 diff --git a/include/libs/sync/sync.h b/include/libs/sync/sync.h index 3a77cc1e19..a369b81d26 100644 --- a/include/libs/sync/sync.h +++ b/include/libs/sync/sync.h @@ -47,8 +47,9 @@ typedef enum { typedef enum { TAOS_SYNC_PROPOSE_SUCCESS = 0, TAOS_SYNC_PROPOSE_NOT_LEADER = 1, - TAOS_SYNC_PROPOSE_OTHER_ERROR = 2, - TAOS_SYNC_ONLY_ONE_REPLICA = 3, + TAOS_SYNC_ONLY_ONE_REPLICA = 2, + TAOS_SYNC_NOT_IN_NEW_CONFIG = 3, + TAOS_SYNC_OTHER_ERROR = 100, } ESyncProposeCode; typedef enum { @@ -110,6 +111,7 @@ typedef struct SSyncFSM { void (*FpRestoreFinishCb)(struct SSyncFSM* pFsm); void (*FpReConfigCb)(struct SSyncFSM* pFsm, const SRpcMsg* pMsg, SReConfigCbMeta cbMeta); + void (*FpLeaderTransferCb)(struct SSyncFSM* pFsm, const SRpcMsg* pMsg, SFsmCbMeta cbMeta); int32_t (*FpGetSnapshot)(struct SSyncFSM* pFsm, SSnapshot* pSnapshot); @@ -199,15 +201,13 @@ 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); + +// build SRpcMsg, need to call syncPropose with SRpcMsg +int32_t syncReconfigBuild(int64_t rid, const SSyncCfg* pNewCfg, SRpcMsg* pRpcMsg); int32_t syncLeaderTransfer(int64_t rid); int32_t syncLeaderTransferTo(int64_t rid, SNodeInfo newLeader); -// to be moved to static -void syncStartNormal(int64_t rid); -void syncStartStandBy(int64_t rid); - #ifdef __cplusplus } #endif diff --git a/include/libs/sync/syncTools.h b/include/libs/sync/syncTools.h index 68a33d48cb..fbf124bf47 100644 --- a/include/libs/sync/syncTools.h +++ b/include/libs/sync/syncTools.h @@ -467,6 +467,7 @@ typedef struct SyncLeaderTransfer { SRaftId srcId; SRaftId destId; */ + SNodeInfo newNodeInfo; SRaftId newLeaderId; } SyncLeaderTransfer; diff --git a/include/util/taoserror.h b/include/util/taoserror.h index 03308e395f..8d9951f9e3 100644 --- a/include/util/taoserror.h +++ b/include/util/taoserror.h @@ -695,6 +695,8 @@ 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) +//index +#define TSDB_CODE_INDEX_REBUILDING TAOS_DEF_ERROR_CODE(0, 0x3200) #ifdef __cplusplus } diff --git a/source/client/src/clientHb.c b/source/client/src/clientHb.c index 2412b32df2..5db27ae438 100644 --- a/source/client/src/clientHb.c +++ b/source/client/src/clientHb.c @@ -99,15 +99,15 @@ static int32_t hbProcessDBInfoRsp(void *value, int32_t valueLen, struct SCatalog static int32_t hbProcessStbInfoRsp(void *value, int32_t valueLen, struct SCatalog *pCatalog) { int32_t code = 0; - STableMetaBatchRsp batchMetaRsp = {0}; - if (tDeserializeSTableMetaBatchRsp(value, valueLen, &batchMetaRsp) != 0) { + SSTbHbRsp hbRsp = {0}; + if (tDeserializeSSTbHbRsp(value, valueLen, &hbRsp) != 0) { terrno = TSDB_CODE_INVALID_MSG; return -1; } - int32_t numOfBatchs = taosArrayGetSize(batchMetaRsp.pArray); - for (int32_t i = 0; i < numOfBatchs; ++i) { - STableMetaRsp *rsp = taosArrayGet(batchMetaRsp.pArray, i); + int32_t numOfMeta = taosArrayGetSize(hbRsp.pMetaRsp); + for (int32_t i = 0; i < numOfMeta; ++i) { + STableMetaRsp *rsp = taosArrayGet(hbRsp.pMetaRsp, i); if (rsp->numOfColumns < 0) { tscDebug("hb remove stb, db:%s, stb:%s", rsp->dbFName, rsp->stbName); @@ -116,7 +116,7 @@ static int32_t hbProcessStbInfoRsp(void *value, int32_t valueLen, struct SCatalo tscDebug("hb update stb, db:%s, stb:%s", rsp->dbFName, rsp->stbName); if (rsp->pSchemas[0].colId != PRIMARYKEY_TIMESTAMP_COL_ID) { tscError("invalid colId[%" PRIi16 "] for the first column in table meta rsp msg", rsp->pSchemas[0].colId); - tFreeSTableMetaBatchRsp(&batchMetaRsp); + tFreeSSTbHbRsp(&hbRsp); return TSDB_CODE_TSC_INVALID_VALUE; } @@ -124,7 +124,17 @@ static int32_t hbProcessStbInfoRsp(void *value, int32_t valueLen, struct SCatalo } } - tFreeSTableMetaBatchRsp(&batchMetaRsp); + int32_t numOfIndex = taosArrayGetSize(hbRsp.pIndexRsp); + for (int32_t i = 0; i < numOfIndex; ++i) { + STableIndexRsp *rsp = taosArrayGet(hbRsp.pIndexRsp, i); + + catalogUpdateTableIndex(pCatalog, rsp); + } + + taosArrayDestroy(hbRsp.pIndexRsp); + hbRsp.pIndexRsp = NULL; + + tFreeSSTbHbRsp(&hbRsp); return TSDB_CODE_SUCCESS; } @@ -455,7 +465,7 @@ int32_t hbGetExpiredDBInfo(SClientHbKey *connKey, struct SCatalog *pCatalog, SCl } int32_t hbGetExpiredStbInfo(SClientHbKey *connKey, struct SCatalog *pCatalog, SClientHbReq *req) { - SSTableMetaVersion *stbs = NULL; + SSTableVersion *stbs = NULL; uint32_t stbNum = 0; int32_t code = 0; @@ -469,15 +479,16 @@ int32_t hbGetExpiredStbInfo(SClientHbKey *connKey, struct SCatalog *pCatalog, SC } for (int32_t i = 0; i < stbNum; ++i) { - SSTableMetaVersion *stb = &stbs[i]; + SSTableVersion *stb = &stbs[i]; stb->suid = htobe64(stb->suid); stb->sversion = htons(stb->sversion); stb->tversion = htons(stb->tversion); + stb->smaVer = htonl(stb->smaVer); } SKv kv = { .key = HEARTBEAT_KEY_STBINFO, - .valueLen = sizeof(SSTableMetaVersion) * stbNum, + .valueLen = sizeof(SSTableVersion) * stbNum, .value = stbs, }; diff --git a/source/client/src/clientImpl.c b/source/client/src/clientImpl.c index 93234a3e80..6c3132a3ca 100644 --- a/source/client/src/clientImpl.c +++ b/source/client/src/clientImpl.c @@ -206,7 +206,7 @@ int32_t execLocalCmd(SRequestObj* pRequest, SQuery* pQuery) { SRetrieveTableRsp* pRsp = NULL; int32_t code = qExecCommand(pQuery->pRoot, &pRsp); if (TSDB_CODE_SUCCESS == code && NULL != pRsp) { - code = setQueryResultFromRsp(&pRequest->body.resInfo, pRsp, false, false); + code = setQueryResultFromRsp(&pRequest->body.resInfo, pRsp, false, true); } return code; @@ -1415,7 +1415,7 @@ int32_t setResultDataPtr(SReqResultInfo* pResultInfo, TAOS_FIELD* pFields, int32 int32_t bytes = *(int32_t*)p; p += sizeof(int32_t); - // ASSERT(type == pFields[i].type && bytes == pFields[i].bytes); + /*ASSERT(type == pFields[i].type && bytes == pFields[i].bytes);*/ } int32_t* colLength = (int32_t*)p; diff --git a/source/client/src/tmq.c b/source/client/src/tmq.c index 7d49c4206f..10c6c3623a 100644 --- a/source/client/src/tmq.c +++ b/source/client/src/tmq.c @@ -132,6 +132,7 @@ typedef struct { // statistics int64_t pollCnt; // offset + int64_t committedOffset; int64_t currentOffset; // connection info int32_t vgId; @@ -193,6 +194,26 @@ typedef struct { void* userParam; } SMqCommitCbParam; +typedef struct { + tmq_t* tmq; + int8_t automatic; + int8_t async; + int8_t freeOffsets; + int32_t waitingRspNum; + int32_t totalRspNum; + tmq_resp_err_t rspErr; + tmq_commit_cb* userCb; + SArray* successfulOffsets; + SArray* failedOffsets; + void* userParam; + tsem_t rspSem; +} SMqCommitCbParamSet; + +typedef struct { + SMqCommitCbParamSet* params; + STqOffset* pOffset; +} SMqCommitCbParam2; + tmq_conf_t* tmq_conf_new() { tmq_conf_t* conf = taosMemoryCalloc(1, sizeof(tmq_conf_t)); conf->withTbName = false; @@ -343,6 +364,139 @@ int32_t tmqCommitCb(void* param, const SDataBuf* pMsg, int32_t code) { return 0; } +int32_t tmqCommitCb2(void* param, const SDataBuf* pBuf, int32_t code) { + SMqCommitCbParam2* pParam = (SMqCommitCbParam2*)param; + SMqCommitCbParamSet* pParamSet = (SMqCommitCbParamSet*)pParam->params; + // push into array + if (code == 0) { + taosArrayPush(pParamSet->failedOffsets, &pParam->pOffset); + } else { + taosArrayPush(pParamSet->successfulOffsets, &pParam->pOffset); + } + + // count down waiting rsp + int32_t waitingRspNum = atomic_sub_fetch_32(&pParamSet->waitingRspNum, 1); + ASSERT(waitingRspNum >= 0); + + if (waitingRspNum == 0) { + // if no more waiting rsp + if (pParamSet->async) { + // call async cb func + if (pParamSet->automatic && pParamSet->tmq->commitCb) { + pParamSet->tmq->commitCb(pParamSet->tmq, pParamSet->rspErr, NULL, pParamSet->tmq->commitCbUserParam); + } else if (!pParamSet->automatic && pParamSet->userCb) { + // sem post + pParamSet->userCb(pParamSet->tmq, pParamSet->rspErr, NULL, pParamSet->userParam); + } + } + + taosArrayDestroyP(pParamSet->successfulOffsets, taosMemoryFree); + taosArrayDestroyP(pParamSet->failedOffsets, taosMemoryFree); + } + return 0; +} + +int32_t tmqCommitInner2(tmq_t* tmq, const tmq_topic_vgroup_list_t* offsets, int8_t automatic, int8_t async, + tmq_commit_cb* userCb, void* userParam) { + int32_t code = -1; + + SMqCommitCbParamSet* pParamSet = taosMemoryCalloc(1, sizeof(SMqCommitCbParamSet)); + if (pParamSet == NULL) { + terrno = TSDB_CODE_OUT_OF_MEMORY; + return -1; + } + pParamSet->tmq = tmq; + pParamSet->automatic = automatic; + pParamSet->async = async; + pParamSet->freeOffsets = 1; + pParamSet->userCb = userCb; + pParamSet->userParam = userParam; + tsem_init(&pParamSet->rspSem, 0, 0); + + for (int32_t i = 0; i < taosArrayGetSize(tmq->clientTopics); i++) { + SMqClientTopic* pTopic = taosArrayGet(tmq->clientTopics, i); + for (int32_t j = 0; j < taosArrayGetSize(pTopic->vgs); j++) { + SMqClientVg* pVg = taosArrayGet(pTopic->vgs, i); + STqOffset* pOffset = taosMemoryCalloc(1, sizeof(STqOffset)); + if (pOffset == NULL) { + terrno = TSDB_CODE_OUT_OF_MEMORY; + return -1; + } + int32_t tlen = strlen(tmq->groupId); + memcpy(pOffset->subKey, tmq->groupId, tlen); + pOffset->subKey[tlen] = TMQ_SEPARATOR; + strcpy(pOffset->subKey + tlen + 1, pTopic->topicName); + int32_t len; + int32_t code; + tEncodeSize(tEncodeSTqOffset, pOffset, len, code); + if (code < 0) { + ASSERT(0); + } + void* buf = taosMemoryCalloc(1, sizeof(SMsgHead) + len); + ((SMsgHead*)buf)->vgId = htonl(pVg->vgId); + + void* abuf = POINTER_SHIFT(buf, sizeof(SMsgHead)); + + SEncoder encoder; + tEncoderInit(&encoder, abuf, len); + tEncodeSTqOffset(&encoder, pOffset); + + // build param + SMqCommitCbParam2* pParam = taosMemoryCalloc(1, sizeof(SMqCommitCbParam2)); + pParam->params = pParamSet; + pParam->pOffset = pOffset; + + // build send info + SMsgSendInfo* pMsgSendInfo = taosMemoryCalloc(1, sizeof(SMsgSendInfo)); + if (pMsgSendInfo == NULL) { + // TODO + continue; + } + pMsgSendInfo->msgInfo = (SDataBuf){ + .pData = buf, + .len = len, + .handle = NULL, + }; + + pMsgSendInfo->requestId = generateRequestId(); + pMsgSendInfo->requestObjRefId = 0; + pMsgSendInfo->param = pParam; + pMsgSendInfo->fp = tmqCommitCb2; + pMsgSendInfo->msgType = TDMT_MND_MQ_COMMIT_OFFSET; + // send msg + + SEpSet epSet = getEpSet_s(&tmq->pTscObj->pAppInfo->mgmtEp); + int64_t transporterId = 0; + asyncSendMsgToServer(tmq->pTscObj->pAppInfo->pTransporter, &epSet, &transporterId, pMsgSendInfo); + pParamSet->waitingRspNum++; + pParamSet->totalRspNum++; + } + } + + if (!async) { + tsem_wait(&pParamSet->rspSem); + code = pParamSet->rspErr; + tsem_destroy(&pParamSet->rspSem); + } else { + code = 0; + } + + if (code != 0 && async) { + if (automatic) { + tmq->commitCb(tmq, code, NULL, tmq->commitCbUserParam); + } else { + userCb(tmq, code, NULL, userParam); + } + } + + if (!async) { + taosArrayDestroyP(pParamSet->successfulOffsets, taosMemoryFree); + taosArrayDestroyP(pParamSet->failedOffsets, taosMemoryFree); + } + + return 0; +} + int32_t tmqCommitInner(tmq_t* tmq, const tmq_topic_vgroup_list_t* offsets, int8_t automatic, int8_t async, tmq_commit_cb* userCb, void* userParam) { SMqCMCommitOffsetReq req; @@ -890,12 +1044,13 @@ bool tmqUpdateEp(tmq_t* tmq, int32_t epoch, SMqAskEpRsp* pRsp) { sprintf(vgKey, "%s:%d", topic.topicName, pVgEp->vgId); int64_t* pOffset = taosHashGet(pHash, vgKey, strlen(vgKey)); int64_t offset = pVgEp->offset; - tscDebug("consumer %ld epoch %d vg %d offset og to %ld", tmq->consumerId, epoch, pVgEp->vgId, offset); + tscDebug("consumer %ld(epoch %d) original offset of vg %d is %ld", tmq->consumerId, epoch, pVgEp->vgId, offset); if (pOffset != NULL) { offset = *pOffset; - tscDebug("consumer %ld epoch %d vg %d found %s", tmq->consumerId, epoch, pVgEp->vgId, vgKey); + tscDebug("consumer %ld(epoch %d) receive offset of vg %d, full key is %s", tmq->consumerId, epoch, pVgEp->vgId, + vgKey); } - tscDebug("consumer %ld epoch %d vg %d offset set to %ld", tmq->consumerId, epoch, pVgEp->vgId, offset); + tscDebug("consumer %ld(epoch %d) offset of vg %d updated to %ld", tmq->consumerId, epoch, pVgEp->vgId, offset); SMqClientVg clientVg = { .pollCnt = 0, .currentOffset = offset, @@ -1226,9 +1381,8 @@ SMqRspObj* tmqHandleAllRsp(tmq_t* tmq, int64_t timeout, bool pollIfReset) { if (rspWrapper->tmqRspType == TMQ_MSG_TYPE__POLL_RSP) { SMqPollRspWrapper* pollRspWrapper = (SMqPollRspWrapper*)rspWrapper; /*atomic_sub_fetch_32(&tmq->readyRequest, 1);*/ - /*printf("handle poll rsp %d\n", rspMsg->head.mqMsgType);*/ - if (pollRspWrapper->msg.head.epoch == atomic_load_32(&tmq->epoch)) { - /*printf("epoch match\n");*/ + int32_t consumerEpoch = atomic_load_32(&tmq->epoch); + if (pollRspWrapper->msg.head.epoch == consumerEpoch) { SMqClientVg* pVg = pollRspWrapper->vgHandle; /*printf("vg %d offset %ld up to %ld\n", pVg->vgId, pVg->currentOffset, rspMsg->msg.rspOffset);*/ pVg->currentOffset = pollRspWrapper->msg.rspOffset; @@ -1243,7 +1397,8 @@ SMqRspObj* tmqHandleAllRsp(tmq_t* tmq, int64_t timeout, bool pollIfReset) { taosFreeQitem(pollRspWrapper); return pRsp; } else { - /*printf("epoch mismatch\n");*/ + tscDebug("msg discard since epoch mismatch: msg epoch %d, consumer epoch %d\n", pollRspWrapper->msg.head.epoch, + consumerEpoch); taosFreeQitem(pollRspWrapper); } } else { @@ -1263,10 +1418,14 @@ TAOS_RES* tmq_consumer_poll(tmq_t* tmq, int64_t timeout) { SMqRspObj* rspObj; int64_t startTime = taosGetTimestampMs(); +#if 0 + tmqHandleAllDelayedTask(tmq); + tmqPollImpl(tmq, timeout); rspObj = tmqHandleAllRsp(tmq, timeout, false); if (rspObj) { return (TAOS_RES*)rspObj; } +#endif // in no topic status also need process delayed task if (atomic_load_8(&tmq->status) == TMQ_CONSUMER_STATUS__INIT) { @@ -1359,8 +1518,7 @@ const char* tmq_get_table_name(TAOS_RES* res) { pRspObj->resIter >= pRspObj->rsp.blockNum) { return NULL; } - const char* name = taosArrayGetP(pRspObj->rsp.blockTbName, pRspObj->resIter); - return name; + return (const char*)taosArrayGetP(pRspObj->rsp.blockTbName, pRspObj->resIter); } return NULL; } diff --git a/source/common/src/tdatablock.c b/source/common/src/tdatablock.c index 5affcf5100..5a2aaed74e 100644 --- a/source/common/src/tdatablock.c +++ b/source/common/src/tdatablock.c @@ -1729,173 +1729,6 @@ char* buildCtbNameByGroupId(const char* stbName, uint64_t groupId) { return rname.childTableName; } -SSubmitReq* tdBlockToSubmit(const SArray* pBlocks, const STSchema* pTSchema, bool createTb, int64_t suid, - const char* stbFullName, int32_t vgId) { - SSubmitReq* ret = NULL; - SArray* tagArray = taosArrayInit(1, sizeof(STagVal)); - if (!tagArray) { - terrno = TSDB_CODE_OUT_OF_MEMORY; - return NULL; - } - - // cal size - int32_t cap = sizeof(SSubmitReq); - int32_t sz = taosArrayGetSize(pBlocks); - for (int32_t i = 0; i < sz; i++) { - SSDataBlock* pDataBlock = taosArrayGet(pBlocks, i); - int32_t rows = pDataBlock->info.rows; - // TODO min - int32_t rowSize = pDataBlock->info.rowSize; - int32_t maxLen = TD_ROW_MAX_BYTES_FROM_SCHEMA(pTSchema); - int32_t schemaLen = 0; - - if (createTb) { - SVCreateTbReq createTbReq = {0}; - char* cname = buildCtbNameByGroupId(stbFullName, pDataBlock->info.groupId); - createTbReq.name = cname; - createTbReq.flags = 0; - createTbReq.type = TSDB_CHILD_TABLE; - createTbReq.ctb.suid = suid; - - STagVal tagVal = {.cid = pDataBlock->info.numOfCols + 1, - .type = TSDB_DATA_TYPE_UBIGINT, - .i64 = (int64_t)pDataBlock->info.groupId, - }; - STag* pTag = NULL; - taosArrayClear(tagArray); - taosArrayPush(tagArray, &tagVal); - tTagNew(tagArray, 1, false, &pTag); - if (pTag == NULL) { - tdDestroySVCreateTbReq(&createTbReq); - taosArrayDestroy(tagArray); - return NULL; - } - createTbReq.ctb.pTag = (uint8_t*)pTag; - - int32_t code; - tEncodeSize(tEncodeSVCreateTbReq, &createTbReq, schemaLen, code); - - tdDestroySVCreateTbReq(&createTbReq); - if (code < 0) { - taosArrayDestroy(tagArray); - return NULL; - } - } - - cap += sizeof(SSubmitBlk) + schemaLen + rows * maxLen; - } - - // assign data - // TODO - ret = taosMemoryCalloc(1, cap + 46); - ret = POINTER_SHIFT(ret, 46); - ret->header.vgId = vgId; - ret->version = htonl(1); - ret->length = sizeof(SSubmitReq); - ret->numOfBlocks = htonl(sz); - - void* submitBlk = POINTER_SHIFT(ret, sizeof(SSubmitReq)); - for (int32_t i = 0; i < sz; i++) { - SSDataBlock* pDataBlock = taosArrayGet(pBlocks, i); - - SSubmitBlk* blkHead = submitBlk; - blkHead->numOfRows = htons(pDataBlock->info.rows); - blkHead->sversion = htonl(pTSchema->version); - // TODO - blkHead->suid = htobe64(suid); - // uid is assigned by vnode - blkHead->uid = 0; - - int32_t rows = pDataBlock->info.rows; - /*int32_t maxLen = TD_ROW_MAX_BYTES_FROM_SCHEMA(pTSchema);*/ - /*blkHead->dataLen = htonl(rows * maxLen);*/ - blkHead->dataLen = 0; - - void* blockData = POINTER_SHIFT(submitBlk, sizeof(SSubmitBlk)); - - int32_t schemaLen = 0; - if (createTb) { - SVCreateTbReq createTbReq = {0}; - char* cname = buildCtbNameByGroupId(stbFullName, pDataBlock->info.groupId); - createTbReq.name = cname; - createTbReq.flags = 0; - createTbReq.type = TSDB_CHILD_TABLE; - createTbReq.ctb.suid = suid; - - STagVal tagVal = {.cid = pDataBlock->info.numOfCols + 1, - .type = TSDB_DATA_TYPE_UBIGINT, - .i64 = (int64_t)pDataBlock->info.groupId, - }; - taosArrayClear(tagArray); - taosArrayPush(tagArray, &tagVal); - STag* pTag = NULL; - tTagNew(tagArray, 1, false, &pTag); - if (pTag == NULL) { - tdDestroySVCreateTbReq(&createTbReq); - taosArrayDestroy(tagArray); - taosMemoryFreeClear(ret); - return NULL; - } - createTbReq.ctb.pTag = (uint8_t*)pTag; - - int32_t code; - tEncodeSize(tEncodeSVCreateTbReq, &createTbReq, schemaLen, code); - if (code < 0) { - tdDestroySVCreateTbReq(&createTbReq); - taosArrayDestroy(tagArray); - taosMemoryFreeClear(ret); - return NULL; - } - - SEncoder encoder = {0}; - tEncoderInit(&encoder, blockData, schemaLen); - code = tEncodeSVCreateTbReq(&encoder, &createTbReq); - tEncoderClear(&encoder); - tdDestroySVCreateTbReq(&createTbReq); - - if (code < 0) { - taosArrayDestroy(tagArray); - taosMemoryFreeClear(ret); - return NULL; - } - } - blkHead->schemaLen = htonl(schemaLen); - - STSRow* rowData = POINTER_SHIFT(blockData, schemaLen); - - for (int32_t j = 0; j < rows; j++) { - SRowBuilder rb = {0}; - tdSRowInit(&rb, pTSchema->version); - tdSRowSetTpInfo(&rb, pTSchema->numOfCols, pTSchema->flen); - tdSRowResetBuf(&rb, rowData); - - for (int32_t k = 0; k < pTSchema->numOfCols; k++) { - const STColumn* pColumn = &pTSchema->columns[k]; - SColumnInfoData* pColData = taosArrayGet(pDataBlock->pDataBlock, k); - if (colDataIsNull_s(pColData, j)) { - tdAppendColValToRow(&rb, pColumn->colId, pColumn->type, TD_VTYPE_NONE, NULL, false, pColumn->offset, k); - } else { - void* data = colDataGetData(pColData, j); - tdAppendColValToRow(&rb, pColumn->colId, pColumn->type, TD_VTYPE_NORM, data, true, pColumn->offset, k); - } - } - int32_t rowLen = TD_ROW_LEN(rowData); - rowData = POINTER_SHIFT(rowData, rowLen); - blkHead->dataLen += rowLen; - } - int32_t dataLen = blkHead->dataLen; - blkHead->dataLen = htonl(dataLen); - - ret->length += sizeof(SSubmitBlk) + schemaLen + dataLen; - blkHead = POINTER_SHIFT(blkHead, schemaLen + dataLen); - /*submitBlk = blkHead;*/ - } - - ret->length = htonl(ret->length); - taosArrayDestroy(tagArray); - return ret; -} - void blockCompressEncode(const SSDataBlock* pBlock, char* data, int32_t* dataLen, int32_t numOfCols, int8_t needCompress) { // todo extract method diff --git a/source/common/src/tmsg.c b/source/common/src/tmsg.c index 2ba2253780..2f7ca249ef 100644 --- a/source/common/src/tmsg.c +++ b/source/common/src/tmsg.c @@ -2438,6 +2438,10 @@ int32_t tSerializeSTableIndexRsp(void *buf, int32_t bufLen, const STableIndexRsp tEncoderInit(&encoder, buf, bufLen); if (tStartEncode(&encoder) < 0) return -1; + if (tEncodeCStr(&encoder, pRsp->tbName) < 0) return -1; + if (tEncodeCStr(&encoder, pRsp->dbFName) < 0) return -1; + if (tEncodeU64(&encoder, pRsp->suid) < 0) return -1; + if (tEncodeI32(&encoder, pRsp->version) < 0) return -1; int32_t num = taosArrayGetSize(pRsp->pIndex); if (tEncodeI32(&encoder, num) < 0) return -1; if (num > 0) { @@ -2472,6 +2476,10 @@ int32_t tDeserializeSTableIndexRsp(void *buf, int32_t bufLen, STableIndexRsp *pR tDecoderInit(&decoder, buf, bufLen); if (tStartDecode(&decoder) < 0) return -1; + if (tDecodeCStrTo(&decoder, pRsp->tbName) < 0) return -1; + if (tDecodeCStrTo(&decoder, pRsp->dbFName) < 0) return -1; + if (tDecodeU64(&decoder, &pRsp->suid) < 0) return -1; + if (tDecodeI32(&decoder, &pRsp->version) < 0) return -1; int32_t num = 0; if (tDecodeI32(&decoder, &num) < 0) return -1; if (num > 0) { @@ -2631,18 +2639,35 @@ int32_t tSerializeSTableMetaRsp(void *buf, int32_t bufLen, STableMetaRsp *pRsp) return tlen; } -int32_t tSerializeSTableMetaBatchRsp(void *buf, int32_t bufLen, STableMetaBatchRsp *pRsp) { +int32_t tSerializeSSTbHbRsp(void *buf, int32_t bufLen, SSTbHbRsp *pRsp) { SEncoder encoder = {0}; tEncoderInit(&encoder, buf, bufLen); if (tStartEncode(&encoder) < 0) return -1; - int32_t numOfBatch = taosArrayGetSize(pRsp->pArray); - if (tEncodeI32(&encoder, numOfBatch) < 0) return -1; - for (int32_t i = 0; i < numOfBatch; ++i) { - STableMetaRsp *pMetaRsp = taosArrayGet(pRsp->pArray, i); + int32_t numOfMeta = taosArrayGetSize(pRsp->pMetaRsp); + if (tEncodeI32(&encoder, numOfMeta) < 0) return -1; + for (int32_t i = 0; i < numOfMeta; ++i) { + STableMetaRsp *pMetaRsp = taosArrayGet(pRsp->pMetaRsp, i); if (tEncodeSTableMetaRsp(&encoder, pMetaRsp) < 0) return -1; } + + int32_t numOfIndex = taosArrayGetSize(pRsp->pIndexRsp); + if (tEncodeI32(&encoder, numOfIndex) < 0) return -1; + for (int32_t i = 0; i < numOfIndex; ++i) { + STableIndexRsp *pIndexRsp = taosArrayGet(pRsp->pIndexRsp, i); + if (tEncodeCStr(&encoder, pIndexRsp->tbName) < 0) return -1; + if (tEncodeCStr(&encoder, pIndexRsp->dbFName) < 0) return -1; + if (tEncodeU64(&encoder, pIndexRsp->suid) < 0) return -1; + if (tEncodeI32(&encoder, pIndexRsp->version) < 0) return -1; + int32_t num = taosArrayGetSize(pIndexRsp->pIndex); + if (tEncodeI32(&encoder, num) < 0) return -1; + for (int32_t i = 0; i < num; ++i) { + STableIndexInfo *pInfo = (STableIndexInfo *)taosArrayGet(pIndexRsp->pIndex, i); + if (tSerializeSTableIndexInfo(&encoder, pInfo) < 0) return -1; + } + } + tEndEncode(&encoder); int32_t tlen = encoder.pos; @@ -2662,26 +2687,58 @@ int32_t tDeserializeSTableMetaRsp(void *buf, int32_t bufLen, STableMetaRsp *pRsp return 0; } -int32_t tDeserializeSTableMetaBatchRsp(void *buf, int32_t bufLen, STableMetaBatchRsp *pRsp) { +int32_t tDeserializeSSTbHbRsp(void *buf, int32_t bufLen, SSTbHbRsp *pRsp) { SDecoder decoder = {0}; tDecoderInit(&decoder, buf, bufLen); if (tStartDecode(&decoder) < 0) return -1; - int32_t numOfBatch = taosArrayGetSize(pRsp->pArray); - if (tDecodeI32(&decoder, &numOfBatch) < 0) return -1; - - pRsp->pArray = taosArrayInit(numOfBatch, sizeof(STableMetaRsp)); - if (pRsp->pArray == NULL) { + int32_t numOfMeta = 0; + if (tDecodeI32(&decoder, &numOfMeta) < 0) return -1; + pRsp->pMetaRsp = taosArrayInit(numOfMeta, sizeof(STableMetaRsp)); + if (pRsp->pMetaRsp == NULL) { terrno = TSDB_CODE_OUT_OF_MEMORY; return -1; } - for (int32_t i = 0; i < numOfBatch; ++i) { + for (int32_t i = 0; i < numOfMeta; ++i) { STableMetaRsp tableMetaRsp = {0}; if (tDecodeSTableMetaRsp(&decoder, &tableMetaRsp) < 0) return -1; - taosArrayPush(pRsp->pArray, &tableMetaRsp); + taosArrayPush(pRsp->pMetaRsp, &tableMetaRsp); } + + int32_t numOfIndex = 0; + if (tDecodeI32(&decoder, &numOfIndex) < 0) return -1; + + pRsp->pIndexRsp = taosArrayInit(numOfIndex, sizeof(STableIndexRsp)); + if (pRsp->pIndexRsp == NULL) { + terrno = TSDB_CODE_OUT_OF_MEMORY; + return -1; + } + + for (int32_t i = 0; i < numOfIndex; ++i) { + STableIndexRsp tableIndexRsp = {0}; + if (tDecodeCStrTo(&decoder, tableIndexRsp.tbName) < 0) return -1; + if (tDecodeCStrTo(&decoder, tableIndexRsp.dbFName) < 0) return -1; + if (tDecodeU64(&decoder, &tableIndexRsp.suid) < 0) return -1; + if (tDecodeI32(&decoder, &tableIndexRsp.version) < 0) return -1; + int32_t num = 0; + if (tDecodeI32(&decoder, &num) < 0) return -1; + if (num > 0) { + tableIndexRsp.pIndex = taosArrayInit(num, sizeof(STableIndexInfo)); + if (NULL == tableIndexRsp.pIndex) return -1; + STableIndexInfo info; + for (int32_t i = 0; i < num; ++i) { + if (tDeserializeSTableIndexInfo(&decoder, &info) < 0) return -1; + if (NULL == taosArrayPush(tableIndexRsp.pIndex, &info)) { + taosMemoryFree(info.expr); + return -1; + } + } + } + taosArrayPush(pRsp->pIndexRsp, &tableIndexRsp); + } + tEndDecode(&decoder); tDecoderClear(&decoder); @@ -2690,14 +2747,32 @@ int32_t tDeserializeSTableMetaBatchRsp(void *buf, int32_t bufLen, STableMetaBatc void tFreeSTableMetaRsp(STableMetaRsp *pRsp) { taosMemoryFreeClear(pRsp->pSchemas); } -void tFreeSTableMetaBatchRsp(STableMetaBatchRsp *pRsp) { - int32_t numOfBatch = taosArrayGetSize(pRsp->pArray); - for (int32_t i = 0; i < numOfBatch; ++i) { - STableMetaRsp *pMetaRsp = taosArrayGet(pRsp->pArray, i); +void tFreeSTableIndexRsp(void *info) { + if (NULL == info) { + return; + } + + STableIndexRsp *pInfo = (STableIndexRsp *)info; + + taosArrayDestroyEx(pInfo->pIndex, tFreeSTableIndexInfo); +} + +void tFreeSSTbHbRsp(SSTbHbRsp *pRsp) { + int32_t numOfMeta = taosArrayGetSize(pRsp->pMetaRsp); + for (int32_t i = 0; i < numOfMeta; ++i) { + STableMetaRsp *pMetaRsp = taosArrayGet(pRsp->pMetaRsp, i); tFreeSTableMetaRsp(pMetaRsp); } - taosArrayDestroy(pRsp->pArray); + taosArrayDestroy(pRsp->pMetaRsp); + + int32_t numOfIndex = taosArrayGetSize(pRsp->pIndexRsp); + for (int32_t i = 0; i < numOfIndex; ++i) { + STableIndexRsp *pIndexRsp = taosArrayGet(pRsp->pIndexRsp, i); + tFreeSTableIndexRsp(pIndexRsp); + } + + taosArrayDestroy(pRsp->pIndexRsp); } int32_t tSerializeSShowRsp(void *buf, int32_t bufLen, SShowRsp *pRsp) { @@ -4776,3 +4851,32 @@ void tFreeSMAlterStbRsp(SMAlterStbRsp *pRsp) { taosMemoryFree(pRsp->pMeta); } } + +int32_t tEncodeSTqOffset(SEncoder *pEncoder, const STqOffset *pOffset) { + if (tEncodeI8(pEncoder, pOffset->type) < 0) return -1; + if (pOffset->type == TMQ_OFFSET__SNAPSHOT) { + if (tEncodeI64(pEncoder, pOffset->uid) < 0) return -1; + if (tEncodeI64(pEncoder, pOffset->ts) < 0) return -1; + } else if (pOffset->type == TMQ_OFFSET__LOG) { + if (tEncodeI64(pEncoder, pOffset->version) < 0) return -1; + } else { + ASSERT(0); + } + if (tEncodeCStr(pEncoder, pOffset->subKey) < 0) return -1; + return 0; +} + +int32_t tDecodeSTqOffset(SDecoder *pDecoder, STqOffset *pOffset) { + if (tDecodeI8(pDecoder, &pOffset->type) < 0) return -1; + if (pOffset->type == TMQ_OFFSET__SNAPSHOT) { + if (tDecodeI64(pDecoder, &pOffset->uid) < 0) return -1; + if (tDecodeI64(pDecoder, &pOffset->ts) < 0) return -1; + } else if (pOffset->type == TMQ_OFFSET__LOG) { + if (tDecodeI64(pDecoder, &pOffset->version) < 0) return -1; + } else { + ASSERT(0); + } + if (tDecodeCStrTo(pDecoder, pOffset->subKey) < 0) return -1; + return 0; +} + diff --git a/source/dnode/mgmt/mgmt_vnode/src/vmHandle.c b/source/dnode/mgmt/mgmt_vnode/src/vmHandle.c index ee120576c3..750006d05f 100644 --- a/source/dnode/mgmt/mgmt_vnode/src/vmHandle.c +++ b/source/dnode/mgmt/mgmt_vnode/src/vmHandle.c @@ -344,6 +344,7 @@ SArray *vmGetMsgHandles() { if (dmSetMgmtHandle(pArray, TDMT_VND_SUBMIT_RSMA, vmPutMsgToWriteQueue, 0) == NULL) goto _OVER; if (dmSetMgmtHandle(pArray, TDMT_VND_MQ_VG_CHANGE, vmPutMsgToWriteQueue, 0) == NULL) goto _OVER; if (dmSetMgmtHandle(pArray, TDMT_VND_MQ_VG_DELETE, vmPutMsgToWriteQueue, 0) == NULL) goto _OVER; + if (dmSetMgmtHandle(pArray, TDMT_VND_MQ_COMMIT_OFFSET, vmPutMsgToWriteQueue, 0) == NULL) goto _OVER; if (dmSetMgmtHandle(pArray, TDMT_VND_CONSUME, vmPutMsgToFetchQueue, 0) == NULL) goto _OVER; if (dmSetMgmtHandle(pArray, TDMT_VND_DELETE, vmPutMsgToWriteQueue, 0) == NULL) goto _OVER; if (dmSetMgmtHandle(pArray, TDMT_VND_QUERY_HEARTBEAT, vmPutMsgToFetchQueue, 0) == NULL) goto _OVER; diff --git a/source/dnode/mnode/impl/inc/mndSma.h b/source/dnode/mnode/impl/inc/mndSma.h index 4a80f619d3..63530b403d 100644 --- a/source/dnode/mnode/impl/inc/mndSma.h +++ b/source/dnode/mnode/impl/inc/mndSma.h @@ -26,6 +26,7 @@ int32_t mndInitSma(SMnode *pMnode); void mndCleanupSma(SMnode *pMnode); SSmaObj *mndAcquireSma(SMnode *pMnode, char *smaName); void mndReleaseSma(SMnode *pMnode, SSmaObj *pSma); +int32_t mndGetTableSma(SMnode *pMnode, char *tbFName, STableIndexRsp *rsp, bool *exist); #ifdef __cplusplus } diff --git a/source/dnode/mnode/impl/inc/mndStb.h b/source/dnode/mnode/impl/inc/mndStb.h index 28d3215b98..d3de4e3b3f 100644 --- a/source/dnode/mnode/impl/inc/mndStb.h +++ b/source/dnode/mnode/impl/inc/mndStb.h @@ -27,7 +27,7 @@ void mndCleanupStb(SMnode *pMnode); SStbObj *mndAcquireStb(SMnode *pMnode, char *stbName); void mndReleaseStb(SMnode *pMnode, SStbObj *pStb); SSdbRaw *mndStbActionEncode(SStbObj *pStb); -int32_t mndValidateStbInfo(SMnode *pMnode, SSTableMetaVersion *pStbs, int32_t numOfStbs, void **ppRsp, +int32_t mndValidateStbInfo(SMnode *pMnode, SSTableVersion *pStbs, int32_t numOfStbs, void **ppRsp, int32_t *pRspLen); int32_t mndGetNumOfStbs(SMnode *pMnode, char *dbName, int32_t *pNumOfStbs); diff --git a/source/dnode/mnode/impl/src/mndDb.c b/source/dnode/mnode/impl/src/mndDb.c index b4237466d4..cfe363fdf0 100644 --- a/source/dnode/mnode/impl/src/mndDb.c +++ b/source/dnode/mnode/impl/src/mndDb.c @@ -1366,7 +1366,7 @@ char *buildRetension(SArray *pRetension) { } static void dumpDbInfoData(SSDataBlock *pBlock, SDbObj *pDb, SShowObj *pShow, int32_t rows, int64_t numOfTables, - bool sysDb) { + bool sysDb, ESdbStatus objStatus) { int32_t cols = 0; int32_t bytes = pShow->pMeta->pSchemas[cols].bytes; @@ -1379,6 +1379,8 @@ static void dumpDbInfoData(SSDataBlock *pBlock, SDbObj *pDb, SShowObj *pShow, in } char *status = "ready"; + if (objStatus == SDB_STATUS_CREATING) status = "creating"; + if (objStatus == SDB_STATUS_DROPPING) status = "dropping"; char statusB[24] = {0}; STR_WITH_SIZE_TO_VARSTR(statusB, status, strlen(status)); @@ -1503,8 +1505,8 @@ static void dumpDbInfoData(SSDataBlock *pBlock, SDbObj *pDb, SShowObj *pShow, in pColInfo = taosArrayGet(pBlock->pDataBlock, cols++); colDataAppend(pColInfo, rows, (const char *)statusB, false); -// pColInfo = taosArrayGet(pBlock->pDataBlock, cols++); -// colDataAppend(pColInfo, rows, (const char *)&pDb->cfg.schemaless, false); + // pColInfo = taosArrayGet(pBlock->pDataBlock, cols++); + // colDataAppend(pColInfo, rows, (const char *)&pDb->cfg.schemaless, false); char *p = buildRetension(pDb->cfg.pRetensions); @@ -1548,29 +1550,30 @@ static bool mndGetTablesOfDbFp(SMnode *pMnode, void *pObj, void *p1, void *p2, v } static int32_t mndRetrieveDbs(SRpcMsg *pReq, SShowObj *pShow, SSDataBlock *pBlock, int32_t rowsCapacity) { - SMnode *pMnode = pReq->info.node; - SSdb *pSdb = pMnode->pSdb; - int32_t numOfRows = 0; - SDbObj *pDb = NULL; + SMnode *pMnode = pReq->info.node; + SSdb *pSdb = pMnode->pSdb; + int32_t numOfRows = 0; + SDbObj *pDb = NULL; + ESdbStatus objStatus = 0; // Append the information_schema database into the result. if (!pShow->sysDbRsp) { SDbObj infoschemaDb = {0}; setInformationSchemaDbCfg(&infoschemaDb); - dumpDbInfoData(pBlock, &infoschemaDb, pShow, numOfRows, 14, true); + dumpDbInfoData(pBlock, &infoschemaDb, pShow, numOfRows, 14, true, 0); numOfRows += 1; SDbObj perfschemaDb = {0}; setPerfSchemaDbCfg(&perfschemaDb); - dumpDbInfoData(pBlock, &perfschemaDb, pShow, numOfRows, 3, true); + dumpDbInfoData(pBlock, &perfschemaDb, pShow, numOfRows, 3, true, 0); numOfRows += 1; pShow->sysDbRsp = true; } while (numOfRows < rowsCapacity) { - pShow->pIter = sdbFetch(pSdb, SDB_DB, pShow->pIter, (void **)&pDb); + pShow->pIter = sdbFetchAll(pSdb, SDB_DB, pShow->pIter, (void **)&pDb, &objStatus); if (pShow->pIter == NULL) { break; } @@ -1578,7 +1581,7 @@ static int32_t mndRetrieveDbs(SRpcMsg *pReq, SShowObj *pShow, SSDataBlock *pBloc int32_t numOfTables = 0; sdbTraverse(pSdb, SDB_VGROUP, mndGetTablesOfDbFp, &numOfTables, NULL, NULL); - dumpDbInfoData(pBlock, pDb, pShow, numOfRows, numOfTables, false); + dumpDbInfoData(pBlock, pDb, pShow, numOfRows, numOfTables, false, objStatus); numOfRows++; sdbRelease(pSdb, pDb); } diff --git a/source/dnode/mnode/impl/src/mndMain.c b/source/dnode/mnode/impl/src/mndMain.c index 892650fa18..3b2eafced8 100644 --- a/source/dnode/mnode/impl/src/mndMain.c +++ b/source/dnode/mnode/impl/src/mndMain.c @@ -380,22 +380,22 @@ void mndStop(SMnode *pMnode) { int32_t mndProcessSyncMsg(SRpcMsg *pMsg) { SMnode *pMnode = pMsg->info.node; SSyncMgmt *pMgmt = &pMnode->syncMgmt; - int32_t code = TAOS_SYNC_PROPOSE_OTHER_ERROR; + int32_t code = TAOS_SYNC_OTHER_ERROR; if (!syncEnvIsStart()) { mError("failed to process sync msg:%p type:%s since syncEnv stop", pMsg, TMSG_INFO(pMsg->msgType)); - return TAOS_SYNC_PROPOSE_OTHER_ERROR; + return TAOS_SYNC_OTHER_ERROR; } SSyncNode *pSyncNode = syncNodeAcquire(pMgmt->sync); if (pSyncNode == NULL) { mError("failed to process sync msg:%p type:%s since syncNode is null", pMsg, TMSG_INFO(pMsg->msgType)); - return TAOS_SYNC_PROPOSE_OTHER_ERROR; + return TAOS_SYNC_OTHER_ERROR; } if (mndAcquireSyncRef(pMnode) != 0) { mError("failed to process sync msg:%p type:%s since %s", pMsg, TMSG_INFO(pMsg->msgType), terrstr()); - return TAOS_SYNC_PROPOSE_OTHER_ERROR; + return TAOS_SYNC_OTHER_ERROR; } char logBuf[512] = {0}; @@ -456,7 +456,7 @@ int32_t mndProcessSyncMsg(SRpcMsg *pMsg) { tmsgSendRsp(&rsp); } else { mError("failed to process msg:%p since invalid type:%s", pMsg, TMSG_INFO(pMsg->msgType)); - code = TAOS_SYNC_PROPOSE_OTHER_ERROR; + code = TAOS_SYNC_OTHER_ERROR; } } else { if (pMsg->msgType == TDMT_SYNC_TIMEOUT) { @@ -497,7 +497,7 @@ int32_t mndProcessSyncMsg(SRpcMsg *pMsg) { tmsgSendRsp(&rsp); } else { mError("failed to process msg:%p since invalid type:%s", pMsg, TMSG_INFO(pMsg->msgType)); - code = TAOS_SYNC_PROPOSE_OTHER_ERROR; + code = TAOS_SYNC_OTHER_ERROR; } } diff --git a/source/dnode/mnode/impl/src/mndMnode.c b/source/dnode/mnode/impl/src/mndMnode.c index 95a2c3d935..f9749a969d 100644 --- a/source/dnode/mnode/impl/src/mndMnode.c +++ b/source/dnode/mnode/impl/src/mndMnode.c @@ -659,7 +659,7 @@ static int32_t mndRetrieveMnodes(SRpcMsg *pReq, SShowObj *pShow, SSDataBlock *pB int32_t numOfRows = 0; int32_t cols = 0; SMnodeObj *pObj = NULL; - ESdbStatus objStatus; + ESdbStatus objStatus = 0; char *pWrite; int64_t curMs = taosGetTimestampMs(); diff --git a/source/dnode/mnode/impl/src/mndProfile.c b/source/dnode/mnode/impl/src/mndProfile.c index bacdf2f366..4cfcf389aa 100644 --- a/source/dnode/mnode/impl/src/mndProfile.c +++ b/source/dnode/mnode/impl/src/mndProfile.c @@ -433,7 +433,7 @@ static int32_t mndProcessQueryHeartBeat(SMnode *pMnode, SRpcMsg *pMsg, SClientHb case HEARTBEAT_KEY_STBINFO: { void *rspMsg = NULL; int32_t rspLen = 0; - mndValidateStbInfo(pMnode, kv->value, kv->valueLen / sizeof(SSTableMetaVersion), &rspMsg, &rspLen); + mndValidateStbInfo(pMnode, kv->value, kv->valueLen / sizeof(SSTableVersion), &rspMsg, &rspLen); if (rspMsg && rspLen > 0) { SKv kv1 = {.key = HEARTBEAT_KEY_STBINFO, .valueLen = rspLen, .value = rspMsg}; taosArrayPush(hbRsp.info, &kv1); diff --git a/source/dnode/mnode/impl/src/mndSma.c b/source/dnode/mnode/impl/src/mndSma.c index c19b558f19..3f0d20348b 100644 --- a/source/dnode/mnode/impl/src/mndSma.c +++ b/source/dnode/mnode/impl/src/mndSma.c @@ -532,6 +532,7 @@ static int32_t mndCreateSma(SMnode *pMnode, SRpcMsg *pReq, SMCreateSmaReq *pCrea SStreamObj streamObj = {0}; tstrncpy(streamObj.name, pCreate->name, TSDB_STREAM_FNAME_LEN); tstrncpy(streamObj.sourceDb, pDb->name, TSDB_DB_FNAME_LEN); + tstrncpy(streamObj.targetDb, streamObj.sourceDb, TSDB_DB_FNAME_LEN); streamObj.createTime = taosGetTimestampMs(); streamObj.updateTime = streamObj.createTime; streamObj.uid = mndGenerateUid(pCreate->name, strlen(pCreate->name)); @@ -899,18 +900,31 @@ static int32_t mndGetSma(SMnode *pMnode, SUserIndexReq *indexReq, SUserIndexRsp return code; } -static int32_t mndGetTableSma(SMnode *pMnode, STableIndexReq *indexReq, STableIndexRsp *rsp, bool *exist) { +int32_t mndGetTableSma(SMnode *pMnode, char *tbFName, STableIndexRsp *rsp, bool *exist) { int32_t code = 0; SSmaObj *pSma = NULL; SSdb *pSdb = pMnode->pSdb; void *pIter = NULL; STableIndexInfo info; + SStbObj* pStb = mndAcquireStb(pMnode, tbFName); + if (NULL == pStb) { + *exist = false; + return TSDB_CODE_SUCCESS; + } + + strcpy(rsp->dbFName, pStb->db); + strcpy(rsp->tbName, pStb->name + strlen(pStb->db) + 1); + rsp->suid = pStb->uid; + rsp->version = pStb->smaVer; + mndReleaseStb(pMnode, pStb); + + while (1) { pIter = sdbFetch(pSdb, SDB_SMA, pIter, (void **)&pSma); if (pIter == NULL) break; - if (pSma->stb[0] != indexReq->tbFName[0] || strcmp(pSma->stb, indexReq->tbFName)) { + if (pSma->stb[0] != tbFName[0] || strcmp(pSma->stb, tbFName)) { continue; } @@ -1022,7 +1036,7 @@ static int32_t mndProcessGetTbSmaReq(SRpcMsg *pReq) { goto _OVER; } - code = mndGetTableSma(pMnode, &indexReq, &rsp, &exist); + code = mndGetTableSma(pMnode, indexReq.tbFName, &rsp, &exist); if (code) { goto _OVER; } @@ -1114,4 +1128,4 @@ static int32_t mndRetrieveSma(SRpcMsg *pReq, SShowObj *pShow, SSDataBlock *pBloc static void mndCancelGetNextSma(SMnode *pMnode, void *pIter) { SSdb *pSdb = pMnode->pSdb; sdbCancelFetch(pSdb, pIter); -} \ No newline at end of file +} diff --git a/source/dnode/mnode/impl/src/mndStb.c b/source/dnode/mnode/impl/src/mndStb.c index ad0c913fb8..6f89c97f83 100644 --- a/source/dnode/mnode/impl/src/mndStb.c +++ b/source/dnode/mnode/impl/src/mndStb.c @@ -27,6 +27,7 @@ #include "mndTrans.h" #include "mndUser.h" #include "mndVgroup.h" +#include "mndSma.h" #include "tname.h" #define STB_VER_NUMBER 1 @@ -1271,7 +1272,7 @@ static int32_t mndBuildStbSchemaImp(SDbObj *pDb, SStbObj *pStb, const char *tbNa return 0; } -static int32_t mndBuildStbSchema(SMnode *pMnode, const char *dbFName, const char *tbName, STableMetaRsp *pRsp) { +static int32_t mndBuildStbSchema(SMnode *pMnode, const char *dbFName, const char *tbName, STableMetaRsp *pRsp, int32_t *smaVer) { char tbFName[TSDB_TABLE_FNAME_LEN] = {0}; snprintf(tbFName, sizeof(tbFName), "%s.%s", dbFName, tbName); @@ -1288,6 +1289,10 @@ static int32_t mndBuildStbSchema(SMnode *pMnode, const char *dbFName, const char return -1; } + if (smaVer) { + *smaVer = pStb->smaVer; + } + int32_t code = mndBuildStbSchemaImp(pDb, pStb, tbName, pRsp); mndReleaseDb(pMnode, pDb); mndReleaseStb(pMnode, pStb); @@ -1634,7 +1639,7 @@ static int32_t mndProcessTableMetaReq(SRpcMsg *pReq) { } } else { mDebug("stb:%s.%s, start to retrieve meta", infoReq.dbFName, infoReq.tbName); - if (mndBuildStbSchema(pMnode, infoReq.dbFName, infoReq.tbName, &metaRsp) != 0) { + if (mndBuildStbSchema(pMnode, infoReq.dbFName, infoReq.tbName, &metaRsp, NULL) != 0) { goto _OVER; } } @@ -1667,51 +1672,86 @@ _OVER: return code; } -int32_t mndValidateStbInfo(SMnode *pMnode, SSTableMetaVersion *pStbVersions, int32_t numOfStbs, void **ppRsp, +int32_t mndValidateStbInfo(SMnode *pMnode, SSTableVersion *pStbVersions, int32_t numOfStbs, void **ppRsp, int32_t *pRspLen) { - STableMetaBatchRsp batchMetaRsp = {0}; - batchMetaRsp.pArray = taosArrayInit(numOfStbs, sizeof(STableMetaRsp)); - if (batchMetaRsp.pArray == NULL) { + SSTbHbRsp hbRsp = {0}; + hbRsp.pMetaRsp = taosArrayInit(numOfStbs, sizeof(STableMetaRsp)); + if (hbRsp.pMetaRsp == NULL) { terrno = TSDB_CODE_OUT_OF_MEMORY; return -1; } + hbRsp.pIndexRsp = taosArrayInit(numOfStbs, sizeof(STableIndexRsp)); + if (NULL == hbRsp.pIndexRsp) { + taosArrayDestroy(hbRsp.pMetaRsp); + terrno = TSDB_CODE_OUT_OF_MEMORY; + return -1; + } + for (int32_t i = 0; i < numOfStbs; ++i) { - SSTableMetaVersion *pStbVersion = &pStbVersions[i]; + SSTableVersion *pStbVersion = &pStbVersions[i]; pStbVersion->suid = be64toh(pStbVersion->suid); pStbVersion->sversion = ntohs(pStbVersion->sversion); pStbVersion->tversion = ntohs(pStbVersion->tversion); + pStbVersion->smaVer = ntohl(pStbVersion->smaVer); STableMetaRsp metaRsp = {0}; + int32_t smaVer = 0; mDebug("stb:%s.%s, start to retrieve meta", pStbVersion->dbFName, pStbVersion->stbName); - if (mndBuildStbSchema(pMnode, pStbVersion->dbFName, pStbVersion->stbName, &metaRsp) != 0) { + if (mndBuildStbSchema(pMnode, pStbVersion->dbFName, pStbVersion->stbName, &metaRsp, &smaVer) != 0) { metaRsp.numOfColumns = -1; metaRsp.suid = pStbVersion->suid; + taosArrayPush(hbRsp.pMetaRsp, &metaRsp); + continue; } if (pStbVersion->sversion != metaRsp.sversion || pStbVersion->tversion != metaRsp.tversion) { - taosArrayPush(batchMetaRsp.pArray, &metaRsp); + taosArrayPush(hbRsp.pMetaRsp, &metaRsp); } else { tFreeSTableMetaRsp(&metaRsp); } + + if (pStbVersion->smaVer && pStbVersion->smaVer != smaVer) { + bool exist = false; + char tbFName[TSDB_TABLE_FNAME_LEN]; + STableIndexRsp indexRsp = {0}; + indexRsp.pIndex = taosArrayInit(10, sizeof(STableIndexInfo)); + if (NULL == indexRsp.pIndex) { + terrno = TSDB_CODE_OUT_OF_MEMORY; + return -1; + } + + sprintf(tbFName, "%s.%s", pStbVersion->dbFName, pStbVersion->stbName); + int32_t code = mndGetTableSma(pMnode, tbFName, &indexRsp, &exist); + if (code || !exist) { + indexRsp.suid = pStbVersion->suid; + indexRsp.version = -1; + indexRsp.pIndex = NULL; + } + + strcpy(indexRsp.dbFName, pStbVersion->dbFName); + strcpy(indexRsp.tbName, pStbVersion->stbName); + + taosArrayPush(hbRsp.pIndexRsp, &indexRsp); + } } - int32_t rspLen = tSerializeSTableMetaBatchRsp(NULL, 0, &batchMetaRsp); + int32_t rspLen = tSerializeSSTbHbRsp(NULL, 0, &hbRsp); if (rspLen < 0) { - tFreeSTableMetaBatchRsp(&batchMetaRsp); + tFreeSSTbHbRsp(&hbRsp); terrno = TSDB_CODE_INVALID_MSG; return -1; } void *pRsp = taosMemoryMalloc(rspLen); if (pRsp == NULL) { - tFreeSTableMetaBatchRsp(&batchMetaRsp); + tFreeSSTbHbRsp(&hbRsp); terrno = TSDB_CODE_OUT_OF_MEMORY; return -1; } - tSerializeSTableMetaBatchRsp(pRsp, rspLen, &batchMetaRsp); - tFreeSTableMetaBatchRsp(&batchMetaRsp); + tSerializeSSTbHbRsp(pRsp, rspLen, &hbRsp); + tFreeSSTbHbRsp(&hbRsp); *ppRsp = pRsp; *pRspLen = rspLen; return 0; diff --git a/source/dnode/mnode/impl/src/mndSync.c b/source/dnode/mnode/impl/src/mndSync.c index e58012b1b7..9940037356 100644 --- a/source/dnode/mnode/impl/src/mndSync.c +++ b/source/dnode/mnode/impl/src/mndSync.c @@ -236,7 +236,7 @@ int32_t mndSyncPropose(SMnode *pMnode, SSdbRaw *pRaw, int32_t transId) { tsem_wait(&pMgmt->syncSem); } else if (code == TAOS_SYNC_PROPOSE_NOT_LEADER) { terrno = TSDB_CODE_APP_NOT_READY; - } else if (code == TAOS_SYNC_PROPOSE_OTHER_ERROR) { + } else if (code == TAOS_SYNC_OTHER_ERROR) { terrno = TSDB_CODE_SYN_INTERNAL_ERROR; } else { terrno = TSDB_CODE_APP_ERROR; @@ -254,13 +254,16 @@ int32_t mndSyncPropose(SMnode *pMnode, SSdbRaw *pRaw, int32_t transId) { void mndSyncStart(SMnode *pMnode) { SSyncMgmt *pMgmt = &pMnode->syncMgmt; syncSetMsgCb(pMgmt->sync, &pMnode->msgCb); + syncStart(pMgmt->sync); + mDebug("mnode sync started, id:%" PRId64 " standby:%d", pMgmt->sync, pMgmt->standby); +/* if (pMgmt->standby) { syncStartStandBy(pMgmt->sync); } else { syncStart(pMgmt->sync); } - mDebug("mnode sync started, id:%" PRId64 " standby:%d", pMgmt->sync, pMgmt->standby); +*/ } void mndSyncStop(SMnode *pMnode) {} diff --git a/source/dnode/vnode/src/inc/tq.h b/source/dnode/vnode/src/inc/tq.h index 5a8564bfd1..2ee0673ce5 100644 --- a/source/dnode/vnode/src/inc/tq.h +++ b/source/dnode/vnode/src/inc/tq.h @@ -41,7 +41,6 @@ extern "C" { #define tqTrace(...) do { if (tqDebugFlag & DEBUG_TRACE) { taosPrintLog("TQ ", DEBUG_TRACE, tqDebugFlag, __VA_ARGS__); }} while(0) // clang-format on -typedef struct STqOffsetCfg STqOffsetCfg; typedef struct STqOffsetStore STqOffsetStore; // tqRead @@ -127,14 +126,15 @@ typedef struct { } STqHandle; struct STQ { - char* path; - SHashObj* pushMgr; // consumerId -> STqHandle* - SHashObj* handles; // subKey -> STqHandle - SHashObj* pStreamTasks; // taksId -> SStreamTask - SVnode* pVnode; - SWal* pWal; - TDB* pMetaStore; - TTB* pExecStore; + char* path; + SHashObj* pushMgr; // consumerId -> STqHandle* + SHashObj* handles; // subKey -> STqHandle + SHashObj* pStreamTasks; // taksId -> SStreamTask + STqOffsetStore* pOffsetStore; + SVnode* pVnode; + SWal* pWal; + TDB* pMetaStore; + TTB* pExecStore; }; typedef struct { @@ -157,17 +157,19 @@ int32_t tqMetaClose(STQ* pTq); int32_t tqMetaSaveHandle(STQ* pTq, const char* key, const STqHandle* pHandle); int32_t tqMetaDeleteHandle(STQ* pTq, const char* key); +typedef struct { + int32_t size; +} STqOffsetHead; + +STqOffsetStore* tqOffsetOpen(); +void tqOffsetClose(STqOffsetStore*); +STqOffset* tqOffsetRead(STqOffsetStore* pStore, const char* subscribeKey); +int32_t tqOffsetWrite(STqOffsetStore* pStore, const STqOffset* pOffset); +int32_t tqOffsetSnapshot(STqOffsetStore* pStore); + // tqSink void tqTableSink(SStreamTask* pTask, void* vnode, int64_t ver, void* data); -// tqOffset -STqOffsetStore* tqOffsetOpen(STqOffsetCfg*); -void tqOffsetClose(STqOffsetStore*); -int64_t tqOffsetFetch(STqOffsetStore* pStore, const char* subscribeKey); -int32_t tqOffsetCommit(STqOffsetStore* pStore, const char* subscribeKey, int64_t offset); -int32_t tqOffsetPersist(STqOffsetStore* pStore, const char* subscribeKey); -int32_t tqOffsetPersistAll(STqOffsetStore* pStore); - #ifdef __cplusplus } #endif diff --git a/source/dnode/vnode/src/inc/vnodeInt.h b/source/dnode/vnode/src/inc/vnodeInt.h index 0c2b09a493..300a5f890e 100644 --- a/source/dnode/vnode/src/inc/vnodeInt.h +++ b/source/dnode/vnode/src/inc/vnodeInt.h @@ -137,6 +137,7 @@ int tqCommit(STQ*); int32_t tqUpdateTbUidList(STQ* pTq, const SArray* tbUidList, bool isAdd); int32_t tqProcessVgChangeReq(STQ* pTq, char* msg, int32_t msgLen); int32_t tqProcessVgDeleteReq(STQ* pTq, char* msg, int32_t msgLen); +int32_t tqProcessOffsetCommitReq(STQ* pTq, char* msg, int32_t msgLen); int32_t tqProcessPollReq(STQ* pTq, SRpcMsg* pMsg, int32_t workerId); int32_t tqProcessTaskDeploy(STQ* pTq, char* msg, int32_t msgLen); int32_t tqProcessStreamTrigger(STQ* pTq, SSubmitReq* data); diff --git a/source/dnode/vnode/src/meta/metaQuery.c b/source/dnode/vnode/src/meta/metaQuery.c index 1d8082084f..a6339125c4 100644 --- a/source/dnode/vnode/src/meta/metaQuery.c +++ b/source/dnode/vnode/src/meta/metaQuery.c @@ -663,12 +663,23 @@ int32_t metaFilteTableIds(SMeta *pMeta, SMetaFltParam *param, SArray *pUids) { void * entryKey = NULL, *entryVal = NULL; int32_t nEntryKey, nEntryVal; + bool first = true; while (1) { valid = tdbTbcGet(pCursor->pCur, (const void **)&entryKey, &nEntryKey, (const void **)&entryVal, &nEntryVal); if (valid < 0) { break; } STagIdxKey *p = entryKey; + if (p->type != pCursor->type) { + if (first) { + valid = param->reverse ? tdbTbcMoveToPrev(pCursor->pCur) : tdbTbcMoveToNext(pCursor->pCur); + if (valid < 0) break; + continue; + } else { + break; + } + } + first = false; if (p != NULL) { int32_t cmp = (*param->filterFunc)(p->data, pKey->data, pKey->type); if (cmp == 0) { diff --git a/source/dnode/vnode/src/tq/tq.c b/source/dnode/vnode/src/tq/tq.c index 4a5ea49d79..9f34ae39c0 100644 --- a/source/dnode/vnode/src/tq/tq.c +++ b/source/dnode/vnode/src/tq/tq.c @@ -47,7 +47,7 @@ void tqCleanUp() { } STQ* tqOpen(const char* path, SVnode* pVnode, SWal* pWal) { - STQ* pTq = taosMemoryMalloc(sizeof(STQ)); + STQ* pTq = taosMemoryCalloc(1, sizeof(STQ)); if (pTq == NULL) { terrno = TSDB_CODE_TQ_OUT_OF_MEMORY; return NULL; @@ -66,19 +66,23 @@ STQ* tqOpen(const char* path, SVnode* pVnode, SWal* pWal) { ASSERT(0); } + if (tqOffsetOpen(pTq) < 0) { + ASSERT(0); + } + return pTq; } void tqClose(STQ* pTq) { if (pTq) { - taosMemoryFreeClear(pTq->path); + tqOffsetClose(pTq->pOffsetStore); taosHashCleanup(pTq->handles); taosHashCleanup(pTq->pStreamTasks); taosHashCleanup(pTq->pushMgr); + taosMemoryFree(pTq->path); tqMetaClose(pTq); taosMemoryFree(pTq); } - // TODO } int32_t tqSendPollRsp(STQ* pTq, const SRpcMsg* pMsg, const SMqPollReq* pReq, const SMqDataBlkRsp* pRsp) { @@ -109,6 +113,33 @@ int32_t tqSendPollRsp(STQ* pTq, const SRpcMsg* pMsg, const SMqPollReq* pReq, con return 0; } +int32_t tqProcessOffsetCommitReq(STQ* pTq, char* msg, int32_t msgLen) { + STqOffset offset = {0}; + SDecoder decoder; + tDecoderInit(&decoder, msg, msgLen); + if (tDecodeSTqOffset(&decoder, &offset) < 0) { + ASSERT(0); + return -1; + } + tDecoderClear(&decoder); + + if (offset.type == TMQ_OFFSET__SNAPSHOT) { + tqDebug("receive offset commit msg to %s, offset(type:snapshot) uid: %ld, ts: %ld", offset.subKey, offset.uid, + offset.ts); + } else if (offset.type == TMQ_OFFSET__LOG) { + tqDebug("receive offset commit msg to %s, offset(type:log) version: %ld", offset.subKey, offset.version); + } else { + ASSERT(0); + } + + if (tqOffsetWrite(pTq->pOffsetStore, &offset) < 0) { + ASSERT(0); + return -1; + } + + return 0; +} + int32_t tqProcessPollReq(STQ* pTq, SRpcMsg* pMsg, int32_t workerId) { SMqPollReq* pReq = pMsg->pCont; int64_t consumerId = pReq->consumerId; diff --git a/source/dnode/vnode/src/tq/tqOffset.c b/source/dnode/vnode/src/tq/tqOffset.c index 4d83a67579..8d6cb28065 100644 --- a/source/dnode/vnode/src/tq/tqOffset.c +++ b/source/dnode/vnode/src/tq/tqOffset.c @@ -16,26 +16,128 @@ #include "tq.h" -enum ETqOffsetPersist { - TQ_OFFSET_PERSIST__LAZY = 1, - TQ_OFFSET_PERSIST__EAGER, -}; - -struct STqOffsetCfg { - int8_t persistPolicy; -}; - struct STqOffsetStore { - STqOffsetCfg cfg; - SHashObj* pHash; // SHashObj + char* fname; + STQ* pTq; + SHashObj* pHash; // SHashObj }; -STqOffsetStore* tqOffsetOpen(STqOffsetCfg* pCfg) { - STqOffsetStore* pStore = taosMemoryMalloc(sizeof(STqOffsetStore)); +static char* buildFileName(const char* path) { + int32_t len = strlen(path); + char* fname = taosMemoryCalloc(1, len + 20); + snprintf(fname, len + 20, "%s/offset", path); + return fname; +} + +STqOffsetStore* tqOffsetOpen(STQ* pTq) { + STqOffsetStore* pStore = taosMemoryCalloc(1, sizeof(STqOffsetStore)); if (pStore == NULL) { return NULL; } - memcpy(&pStore->cfg, pCfg, sizeof(STqOffsetCfg)); + pStore->pTq = pTq; + pTq->pOffsetStore = pStore; + pStore->pHash = taosHashInit(64, MurmurHash3_32, true, HASH_NO_LOCK); + if (pStore->pHash == NULL) { + if (pStore->pHash) taosHashCleanup(pStore->pHash); + return NULL; + } + char* fname = buildFileName(pStore->pTq->path); + TdFilePtr pFile = taosOpenFile(fname, TD_FILE_READ); + if (pFile != NULL) { + STqOffsetHead head = {0}; + int64_t code; + + while (1) { + if ((code = taosReadFile(pFile, &head, sizeof(STqOffsetHead))) != sizeof(STqOffsetHead)) { + if (code == 0) { + break; + } else { + ASSERT(0); + // TODO handle error + } + } + int32_t size = htonl(head.size); + void* memBuf = taosMemoryCalloc(1, size); + if ((code = taosReadFile(pFile, memBuf, size)) != size) { + ASSERT(0); + // TODO handle error + } + STqOffset offset; + SDecoder decoder; + tDecoderInit(&decoder, memBuf, size); + if (tDecodeSTqOffset(&decoder, &offset) < 0) { + ASSERT(0); + } + tDecoderClear(&decoder); + if (taosHashPut(pStore->pHash, offset.subKey, strlen(offset.subKey), &offset, sizeof(STqOffset)) < 0) { + ASSERT(0); + // TODO + } + } + + taosCloseFile(&pFile); + taosMemoryFree(fname); + } return pStore; } + +void tqOffsetClose(STqOffsetStore* pStore) { + tqOffsetSnapshot(pStore); + taosHashCleanup(pStore->pHash); +} + +STqOffset* tqOffsetRead(STqOffsetStore* pStore, const char* subscribeKey) { + return (STqOffset*)taosHashGet(pStore->pHash, subscribeKey, strlen(subscribeKey)); +} + +int32_t tqOffsetWrite(STqOffsetStore* pStore, const STqOffset* pOffset) { + return taosHashPut(pStore->pHash, pOffset->subKey, strlen(pOffset->subKey), pOffset, sizeof(STqOffset)); +} + +int32_t tqOffsetSnapshot(STqOffsetStore* pStore) { + // open file + // TODO file name should be with a version + char* fname = buildFileName(pStore->pTq->path); + TdFilePtr pFile = taosOpenFile(fname, TD_FILE_CREATE | TD_FILE_WRITE | TD_FILE_APPEND); + if (pFile == NULL) { + ASSERT(0); + return -1; + } + void* pIter = NULL; + while (1) { + pIter = taosHashIterate(pStore->pHash, pIter); + if (pIter == NULL) break; + STqOffset* pOffset = (STqOffset*)pIter; + int32_t bodyLen; + int32_t code; + tEncodeSize(tEncodeSTqOffset, pOffset, bodyLen, code); + ASSERT(code == 0); + if (code < 0) { + ASSERT(0); + taosHashCancelIterate(pStore->pHash, pIter); + return -1; + } + + int32_t totLen = sizeof(STqOffsetHead) + bodyLen; + void* buf = taosMemoryCalloc(1, totLen); + void* abuf = POINTER_SHIFT(buf, sizeof(STqOffsetHead)); + + ((STqOffsetHead*)buf)->size = htonl(bodyLen); + SEncoder encoder; + tEncoderInit(&encoder, abuf, bodyLen); + tEncodeSTqOffset(&encoder, pOffset); + // write file + int64_t writeLen; + if ((writeLen = taosWriteFile(pFile, buf, totLen)) != bodyLen) { + ASSERT(0); + tqError("write offset incomplete, len %d, write len %ld", bodyLen, writeLen); + taosHashCancelIterate(pStore->pHash, pIter); + return -1; + } + } + // close and rename file + taosCloseFile(&pFile); + taosMemoryFree(fname); + return 0; +} diff --git a/source/dnode/vnode/src/tq/tqSink.c b/source/dnode/vnode/src/tq/tqSink.c index 5c0bf971fb..0cca1d2e10 100644 --- a/source/dnode/vnode/src/tq/tqSink.c +++ b/source/dnode/vnode/src/tq/tqSink.c @@ -15,6 +15,177 @@ #include "tq.h" +static SSubmitReq* tdBlockToSubmit(const SArray* pBlocks, const STSchema* pSchema, bool createTb, int64_t suid, + const char* stbFullName, int32_t vgId); + +static SSubmitReq* tdBlockToSubmit(const SArray* pBlocks, const STSchema* pTSchema, bool createTb, int64_t suid, + const char* stbFullName, int32_t vgId) { + SSubmitReq* ret = NULL; + SArray* tagArray = taosArrayInit(1, sizeof(STagVal)); + if (!tagArray) { + terrno = TSDB_CODE_OUT_OF_MEMORY; + return NULL; + } + + // cal size + int32_t cap = sizeof(SSubmitReq); + int32_t sz = taosArrayGetSize(pBlocks); + for (int32_t i = 0; i < sz; i++) { + SSDataBlock* pDataBlock = taosArrayGet(pBlocks, i); + int32_t rows = pDataBlock->info.rows; + // TODO min + int32_t rowSize = pDataBlock->info.rowSize; + int32_t maxLen = TD_ROW_MAX_BYTES_FROM_SCHEMA(pTSchema); + int32_t schemaLen = 0; + + if (createTb) { + SVCreateTbReq createTbReq = {0}; + char* cname = buildCtbNameByGroupId(stbFullName, pDataBlock->info.groupId); + createTbReq.name = cname; + createTbReq.flags = 0; + createTbReq.type = TSDB_CHILD_TABLE; + createTbReq.ctb.suid = suid; + + STagVal tagVal = { + .cid = pDataBlock->info.numOfCols + 1, + .type = TSDB_DATA_TYPE_UBIGINT, + .i64 = (int64_t)pDataBlock->info.groupId, + }; + STag* pTag = NULL; + taosArrayClear(tagArray); + taosArrayPush(tagArray, &tagVal); + tTagNew(tagArray, 1, false, &pTag); + if (pTag == NULL) { + tdDestroySVCreateTbReq(&createTbReq); + taosArrayDestroy(tagArray); + return NULL; + } + createTbReq.ctb.pTag = (uint8_t*)pTag; + + int32_t code; + tEncodeSize(tEncodeSVCreateTbReq, &createTbReq, schemaLen, code); + + tdDestroySVCreateTbReq(&createTbReq); + if (code < 0) { + taosArrayDestroy(tagArray); + return NULL; + } + } + + cap += sizeof(SSubmitBlk) + schemaLen + rows * maxLen; + } + + // assign data + // TODO + ret = rpcMallocCont(cap); + ret->header.vgId = vgId; + ret->version = htonl(1); + ret->length = sizeof(SSubmitReq); + ret->numOfBlocks = htonl(sz); + + void* submitBlk = POINTER_SHIFT(ret, sizeof(SSubmitReq)); + for (int32_t i = 0; i < sz; i++) { + SSDataBlock* pDataBlock = taosArrayGet(pBlocks, i); + + SSubmitBlk* blkHead = submitBlk; + blkHead->numOfRows = htons(pDataBlock->info.rows); + blkHead->sversion = htonl(pTSchema->version); + // TODO + blkHead->suid = htobe64(suid); + // uid is assigned by vnode + blkHead->uid = 0; + + int32_t rows = pDataBlock->info.rows; + /*int32_t maxLen = TD_ROW_MAX_BYTES_FROM_SCHEMA(pTSchema);*/ + /*blkHead->dataLen = htonl(rows * maxLen);*/ + blkHead->dataLen = 0; + + void* blockData = POINTER_SHIFT(submitBlk, sizeof(SSubmitBlk)); + + int32_t schemaLen = 0; + if (createTb) { + SVCreateTbReq createTbReq = {0}; + char* cname = buildCtbNameByGroupId(stbFullName, pDataBlock->info.groupId); + createTbReq.name = cname; + createTbReq.flags = 0; + createTbReq.type = TSDB_CHILD_TABLE; + createTbReq.ctb.suid = suid; + + STagVal tagVal = { + .cid = pDataBlock->info.numOfCols + 1, + .type = TSDB_DATA_TYPE_UBIGINT, + .i64 = (int64_t)pDataBlock->info.groupId, + }; + taosArrayClear(tagArray); + taosArrayPush(tagArray, &tagVal); + STag* pTag = NULL; + tTagNew(tagArray, 1, false, &pTag); + if (pTag == NULL) { + tdDestroySVCreateTbReq(&createTbReq); + taosArrayDestroy(tagArray); + taosMemoryFreeClear(ret); + return NULL; + } + createTbReq.ctb.pTag = (uint8_t*)pTag; + + int32_t code; + tEncodeSize(tEncodeSVCreateTbReq, &createTbReq, schemaLen, code); + if (code < 0) { + tdDestroySVCreateTbReq(&createTbReq); + taosArrayDestroy(tagArray); + taosMemoryFreeClear(ret); + return NULL; + } + + SEncoder encoder = {0}; + tEncoderInit(&encoder, blockData, schemaLen); + code = tEncodeSVCreateTbReq(&encoder, &createTbReq); + tEncoderClear(&encoder); + tdDestroySVCreateTbReq(&createTbReq); + + if (code < 0) { + taosArrayDestroy(tagArray); + taosMemoryFreeClear(ret); + return NULL; + } + } + blkHead->schemaLen = htonl(schemaLen); + + STSRow* rowData = POINTER_SHIFT(blockData, schemaLen); + + for (int32_t j = 0; j < rows; j++) { + SRowBuilder rb = {0}; + tdSRowInit(&rb, pTSchema->version); + tdSRowSetTpInfo(&rb, pTSchema->numOfCols, pTSchema->flen); + tdSRowResetBuf(&rb, rowData); + + for (int32_t k = 0; k < pTSchema->numOfCols; k++) { + const STColumn* pColumn = &pTSchema->columns[k]; + SColumnInfoData* pColData = taosArrayGet(pDataBlock->pDataBlock, k); + if (colDataIsNull_s(pColData, j)) { + tdAppendColValToRow(&rb, pColumn->colId, pColumn->type, TD_VTYPE_NONE, NULL, false, pColumn->offset, k); + } else { + void* data = colDataGetData(pColData, j); + tdAppendColValToRow(&rb, pColumn->colId, pColumn->type, TD_VTYPE_NORM, data, true, pColumn->offset, k); + } + } + int32_t rowLen = TD_ROW_LEN(rowData); + rowData = POINTER_SHIFT(rowData, rowLen); + blkHead->dataLen += rowLen; + } + int32_t dataLen = blkHead->dataLen; + blkHead->dataLen = htonl(dataLen); + + ret->length += sizeof(SSubmitBlk) + schemaLen + dataLen; + blkHead = POINTER_SHIFT(blkHead, schemaLen + dataLen); + /*submitBlk = blkHead;*/ + } + + ret->length = htonl(ret->length); + taosArrayDestroy(tagArray); + return ret; +} + void tqTableSink(SStreamTask* pTask, void* vnode, int64_t ver, void* data) { const SArray* pRes = (const SArray*)data; SVnode* pVnode = (SVnode*)vnode; diff --git a/source/dnode/vnode/src/vnd/vnodeSvr.c b/source/dnode/vnode/src/vnd/vnodeSvr.c index ab2efa4791..10e0888988 100644 --- a/source/dnode/vnode/src/vnd/vnodeSvr.c +++ b/source/dnode/vnode/src/vnd/vnodeSvr.c @@ -148,17 +148,24 @@ int32_t vnodeProcessWriteReq(SVnode *pVnode, SRpcMsg *pMsg, int64_t version, SRp case TDMT_VND_MQ_VG_CHANGE: if (tqProcessVgChangeReq(pVnode->pTq, POINTER_SHIFT(pMsg->pCont, sizeof(SMsgHead)), pMsg->contLen - sizeof(SMsgHead)) < 0) { - // TODO: handle error + goto _err; } break; case TDMT_VND_MQ_VG_DELETE: if (tqProcessVgDeleteReq(pVnode->pTq, pMsg->pCont, pMsg->contLen) < 0) { - // TODO: handle error + goto _err; + } + break; + case TDMT_VND_MQ_COMMIT_OFFSET: + if (tqProcessOffsetCommitReq(pVnode->pTq, POINTER_SHIFT(pMsg->pCont, sizeof(SMsgHead)), + pMsg->contLen - sizeof(SMsgHead)) < 0) { + goto _err; } break; case TDMT_STREAM_TASK_DEPLOY: { if (tqProcessTaskDeploy(pVnode->pTq, POINTER_SHIFT(pMsg->pCont, sizeof(SMsgHead)), pMsg->contLen - sizeof(SMsgHead)) < 0) { + goto _err; } } break; case TDMT_VND_ALTER_CONFIRM: @@ -289,7 +296,7 @@ void vnodeUpdateMetaRsp(SVnode *pVnode, STableMetaRsp *pMetaRsp) { } int32_t vnodeProcessSyncReq(SVnode *pVnode, SRpcMsg *pMsg, SRpcMsg **pRsp) { - int32_t ret = TAOS_SYNC_PROPOSE_OTHER_ERROR; + int32_t ret = TAOS_SYNC_OTHER_ERROR; if (syncEnvIsStart()) { SSyncNode *pSyncNode = syncNodeAcquire(pVnode->sync); @@ -370,13 +377,13 @@ int32_t vnodeProcessSyncReq(SVnode *pVnode, SRpcMsg *pMsg, SRpcMsg **pRsp) { } else { vError("==vnodeProcessSyncReq== error msg type:%d", pRpcMsg->msgType); - ret = TAOS_SYNC_PROPOSE_OTHER_ERROR; + ret = TAOS_SYNC_OTHER_ERROR; } syncNodeRelease(pSyncNode); } else { vError("==vnodeProcessSyncReq== error syncEnv stop"); - ret = TAOS_SYNC_PROPOSE_OTHER_ERROR; + ret = TAOS_SYNC_OTHER_ERROR; } return ret; @@ -901,8 +908,8 @@ static int32_t vnodeProcessAlterConfirmReq(SVnode *pVnode, int64_t version, void static int32_t vnodeProcessAlterHasnRangeReq(SVnode *pVnode, int64_t version, void *pReq, int32_t len, SRpcMsg *pRsp) { vInfo("vgId:%d, alter hashrange msg will be processed", TD_VID(pVnode)); - // todo - // 1. stop work + // todo + // 1. stop work // 2. adjust hash range / compact / remove wals / rename vgroups // 3. reload sync return 0; diff --git a/source/dnode/vnode/src/vnd/vnodeSync.c b/source/dnode/vnode/src/vnd/vnodeSync.c index 087bdb2b49..be27824570 100644 --- a/source/dnode/vnode/src/vnd/vnodeSync.c +++ b/source/dnode/vnode/src/vnd/vnodeSync.c @@ -283,11 +283,14 @@ int32_t vnodeSyncOpen(SVnode *pVnode, char *path) { void vnodeSyncStart(SVnode *pVnode) { syncSetMsgCb(pVnode->sync, &pVnode->msgCb); + syncStart(pVnode->sync); + /* if (pVnode->config.standby) { syncStartStandBy(pVnode->sync); } else { syncStart(pVnode->sync); } + */ } void vnodeSyncClose(SVnode *pVnode) { syncStop(pVnode->sync); } diff --git a/source/libs/catalog/inc/catalogInt.h b/source/libs/catalog/inc/catalogInt.h index 1d56b6d5d4..df4b9b6ea6 100644 --- a/source/libs/catalog/inc/catalogInt.h +++ b/source/libs/catalog/inc/catalogInt.h @@ -57,6 +57,8 @@ enum { CTG_OP_DROP_TB_META, CTG_OP_UPDATE_USER, CTG_OP_UPDATE_VG_EPSET, + CTG_OP_UPDATE_TB_INDEX, + CTG_OP_DROP_TB_INDEX, CTG_OP_MAX }; @@ -128,25 +130,33 @@ typedef struct SCtgUserCtx { SUserAuthInfo user; } SCtgUserCtx; -typedef struct SCtgTbMetaCache { - SRWLatch stbLock; - SRWLatch metaLock; // RC between cache destroy and all other operations - SHashObj *metaCache; //key:tbname, value:STableMeta - SHashObj *stbCache; //key:suid, value:STableMeta* -} SCtgTbMetaCache; +typedef STableIndexRsp STableIndex; + +typedef struct SCtgTbCache { + SRWLatch metaLock; + STableMeta *pMeta; + SRWLatch indexLock; + STableIndex *pIndex; +} SCtgTbCache; + +typedef struct SCtgVgCache { + SRWLatch vgLock; + SDBVgInfo *vgInfo; +} SCtgVgCache; typedef struct SCtgDBCache { - SRWLatch vgLock; + SRWLatch dbLock; // RC between destroy tbCache/stbCache and all reads uint64_t dbId; int8_t deleted; - SDBVgInfo *vgInfo; - SCtgTbMetaCache tbCache; + SCtgVgCache vgCache; + SHashObj *tbCache; // key:tbname, value:SCtgTbCache + SHashObj *stbCache; // key:suid, value:STableMeta* } SCtgDBCache; typedef struct SCtgRentSlot { SRWLatch lock; bool needSort; - SArray *meta; // element is SDbVgVersion or SSTableMetaVersion + SArray *meta; // element is SDbVgVersion or SSTableVersion } SCtgRentSlot; typedef struct SCtgRentMgmt { @@ -245,8 +255,10 @@ typedef struct SCtgCacheStat { uint64_t userNum; uint64_t vgHitNum; uint64_t vgMissNum; - uint64_t tblHitNum; - uint64_t tblMissNum; + uint64_t tbMetaHitNum; + uint64_t tbMetaMissNum; + uint64_t tbIndexHitNum; + uint64_t tbIndexMissNum; uint64_t userHitNum; uint64_t userMissNum; } SCtgCacheStat; @@ -268,10 +280,10 @@ typedef struct SCtgUpdateVgMsg { SDBVgInfo* dbInfo; } SCtgUpdateVgMsg; -typedef struct SCtgUpdateTblMsg { - SCatalog* pCtg; - STableMetaOutput* output; -} SCtgUpdateTblMsg; +typedef struct SCtgUpdateTbMetaMsg { + SCatalog* pCtg; + STableMetaOutput* pMeta; +} SCtgUpdateTbMetaMsg; typedef struct SCtgDropDBMsg { SCatalog* pCtg; @@ -305,6 +317,17 @@ typedef struct SCtgUpdateUserMsg { SGetUserAuthRsp userAuth; } SCtgUpdateUserMsg; +typedef struct SCtgUpdateTbIndexMsg { + SCatalog* pCtg; + STableIndex* pIndex; +} SCtgUpdateTbIndexMsg; + +typedef struct SCtgDropTbIndexMsg { + SCatalog* pCtg; + char dbFName[TSDB_DB_FNAME_LEN]; + char tbName[TSDB_TABLE_NAME_LEN]; +} SCtgDropTbIndexMsg; + typedef struct SCtgUpdateEpsetMsg { SCatalog* pCtg; char dbFName[TSDB_DB_FNAME_LEN]; @@ -465,12 +488,11 @@ int32_t ctgOpUpdateUser(SCtgCacheOperation *action); int32_t ctgOpUpdateEpset(SCtgCacheOperation *operation); int32_t ctgAcquireVgInfoFromCache(SCatalog* pCtg, const char *dbFName, SCtgDBCache **pCache); void ctgReleaseDBCache(SCatalog *pCtg, SCtgDBCache *dbCache); -void ctgReleaseVgInfo(SCtgDBCache *dbCache); -int32_t ctgAcquireVgInfoFromCache(SCatalog* pCtg, const char *dbFName, SCtgDBCache **pCache); +void ctgRUnlockVgInfo(SCtgDBCache *dbCache); int32_t ctgTbMetaExistInCache(SCatalog* pCtg, char *dbFName, char* tbName, int32_t *exist); int32_t ctgReadTbMetaFromCache(SCatalog* pCtg, SCtgTbMetaCtx* ctx, STableMeta** pTableMeta); -int32_t ctgReadTbVerFromCache(SCatalog *pCtg, const SName *pTableName, int32_t *sver, int32_t *tver, int32_t *tbType, uint64_t *suid, char *stbName); -int32_t ctgChkAuthFromCache(SCatalog* pCtg, const char* user, const char* dbFName, AUTH_TYPE type, bool *inCache, bool *pass); +int32_t ctgReadTbVerFromCache(SCatalog *pCtg, SName *pTableName, int32_t *sver, int32_t *tver, int32_t *tbType, uint64_t *suid, char *stbName); +int32_t ctgChkAuthFromCache(SCatalog* pCtg, char* user, char* dbFName, AUTH_TYPE type, bool *inCache, bool *pass); int32_t ctgDropDbCacheEnqueue(SCatalog* pCtg, const char *dbFName, int64_t dbId); int32_t ctgDropDbVgroupEnqueue(SCatalog* pCtg, const char *dbFName, bool syncReq); int32_t ctgDropStbMetaEnqueue(SCatalog* pCtg, const char *dbFName, int64_t dbId, const char *stbName, uint64_t suid, bool syncReq); @@ -479,12 +501,18 @@ int32_t ctgUpdateVgroupEnqueue(SCatalog* pCtg, const char *dbFName, int64_t dbId int32_t ctgUpdateTbMetaEnqueue(SCatalog* pCtg, STableMetaOutput *output, bool syncReq); int32_t ctgUpdateUserEnqueue(SCatalog* pCtg, SGetUserAuthRsp *pAuth, bool syncReq); int32_t ctgUpdateVgEpsetEnqueue(SCatalog* pCtg, char *dbFName, int32_t vgId, SEpSet* pEpSet); +int32_t ctgUpdateTbIndexEnqueue(SCatalog* pCtg, STableIndex **pIndex, bool syncOp); int32_t ctgMetaRentInit(SCtgRentMgmt *mgmt, uint32_t rentSec, int8_t type); int32_t ctgMetaRentAdd(SCtgRentMgmt *mgmt, void *meta, int64_t id, int32_t size); int32_t ctgMetaRentGet(SCtgRentMgmt *mgmt, void **res, uint32_t *num, int32_t size); int32_t ctgUpdateTbMetaToCache(SCatalog* pCtg, STableMetaOutput* pOut, bool syncReq); int32_t ctgStartUpdateThread(); int32_t ctgRelaunchGetTbMetaTask(SCtgTask *pTask); +void ctgReleaseVgInfoToCache(SCatalog* pCtg, SCtgDBCache *dbCache); +int32_t ctgReadTbIndexFromCache(SCatalog* pCtg, SName* pTableName, SArray** pRes); +int32_t ctgDropTbIndexEnqueue(SCatalog* pCtg, SName* pName, bool syncOp); +int32_t ctgOpDropTbIndex(SCtgCacheOperation *operation); +int32_t ctgOpUpdateTbIndex(SCtgCacheOperation *operation); @@ -493,7 +521,7 @@ int32_t ctgGetDBVgInfoFromMnode(SCatalog* pCtg, SRequestConnInfo *pConn, SBuildU int32_t ctgGetQnodeListFromMnode(SCatalog* pCtg, SRequestConnInfo *pConn, SArray *out, SCtgTask* pTask); int32_t ctgGetDBCfgFromMnode(SCatalog* pCtg, SRequestConnInfo *pConn, const char *dbFName, SDbCfgInfo *out, SCtgTask* pTask); int32_t ctgGetIndexInfoFromMnode(SCatalog* pCtg, SRequestConnInfo *pConn, const char *indexName, SIndexInfo *out, SCtgTask* pTask); -int32_t ctgGetTbIndexFromMnode(SCatalog* pCtg, SRequestConnInfo *pConn, SName* name, SArray** out, SCtgTask* pTask); +int32_t ctgGetTbIndexFromMnode(SCatalog* pCtg, SRequestConnInfo *pConn, SName *name, STableIndex* out, SCtgTask* pTask); int32_t ctgGetUdfInfoFromMnode(SCatalog* pCtg, SRequestConnInfo *pConn, const char *funcName, SFuncInfo *out, SCtgTask* pTask); int32_t ctgGetUserDbAuthFromMnode(SCatalog* pCtg, SRequestConnInfo *pConn, const char *user, SGetUserAuthRsp *out, SCtgTask* pTask); int32_t ctgGetTbMetaFromMnodeImpl(SCatalog* pCtg, SRequestConnInfo *pConn, char *dbFName, char* tbName, STableMetaOutput* out, SCtgTask* pTask); @@ -521,6 +549,8 @@ void ctgFreeSTableMetaOutput(STableMetaOutput* pOutput); int32_t ctgUpdateMsgCtx(SCtgMsgCtx* pCtx, int32_t reqType, void* out, char* target); char *ctgTaskTypeStr(CTG_TASK_TYPE type); int32_t ctgUpdateSendTargetInfo(SMsgSendInfo *pMsgSendInfo, int32_t msgType, SCtgTask* pTask); +int32_t ctgCloneTableIndex(SArray* pIndex, SArray** pRes); +void ctgFreeSTableIndex(void *info); extern SCatalogMgmt gCtgMgmt; diff --git a/source/libs/catalog/src/catalog.c b/source/libs/catalog/src/catalog.c index bc94de89ea..a3576f8738 100644 --- a/source/libs/catalog/src/catalog.c +++ b/source/libs/catalog/src/catalog.c @@ -96,8 +96,7 @@ int32_t ctgRefreshDBVgInfo(SCatalog* pCtg, SRequestConnInfo *pConn, const char* if (NULL != dbCache) { input.dbId = dbCache->dbId; - ctgReleaseVgInfo(dbCache); - ctgReleaseDBCache(pCtg, dbCache); + ctgReleaseVgInfoToCache(pCtg, dbCache); } input.vgVersion = CTG_DEFAULT_INVALID_VERSION; @@ -349,8 +348,8 @@ int32_t ctgChkAuth(SCatalog* pCtg, SRequestConnInfo *pConn, const char* user, co int32_t code = 0; *pass = false; - - CTG_ERR_RET(ctgChkAuthFromCache(pCtg, user, dbFName, type, &inCache, pass)); + + CTG_ERR_RET(ctgChkAuthFromCache(pCtg, (char*)user, (char*)dbFName, type, &inCache, pass)); if (inCache) { return TSDB_CODE_SUCCESS; @@ -382,6 +381,45 @@ _return: return TSDB_CODE_SUCCESS; } +int32_t ctgGetTbIndex(SCatalog* pCtg, SRequestConnInfo *pConn, SName* pTableName, SArray** pRes) { + CTG_ERR_RET(ctgReadTbIndexFromCache(pCtg, pTableName, pRes)); + if (*pRes) { + return TSDB_CODE_SUCCESS; + } + + STableIndex *pIndex = taosMemoryCalloc(1, sizeof(STableIndex)); + if (NULL == pIndex) { + CTG_ERR_RET(TSDB_CODE_OUT_OF_MEMORY); + } + + int32_t code = ctgGetTbIndexFromMnode(pCtg, pConn, (SName*)pTableName, pIndex, NULL); + if (TSDB_CODE_MND_DB_INDEX_NOT_EXIST == code) { + code = 0; + goto _return; + } + CTG_ERR_JRET(code); + + SArray* pInfo = NULL; + CTG_ERR_JRET(ctgCloneTableIndex(pIndex->pIndex, &pInfo)); + + *pRes = pInfo; + + CTG_ERR_JRET(ctgUpdateTbIndexEnqueue(pCtg, &pIndex, false)); + + return TSDB_CODE_SUCCESS; + +_return: + + tFreeSTableIndexRsp(pIndex); + taosMemoryFree(pIndex); + + taosArrayDestroyEx(*pRes, tFreeSTableIndexInfo); + *pRes = NULL; + + CTG_RET(code); +} + + int32_t ctgGetTbDistVgInfo(SCatalog* pCtg, SRequestConnInfo *pConn, SName* pTableName, SArray** pVgList) { STableMeta *tbMeta = NULL; int32_t code = 0; @@ -404,7 +442,7 @@ int32_t ctgGetTbDistVgInfo(SCatalog* pCtg, SRequestConnInfo *pConn, SName* pTabl CTG_ERR_JRET(ctgGetDBVgInfo(pCtg, pConn, db, &dbCache, &vgInfo)); if (dbCache) { - vgHash = dbCache->vgInfo->vgHash; + vgHash = dbCache->vgCache.vgInfo->vgHash; } else { vgHash = vgInfo->vgHash; } @@ -442,7 +480,7 @@ int32_t ctgGetTbDistVgInfo(SCatalog* pCtg, SRequestConnInfo *pConn, SName* pTabl _return: if (dbCache) { - ctgReleaseVgInfo(dbCache); + ctgRUnlockVgInfo(dbCache); ctgReleaseDBCache(pCtg, dbCache); } @@ -631,12 +669,11 @@ int32_t catalogGetDBVgVersion(SCatalog* pCtg, const char* dbFName, int32_t* vers CTG_API_LEAVE(TSDB_CODE_SUCCESS); } - *version = dbCache->vgInfo->vgVersion; + *version = dbCache->vgCache.vgInfo->vgVersion; *dbId = dbCache->dbId; - *tableNum = dbCache->vgInfo->numOfTable; + *tableNum = dbCache->vgCache.vgInfo->numOfTable; - ctgReleaseVgInfo(dbCache); - ctgReleaseDBCache(pCtg, dbCache); + ctgReleaseVgInfoToCache(pCtg, dbCache); ctgDebug("Got db vgVersion from cache, dbFName:%s, vgVersion:%d", dbFName, *version); @@ -661,7 +698,7 @@ int32_t catalogGetDBVgInfo(SCatalog* pCtg, SRequestConnInfo *pConn, const char* SDBVgInfo *vgInfo = NULL; CTG_ERR_JRET(ctgGetDBVgInfo(pCtg, pConn, dbFName, &dbCache, &vgInfo)); if (dbCache) { - vgHash = dbCache->vgInfo->vgHash; + vgHash = dbCache->vgCache.vgInfo->vgHash; } else { vgHash = vgInfo->vgHash; } @@ -674,7 +711,7 @@ int32_t catalogGetDBVgInfo(SCatalog* pCtg, SRequestConnInfo *pConn, const char* _return: if (dbCache) { - ctgReleaseVgInfo(dbCache); + ctgRUnlockVgInfo(dbCache); ctgReleaseDBCache(pCtg, dbCache); } @@ -741,6 +778,30 @@ _return: CTG_API_LEAVE(code); } +int32_t catalogUpdateTableIndex(SCatalog* pCtg, STableIndexRsp *pRsp) { + CTG_API_ENTER(); + + int32_t code = 0; + + if (NULL == pCtg || NULL == pRsp) { + CTG_API_LEAVE(TSDB_CODE_CTG_INVALID_INPUT); + } + + STableIndex* pIndex = taosMemoryCalloc(1, sizeof(STableIndex)); + if (NULL == pIndex) { + CTG_API_LEAVE(TSDB_CODE_OUT_OF_MEMORY); + } + + memcpy(pIndex, pRsp, sizeof(STableIndex)); + + CTG_ERR_JRET(ctgUpdateTbIndexEnqueue(pCtg, &pIndex, false)); + +_return: + + CTG_API_LEAVE(code); +} + + int32_t catalogRemoveTableMeta(SCatalog* pCtg, SName* pTableName) { CTG_API_ENTER(); @@ -932,12 +993,12 @@ int32_t catalogGetTableHashVgroup(SCatalog *pCtg, SRequestConnInfo *pConn, const SDBVgInfo *vgInfo = NULL; CTG_ERR_JRET(ctgGetDBVgInfo(pCtg, pConn, db, &dbCache, &vgInfo)); - CTG_ERR_JRET(ctgGetVgInfoFromHashValue(pCtg, vgInfo ? vgInfo : dbCache->vgInfo, pTableName, pVgroup)); + CTG_ERR_JRET(ctgGetVgInfoFromHashValue(pCtg, vgInfo ? vgInfo : dbCache->vgCache.vgInfo, pTableName, pVgroup)); _return: if (dbCache) { - ctgReleaseVgInfo(dbCache); + ctgRUnlockVgInfo(dbCache); ctgReleaseDBCache(pCtg, dbCache); } @@ -1060,14 +1121,14 @@ _return: CTG_API_LEAVE(TSDB_CODE_SUCCESS); } -int32_t catalogGetExpiredSTables(SCatalog* pCtg, SSTableMetaVersion** stables, uint32_t* num) { +int32_t catalogGetExpiredSTables(SCatalog* pCtg, SSTableVersion **stables, uint32_t *num) { CTG_API_ENTER(); if (NULL == pCtg || NULL == stables || NULL == num) { CTG_API_LEAVE(TSDB_CODE_CTG_INVALID_INPUT); } - CTG_API_LEAVE(ctgMetaRentGet(&pCtg->stbRent, (void**)stables, num, sizeof(SSTableMetaVersion))); + CTG_API_LEAVE(ctgMetaRentGet(&pCtg->stbRent, (void **)stables, num, sizeof(SSTableVersion))); } int32_t catalogGetExpiredDBs(SCatalog* pCtg, SDbVgVersion** dbs, uint32_t* num) { @@ -1138,7 +1199,12 @@ int32_t catalogGetTableIndex(SCatalog* pCtg, SRequestConnInfo *pConn, const SNam CTG_API_LEAVE(TSDB_CODE_CTG_INVALID_INPUT); } - CTG_API_LEAVE(ctgGetTbIndexFromMnode(pCtg, pConn, (SName*)pTableName, pRes, NULL)); + int32_t code = 0; + CTG_ERR_JRET(ctgGetTbIndex(pCtg, pConn, (SName*)pTableName, pRes)); + +_return: + + CTG_API_LEAVE(code); } int32_t catalogGetUdfInfo(SCatalog* pCtg, SRequestConnInfo *pConn, const char* funcName, SFuncInfo* pInfo) { diff --git a/source/libs/catalog/src/ctgAsync.c b/source/libs/catalog/src/ctgAsync.c index aff04fff06..159c120dde 100644 --- a/source/libs/catalog/src/ctgAsync.c +++ b/source/libs/catalog/src/ctgAsync.c @@ -344,6 +344,11 @@ int32_t ctgHandleForceUpdate(SCatalog* pCtg, SCtgJob *pJob, const SCatalogReq* p } } + for (int32_t i = 0; i < pJob->tbIndexNum; ++i) { + SName* name = taosArrayGet(pReq->pTableIndex, i); + ctgDropTbIndexEnqueue(pCtg, name, true); + } + return TSDB_CODE_SUCCESS; } @@ -680,15 +685,14 @@ int32_t ctgHandleGetTbMetaRsp(SCtgTask* pTask, int32_t reqType, const SDataBuf * CTG_ERR_RET(ctgAcquireVgInfoFromCache(pCtg, dbFName, &dbCache)); if (NULL != dbCache) { SVgroupInfo vgInfo = {0}; - CTG_ERR_JRET(ctgGetVgInfoFromHashValue(pCtg, dbCache->vgInfo, ctx->pName, &vgInfo)); + CTG_ERR_JRET(ctgGetVgInfoFromHashValue(pCtg, dbCache->vgCache.vgInfo, ctx->pName, &vgInfo)); ctgDebug("will refresh tbmeta, not supposed to be stb, tbName:%s, flag:%d", tNameGetTableName(ctx->pName), ctx->flag); ctx->vgId = vgInfo.vgId; CTG_ERR_JRET(ctgGetTbMetaFromVnode(pCtg, pConn, ctx->pName, &vgInfo, NULL, pTask)); - ctgReleaseVgInfo(dbCache); - ctgReleaseDBCache(pCtg, dbCache); + ctgReleaseVgInfoToCache(pCtg, dbCache); } else { SBuildUseDBInput input = {0}; @@ -786,7 +790,7 @@ int32_t ctgHandleGetTbMetaRsp(SCtgTask* pTask, int32_t reqType, const SDataBuf * _return: if (dbCache) { - ctgReleaseVgInfo(dbCache); + ctgRUnlockVgInfo(dbCache); ctgReleaseDBCache(pCtg, dbCache); } @@ -866,9 +870,15 @@ _return: int32_t ctgHandleGetTbIndexRsp(SCtgTask* pTask, int32_t reqType, const SDataBuf *pMsg, int32_t rspCode) { int32_t code = 0; - CTG_ERR_JRET(ctgProcessRspMsg(pTask->msgCtx.out, reqType, pMsg->pData, pMsg->len, rspCode, pTask->msgCtx.target)); + CTG_ERR_JRET(ctgProcessRspMsg(&pTask->msgCtx.out, reqType, pMsg->pData, pMsg->len, rspCode, pTask->msgCtx.target)); - TSWAP(pTask->res, pTask->msgCtx.out); + STableIndex* pOut = (STableIndex*)pTask->msgCtx.out; + SArray* pInfo = NULL; + CTG_ERR_JRET(ctgCloneTableIndex(pOut->pIndex, &pInfo)); + pTask->res = pInfo; + + SCtgTbIndexCtx* ctx = pTask->taskCtx; + CTG_ERR_JRET(ctgUpdateTbIndexEnqueue(pTask->pJob->pCtg, (STableIndex**)&pTask->msgCtx.out, false)); _return: if (TSDB_CODE_MND_DB_INDEX_NOT_EXIST == code) { @@ -1008,7 +1018,7 @@ int32_t ctgAsyncRefreshTbMeta(SCtgTask *pTask) { CTG_ERR_RET(ctgAcquireVgInfoFromCache(pCtg, dbFName, &dbCache)); if (dbCache) { SVgroupInfo vgInfo = {0}; - CTG_ERR_RET(ctgGetVgInfoFromHashValue(pCtg, dbCache->vgInfo, ctx->pName, &vgInfo)); + CTG_ERR_RET(ctgGetVgInfoFromHashValue(pCtg, dbCache->vgCache.vgInfo, ctx->pName, &vgInfo)); ctgDebug("will refresh tbmeta, not supposed to be stb, tbName:%s, flag:%d", tNameGetTableName(ctx->pName), ctx->flag); @@ -1026,8 +1036,7 @@ int32_t ctgAsyncRefreshTbMeta(SCtgTask *pTask) { _return: if (dbCache) { - ctgReleaseVgInfo(dbCache); - ctgReleaseDBCache(pCtg, dbCache); + ctgReleaseVgInfoToCache(pCtg, dbCache); } CTG_RET(code); @@ -1057,7 +1066,7 @@ int32_t ctgLaunchGetDbVgTask(SCtgTask *pTask) { CTG_ERR_RET(ctgAcquireVgInfoFromCache(pCtg, pCtx->dbFName, &dbCache)); if (NULL != dbCache) { - CTG_ERR_JRET(ctgGenerateVgList(pCtg, dbCache->vgInfo->vgHash, (SArray**)&pTask->res)); + CTG_ERR_JRET(ctgGenerateVgList(pCtg, dbCache->vgCache.vgInfo->vgHash, (SArray**)&pTask->res)); CTG_ERR_JRET(ctgHandleTaskEnd(pTask, 0)); } else { @@ -1072,8 +1081,7 @@ int32_t ctgLaunchGetDbVgTask(SCtgTask *pTask) { _return: if (dbCache) { - ctgReleaseVgInfo(dbCache); - ctgReleaseDBCache(pCtg, dbCache); + ctgReleaseVgInfoToCache(pCtg, dbCache); } CTG_RET(code); @@ -1092,7 +1100,7 @@ int32_t ctgLaunchGetTbHashTask(SCtgTask *pTask) { if (NULL == pTask->res) { CTG_ERR_JRET(TSDB_CODE_OUT_OF_MEMORY); } - CTG_ERR_JRET(ctgGetVgInfoFromHashValue(pCtg, dbCache->vgInfo, pCtx->pName, (SVgroupInfo*)pTask->res)); + CTG_ERR_JRET(ctgGetVgInfoFromHashValue(pCtg, dbCache->vgCache.vgInfo, pCtx->pName, (SVgroupInfo*)pTask->res)); CTG_ERR_JRET(ctgHandleTaskEnd(pTask, 0)); } else { @@ -1107,8 +1115,7 @@ int32_t ctgLaunchGetTbHashTask(SCtgTask *pTask) { _return: if (dbCache) { - ctgReleaseVgInfo(dbCache); - ctgReleaseDBCache(pCtg, dbCache); + ctgReleaseVgInfoToCache(pCtg, dbCache); } CTG_RET(code); @@ -1119,6 +1126,15 @@ int32_t ctgLaunchGetTbIndexTask(SCtgTask *pTask) { SCatalog* pCtg = pTask->pJob->pCtg; SRequestConnInfo* pConn = &pTask->pJob->conn; SCtgTbIndexCtx* pCtx = (SCtgTbIndexCtx*)pTask->taskCtx; + SArray* pRes = NULL; + + CTG_ERR_RET(ctgReadTbIndexFromCache(pCtg, pCtx->pName, &pRes)); + if (pRes) { + pTask->res = pRes; + + CTG_ERR_RET(ctgHandleTaskEnd(pTask, 0)); + return TSDB_CODE_SUCCESS; + } CTG_ERR_RET(ctgGetTbIndexFromMnode(pCtg, pConn, pCtx->pName, NULL, pTask)); return TSDB_CODE_SUCCESS; @@ -1157,9 +1173,9 @@ int32_t ctgLaunchGetDbInfoTask(SCtgTask *pTask) { SDbInfo* pInfo = (SDbInfo*)pTask->res; CTG_ERR_RET(ctgAcquireVgInfoFromCache(pCtg, pCtx->dbFName, &dbCache)); if (NULL != dbCache) { - pInfo->vgVer = dbCache->vgInfo->vgVersion; + pInfo->vgVer = dbCache->vgCache.vgInfo->vgVersion; pInfo->dbId = dbCache->dbId; - pInfo->tbNum = dbCache->vgInfo->numOfTable; + pInfo->tbNum = dbCache->vgCache.vgInfo->numOfTable; } else { pInfo->vgVer = CTG_DEFAULT_INVALID_VERSION; } @@ -1169,8 +1185,7 @@ int32_t ctgLaunchGetDbInfoTask(SCtgTask *pTask) { _return: if (dbCache) { - ctgReleaseVgInfo(dbCache); - ctgReleaseDBCache(pCtg, dbCache); + ctgReleaseVgInfoToCache(pCtg, dbCache); } CTG_RET(code); diff --git a/source/libs/catalog/src/ctgCache.c b/source/libs/catalog/src/ctgCache.c index add44d6a16..62890b8326 100644 --- a/source/libs/catalog/src/ctgCache.c +++ b/source/libs/catalog/src/ctgCache.c @@ -59,18 +59,27 @@ SCtgOperation gCtgCacheOperation[CTG_OP_MAX] = { CTG_OP_UPDATE_VG_EPSET, "update epset", ctgOpUpdateEpset - } - + }, + { + CTG_OP_UPDATE_TB_INDEX, + "update tbIndex", + ctgOpUpdateTbIndex + }, + { + CTG_OP_DROP_TB_INDEX, + "drop tbIndex", + ctgOpDropTbIndex + } }; -int32_t ctgAcquireVgInfo(SCatalog *pCtg, SCtgDBCache *dbCache, bool *inCache) { - CTG_LOCK(CTG_READ, &dbCache->vgLock); +int32_t ctgRLockVgInfo(SCatalog *pCtg, SCtgDBCache *dbCache, bool *inCache) { + CTG_LOCK(CTG_READ, &dbCache->vgCache.vgLock); if (dbCache->deleted) { - CTG_UNLOCK(CTG_READ, &dbCache->vgLock); + CTG_UNLOCK(CTG_READ, &dbCache->vgCache.vgLock); ctgDebug("db is dropping, dbId:%"PRIx64, dbCache->dbId); @@ -79,8 +88,8 @@ int32_t ctgAcquireVgInfo(SCatalog *pCtg, SCtgDBCache *dbCache, bool *inCache) { } - if (NULL == dbCache->vgInfo) { - CTG_UNLOCK(CTG_READ, &dbCache->vgLock); + if (NULL == dbCache->vgCache.vgInfo) { + CTG_UNLOCK(CTG_READ, &dbCache->vgCache.vgLock); *inCache = false; ctgDebug("db vgInfo is empty, dbId:%"PRIx64, dbCache->dbId); @@ -92,50 +101,47 @@ int32_t ctgAcquireVgInfo(SCatalog *pCtg, SCtgDBCache *dbCache, bool *inCache) { return TSDB_CODE_SUCCESS; } -int32_t ctgWAcquireVgInfo(SCatalog *pCtg, SCtgDBCache *dbCache) { - CTG_LOCK(CTG_WRITE, &dbCache->vgLock); +int32_t ctgWLockVgInfo(SCatalog *pCtg, SCtgDBCache *dbCache) { + CTG_LOCK(CTG_WRITE, &dbCache->vgCache.vgLock); if (dbCache->deleted) { ctgDebug("db is dropping, dbId:%"PRIx64, dbCache->dbId); - CTG_UNLOCK(CTG_WRITE, &dbCache->vgLock); + CTG_UNLOCK(CTG_WRITE, &dbCache->vgCache.vgLock); CTG_ERR_RET(TSDB_CODE_CTG_DB_DROPPED); } return TSDB_CODE_SUCCESS; } +void ctgRUnlockVgInfo(SCtgDBCache *dbCache) { + CTG_UNLOCK(CTG_READ, &dbCache->vgCache.vgLock); +} + +void ctgWUnlockVgInfo(SCtgDBCache *dbCache) { + CTG_UNLOCK(CTG_WRITE, &dbCache->vgCache.vgLock); +} + void ctgReleaseDBCache(SCatalog *pCtg, SCtgDBCache *dbCache) { - taosHashRelease(pCtg->dbCache, dbCache); + CTG_UNLOCK(CTG_READ, &dbCache->dbLock); } -void ctgReleaseVgInfo(SCtgDBCache *dbCache) { - CTG_UNLOCK(CTG_READ, &dbCache->vgLock); -} - -void ctgWReleaseVgInfo(SCtgDBCache *dbCache) { - CTG_UNLOCK(CTG_WRITE, &dbCache->vgLock); -} - - int32_t ctgAcquireDBCacheImpl(SCatalog* pCtg, const char *dbFName, SCtgDBCache **pCache, bool acquire) { char *p = strchr(dbFName, '.'); if (p && CTG_IS_SYS_DBNAME(p + 1)) { dbFName = p + 1; } - SCtgDBCache *dbCache = NULL; - if (acquire) { - dbCache = (SCtgDBCache *)taosHashAcquire(pCtg->dbCache, dbFName, strlen(dbFName)); - } else { - dbCache = (SCtgDBCache *)taosHashGet(pCtg->dbCache, dbFName, strlen(dbFName)); - } - + SCtgDBCache *dbCache = (SCtgDBCache *)taosHashGet(pCtg->dbCache, dbFName, strlen(dbFName)); if (NULL == dbCache) { *pCache = NULL; ctgDebug("db not in cache, dbFName:%s", dbFName); return TSDB_CODE_SUCCESS; } + if (acquire) { + CTG_LOCK(CTG_READ, &dbCache->dbLock); + } + if (dbCache->deleted) { if (acquire) { ctgReleaseDBCache(pCtg, dbCache); @@ -159,15 +165,35 @@ int32_t ctgGetDBCache(SCatalog* pCtg, const char *dbFName, SCtgDBCache **pCache) CTG_RET(ctgAcquireDBCacheImpl(pCtg, dbFName, pCache, false)); } +void ctgReleaseVgInfoToCache(SCatalog* pCtg, SCtgDBCache *dbCache) { + ctgRUnlockVgInfo(dbCache); + ctgReleaseDBCache(pCtg, dbCache); +} + +void ctgReleaseTbMetaToCache(SCatalog* pCtg, SCtgDBCache *dbCache, SCtgTbCache* pCache) { + if (pCache) { + CTG_UNLOCK(CTG_READ, &pCache->metaLock); + taosHashRelease(dbCache->tbCache, pCache); + } + + if (dbCache) { + ctgReleaseDBCache(pCtg, dbCache); + } +} + +void ctgReleaseTbIndexToCache(SCatalog* pCtg, SCtgDBCache *dbCache, SCtgTbCache* pCache) { + if (pCache) { + CTG_UNLOCK(CTG_READ, &pCache->indexLock); + taosHashRelease(dbCache->tbCache, pCache); + } + + if (dbCache) { + ctgReleaseDBCache(pCtg, dbCache); + } +} int32_t ctgAcquireVgInfoFromCache(SCatalog* pCtg, const char *dbFName, SCtgDBCache **pCache) { SCtgDBCache *dbCache = NULL; - - if (NULL == pCtg->dbCache) { - ctgDebug("empty db cache, dbFName:%s", dbFName); - goto _return; - } - ctgAcquireDBCache(pCtg, dbFName, &dbCache); if (NULL == dbCache) { ctgDebug("db %s not in cache", dbFName); @@ -175,7 +201,7 @@ int32_t ctgAcquireVgInfoFromCache(SCatalog* pCtg, const char *dbFName, SCtgDBCac } bool inCache = false; - ctgAcquireVgInfo(pCtg, dbCache, &inCache); + ctgRLockVgInfo(pCtg, dbCache, &inCache); if (!inCache) { ctgDebug("vgInfo of db %s not in cache", dbFName); goto _return; @@ -202,54 +228,160 @@ _return: return TSDB_CODE_SUCCESS; } -int32_t ctgTbMetaExistInCache(SCatalog* pCtg, char *dbFName, char* tbName, int32_t *exist) { - if (NULL == pCtg->dbCache) { - *exist = 0; - ctgWarn("empty db cache, dbFName:%s, tbName:%s", dbFName, tbName); - return TSDB_CODE_SUCCESS; - } - +int32_t ctgAcquireTbMetaFromCache(SCatalog* pCtg, char *dbFName, char* tbName, SCtgDBCache **pDb, SCtgTbCache** pTb) { SCtgDBCache *dbCache = NULL; + SCtgTbCache* pCache = NULL; ctgAcquireDBCache(pCtg, dbFName, &dbCache); if (NULL == dbCache) { - *exist = 0; - return TSDB_CODE_SUCCESS; + ctgDebug("db %s not in cache", dbFName); + goto _return; + } + + int32_t sz = 0; + pCache = taosHashAcquire(dbCache->tbCache, tbName, strlen(tbName)); + if (NULL == pCache) { + ctgDebug("tb %s not in cache, dbFName:%s", tbName, dbFName); + goto _return; } - size_t sz = 0; - CTG_LOCK(CTG_READ, &dbCache->tbCache.metaLock); - STableMeta *tbMeta = taosHashGet(dbCache->tbCache.metaCache, tbName, strlen(tbName)); - CTG_UNLOCK(CTG_READ, &dbCache->tbCache.metaLock); - - if (NULL == tbMeta) { - ctgReleaseDBCache(pCtg, dbCache); - - *exist = 0; - ctgDebug("tbmeta not in cache, dbFName:%s, tbName:%s", dbFName, tbName); - return TSDB_CODE_SUCCESS; + CTG_LOCK(CTG_READ, &pCache->metaLock); + if (NULL == pCache->pMeta) { + ctgDebug("tb %s meta not in cache, dbFName:%s", tbName, dbFName); + goto _return; } - *exist = 1; + *pDb = dbCache; + *pTb = pCache; - ctgReleaseDBCache(pCtg, dbCache); + ctgDebug("tb %s meta got in cache, dbFName:%s", tbName, dbFName); - ctgDebug("tbmeta is in cache, dbFName:%s, tbName:%s", dbFName, tbName); + CTG_CACHE_STAT_INC(tbMetaHitNum, 1); + + return TSDB_CODE_SUCCESS; + +_return: + + ctgReleaseTbMetaToCache(pCtg, dbCache, pCache); + + CTG_CACHE_STAT_INC(tbMetaMissNum, 1); + + return TSDB_CODE_SUCCESS; +} + +int32_t ctgAcquireStbMetaFromCache(SCatalog* pCtg, char *dbFName, uint64_t suid, SCtgDBCache **pDb, SCtgTbCache** pTb) { + SCtgDBCache* dbCache = NULL; + SCtgTbCache* pCache = NULL; + ctgAcquireDBCache(pCtg, dbFName, &dbCache); + if (NULL == dbCache) { + ctgDebug("db %s not in cache", dbFName); + goto _return; + } + + int32_t sz = 0; + char* stName = taosHashAcquire(dbCache->stbCache, &suid, sizeof(suid)); + if (NULL == stName) { + ctgDebug("stb %" PRIx64 " not in cache, dbFName:%s", suid, dbFName); + goto _return; + } + + pCache = taosHashAcquire(dbCache->tbCache, stName, strlen(stName)); + if (NULL == pCache) { + ctgDebug("stb %" PRIx64 " name %s not in cache, dbFName:%s", suid, stName, dbFName); + taosHashRelease(dbCache->stbCache, stName); + goto _return; + } + + CTG_LOCK(CTG_READ, &pCache->metaLock); + if (NULL == pCache->pMeta) { + ctgDebug("stb %" PRIx64 " meta not in cache, dbFName:%s", suid, dbFName); + goto _return; + } + + *pDb = dbCache; + *pTb = pCache; + + ctgDebug("stb %" PRIx64 " meta got in cache, dbFName:%s", suid, dbFName); + + CTG_CACHE_STAT_INC(tbMetaHitNum, 1); + + return TSDB_CODE_SUCCESS; + +_return: + + ctgReleaseTbMetaToCache(pCtg, dbCache, pCache); + + CTG_CACHE_STAT_INC(tbMetaMissNum, 1); + + *pDb = NULL; + *pTb = NULL; return TSDB_CODE_SUCCESS; } +int32_t ctgAcquireTbIndexFromCache(SCatalog* pCtg, char *dbFName, char* tbName, SCtgDBCache **pDb, SCtgTbCache** pTb) { + SCtgDBCache *dbCache = NULL; + ctgAcquireDBCache(pCtg, dbFName, &dbCache); + if (NULL == dbCache) { + ctgDebug("db %s not in cache", dbFName); + goto _return; + } + + int32_t sz = 0; + SCtgTbCache* pCache = taosHashAcquire(dbCache->tbCache, tbName, strlen(tbName)); + if (NULL == pCache) { + ctgDebug("tb %s not in cache, dbFName:%s", tbName, dbFName); + goto _return; + } + + CTG_LOCK(CTG_READ, &pCache->indexLock); + if (NULL == pCache->pIndex) { + ctgDebug("tb %s index not in cache, dbFName:%s", tbName, dbFName); + goto _return; + } + + *pDb = dbCache; + *pTb = pCache; + + ctgDebug("tb %s index got in cache, dbFName:%s", tbName, dbFName); + + CTG_CACHE_STAT_INC(tbIndexHitNum, 1); + + return TSDB_CODE_SUCCESS; + +_return: + + ctgReleaseTbIndexToCache(pCtg, dbCache, pCache); + + CTG_CACHE_STAT_INC(tbIndexMissNum, 1); + + return TSDB_CODE_SUCCESS; +} + + +int32_t ctgTbMetaExistInCache(SCatalog* pCtg, char *dbFName, char* tbName, int32_t *exist) { + SCtgDBCache *dbCache = NULL; + SCtgTbCache *tbCache = NULL; + ctgAcquireTbMetaFromCache(pCtg, dbFName, tbName, &dbCache, &tbCache); + if (NULL == tbCache) { + ctgReleaseTbMetaToCache(pCtg, dbCache, tbCache); + + *exist = 0; + return TSDB_CODE_SUCCESS; + } + + *exist = 1; + ctgReleaseTbMetaToCache(pCtg, dbCache, tbCache); + + return TSDB_CODE_SUCCESS; +} + int32_t ctgReadTbMetaFromCache(SCatalog* pCtg, SCtgTbMetaCtx* ctx, STableMeta** pTableMeta) { int32_t code = 0; SCtgDBCache *dbCache = NULL; - + SCtgTbCache *tbCache = NULL; *pTableMeta = NULL; - if (NULL == pCtg->dbCache) { - ctgDebug("empty tbmeta cache, tbName:%s", ctx->pName->tname); - return TSDB_CODE_SUCCESS; - } - char dbFName[TSDB_DB_FNAME_LEN] = {0}; if (CTG_FLAG_IS_SYS_DB(ctx->flag)) { strcpy(dbFName, ctx->pName->dbname); @@ -257,202 +389,194 @@ int32_t ctgReadTbMetaFromCache(SCatalog* pCtg, SCtgTbMetaCtx* ctx, STableMeta** tNameGetFullDbName(ctx->pName, dbFName); } - ctgAcquireDBCache(pCtg, dbFName, &dbCache); - if (NULL == dbCache) { - ctgDebug("db %d.%s not in cache", ctx->pName->acctId, ctx->pName->dbname); - return TSDB_CODE_SUCCESS; - } - - int32_t sz = 0; - CTG_LOCK(CTG_READ, &dbCache->tbCache.metaLock); - taosHashGetDup_m(dbCache->tbCache.metaCache, ctx->pName->tname, strlen(ctx->pName->tname), (void **)pTableMeta, &sz); - CTG_UNLOCK(CTG_READ, &dbCache->tbCache.metaLock); - - if (NULL == *pTableMeta) { - ctgReleaseDBCache(pCtg, dbCache); - ctgDebug("tbl not in cache, dbFName:%s, tbName:%s", dbFName, ctx->pName->tname); + ctgAcquireTbMetaFromCache(pCtg, dbFName, ctx->pName->tname, &dbCache, &tbCache); + if (NULL == tbCache) { + ctgReleaseTbMetaToCache(pCtg, dbCache, tbCache); return TSDB_CODE_SUCCESS; } - STableMeta* tbMeta = *pTableMeta; + STableMeta* tbMeta = tbCache->pMeta; ctx->tbInfo.inCache = true; ctx->tbInfo.dbId = dbCache->dbId; ctx->tbInfo.suid = tbMeta->suid; ctx->tbInfo.tbType = tbMeta->tableType; - + if (tbMeta->tableType != TSDB_CHILD_TABLE) { - ctgReleaseDBCache(pCtg, dbCache); - ctgDebug("Got meta from cache, type:%d, dbFName:%s, tbName:%s", tbMeta->tableType, dbFName, ctx->pName->tname); + int32_t metaSize = CTG_META_SIZE(tbMeta); + *pTableMeta = taosMemoryCalloc(1, metaSize); + if (NULL == *pTableMeta) { + ctgReleaseTbMetaToCache(pCtg, dbCache, tbCache); + CTG_ERR_JRET(TSDB_CODE_OUT_OF_MEMORY); + } - CTG_CACHE_STAT_INC(tblHitNum, 1); + memcpy(*pTableMeta, tbMeta, metaSize); + + ctgReleaseTbMetaToCache(pCtg, dbCache, tbCache); + ctgDebug("Got tb %s meta from cache, type:%d, dbFName:%s", ctx->pName->tname, tbMeta->tableType, dbFName); + return TSDB_CODE_SUCCESS; + } + + // PROCESS FOR CHILD TABLE + + int32_t metaSize = sizeof(SCTableMeta); + *pTableMeta = taosMemoryCalloc(1, metaSize); + if (NULL == *pTableMeta) { + CTG_ERR_JRET(TSDB_CODE_OUT_OF_MEMORY); + } + + memcpy(*pTableMeta, tbMeta, metaSize); + + ctgReleaseTbMetaToCache(pCtg, dbCache, tbCache); + ctgDebug("Got ctb %s meta from cache, will continue to get its stb meta, type:%d, dbFName:%s", + ctx->pName->tname, ctx->tbInfo.tbType, dbFName); + + ctgAcquireStbMetaFromCache(pCtg, dbFName, ctx->tbInfo.suid, &dbCache, &tbCache); + if (NULL == tbCache) { + ctgReleaseTbMetaToCache(pCtg, dbCache, tbCache); + taosMemoryFreeClear(*pTableMeta); + ctgDebug("stb %" PRIx64 " meta not in cache", ctx->tbInfo.suid); return TSDB_CODE_SUCCESS; } - CTG_LOCK(CTG_READ, &dbCache->tbCache.stbLock); - - STableMeta **stbMeta = taosHashGet(dbCache->tbCache.stbCache, &tbMeta->suid, sizeof(tbMeta->suid)); - if (NULL == stbMeta || NULL == *stbMeta) { - CTG_UNLOCK(CTG_READ, &dbCache->tbCache.stbLock); - ctgError("stb not in stbCache, suid:%"PRIx64, tbMeta->suid); - goto _return; - } - - if ((*stbMeta)->suid != tbMeta->suid) { - CTG_UNLOCK(CTG_READ, &dbCache->tbCache.stbLock); - ctgError("stable suid in stbCache mis-match, expected suid:%"PRIx64 ",actual suid:%"PRIx64, tbMeta->suid, (*stbMeta)->suid); + STableMeta* stbMeta = tbCache->pMeta; + if (stbMeta->suid != ctx->tbInfo.suid) { + ctgReleaseTbMetaToCache(pCtg, dbCache, tbCache); + ctgError("stb suid %" PRIx64 " in stbCache mis-match, expected suid:%"PRIx64 , stbMeta->suid, ctx->tbInfo.suid); CTG_ERR_JRET(TSDB_CODE_CTG_INTERNAL_ERROR); } - int32_t metaSize = CTG_META_SIZE(*stbMeta); + metaSize = CTG_META_SIZE(stbMeta); *pTableMeta = taosMemoryRealloc(*pTableMeta, metaSize); if (NULL == *pTableMeta) { - CTG_UNLOCK(CTG_READ, &dbCache->tbCache.stbLock); - ctgError("realloc size[%d] failed", metaSize); + ctgReleaseTbMetaToCache(pCtg, dbCache, tbCache); CTG_ERR_JRET(TSDB_CODE_CTG_MEM_ERROR); } - memcpy(&(*pTableMeta)->sversion, &(*stbMeta)->sversion, metaSize - sizeof(SCTableMeta)); + memcpy(&(*pTableMeta)->sversion, &stbMeta->sversion, metaSize - sizeof(SCTableMeta)); - CTG_UNLOCK(CTG_READ, &dbCache->tbCache.stbLock); + ctgReleaseTbMetaToCache(pCtg, dbCache, tbCache); - ctgReleaseDBCache(pCtg, dbCache); - - CTG_CACHE_STAT_INC(tblHitNum, 1); - - ctgDebug("Got tbmeta from cache, dbFName:%s, tbName:%s", dbFName, ctx->pName->tname); + ctgDebug("Got tb %s meta from cache, dbFName:%s", ctx->pName->tname, dbFName); return TSDB_CODE_SUCCESS; _return: - ctgReleaseDBCache(pCtg, dbCache); + ctgReleaseTbMetaToCache(pCtg, dbCache, tbCache); taosMemoryFreeClear(*pTableMeta); - - CTG_CACHE_STAT_INC(tblMissNum, 1); CTG_RET(code); } -int32_t ctgReadTbVerFromCache(SCatalog *pCtg, const SName *pTableName, int32_t *sver, int32_t *tver, int32_t *tbType, uint64_t *suid, +int32_t ctgReadTbVerFromCache(SCatalog *pCtg, SName *pTableName, int32_t *sver, int32_t *tver, int32_t *tbType, uint64_t *suid, char *stbName) { *sver = -1; *tver = -1; - if (NULL == pCtg->dbCache) { - ctgDebug("empty tbmeta cache, tbName:%s", pTableName->tname); - return TSDB_CODE_SUCCESS; - } - SCtgDBCache *dbCache = NULL; + SCtgTbCache *tbCache = NULL; char dbFName[TSDB_DB_FNAME_LEN] = {0}; tNameGetFullDbName(pTableName, dbFName); - ctgAcquireDBCache(pCtg, dbFName, &dbCache); - if (NULL == dbCache) { - ctgDebug("db %s not in cache", pTableName->tname); + ctgAcquireTbMetaFromCache(pCtg, dbFName, pTableName->tname, &dbCache, &tbCache); + if (NULL == tbCache) { + ctgReleaseTbMetaToCache(pCtg, dbCache, tbCache); return TSDB_CODE_SUCCESS; } - CTG_LOCK(CTG_READ, &dbCache->tbCache.metaLock); - STableMeta *tbMeta = taosHashGet(dbCache->tbCache.metaCache, pTableName->tname, strlen(pTableName->tname)); - if (tbMeta) { - *tbType = tbMeta->tableType; - *suid = tbMeta->suid; - if (*tbType != TSDB_CHILD_TABLE) { - *sver = tbMeta->sversion; - *tver = tbMeta->tversion; - } - } - CTG_UNLOCK(CTG_READ, &dbCache->tbCache.metaLock); - - if (NULL == tbMeta) { - ctgReleaseDBCache(pCtg, dbCache); - return TSDB_CODE_SUCCESS; - } + STableMeta* tbMeta = tbCache->pMeta; + *tbType = tbMeta->tableType; + *suid = tbMeta->suid; if (*tbType != TSDB_CHILD_TABLE) { - ctgReleaseDBCache(pCtg, dbCache); - ctgDebug("Got sver %d tver %d from cache, type:%d, dbFName:%s, tbName:%s", *sver, *tver, *tbType, dbFName, pTableName->tname); + *sver = tbMeta->sversion; + *tver = tbMeta->tversion; + ctgDebug("Got tb %s ver from cache, dbFName:%s, tbType:%d, sver:%d, tver:%d, suid:%" PRIx64, + pTableName->tname, dbFName, *tbType, *sver, *tver, *suid); + + ctgReleaseTbMetaToCache(pCtg, dbCache, tbCache); return TSDB_CODE_SUCCESS; } - ctgDebug("Got subtable meta from cache, dbFName:%s, tbName:%s, suid:%" PRIx64, dbFName, pTableName->tname, *suid); - - CTG_LOCK(CTG_READ, &dbCache->tbCache.stbLock); - - STableMeta **stbMeta = taosHashGet(dbCache->tbCache.stbCache, suid, sizeof(*suid)); - if (NULL == stbMeta || NULL == *stbMeta) { - CTG_UNLOCK(CTG_READ, &dbCache->tbCache.stbLock); - ctgReleaseDBCache(pCtg, dbCache); - ctgDebug("stb not in stbCache, suid:%" PRIx64, *suid); + // PROCESS FOR CHILD TABLE + + ctgReleaseTbMetaToCache(pCtg, dbCache, tbCache); + ctgDebug("Got ctb %s ver from cache, will continue to get its stb ver, dbFName:%s", pTableName->tname, dbFName); + + ctgAcquireStbMetaFromCache(pCtg, dbFName, *suid, &dbCache, &tbCache); + if (NULL == tbCache) { + ctgReleaseTbMetaToCache(pCtg, dbCache, tbCache); + ctgDebug("stb %" PRIx64 " meta not in cache", *suid); return TSDB_CODE_SUCCESS; } - - if ((*stbMeta)->suid != *suid) { - CTG_UNLOCK(CTG_READ, &dbCache->tbCache.stbLock); - ctgReleaseDBCache(pCtg, dbCache); - ctgError("stable suid in stbCache mis-match, expected suid:%" PRIx64 ",actual suid:%" PRIx64, *suid, - (*stbMeta)->suid); + + STableMeta* stbMeta = tbCache->pMeta; + if (stbMeta->suid != *suid) { + ctgReleaseTbMetaToCache(pCtg, dbCache, tbCache); + ctgError("stb suid %" PRIx64 " in stbCache mis-match, expected suid:%" PRIx64 , stbMeta->suid, *suid); CTG_ERR_RET(TSDB_CODE_CTG_INTERNAL_ERROR); } size_t nameLen = 0; - char *name = taosHashGetKey(*stbMeta, &nameLen); + char *name = taosHashGetKey(tbCache, &nameLen); strncpy(stbName, name, nameLen); stbName[nameLen] = 0; - *sver = (*stbMeta)->sversion; - *tver = (*stbMeta)->tversion; + *sver = stbMeta->sversion; + *tver = stbMeta->tversion; - CTG_UNLOCK(CTG_READ, &dbCache->tbCache.stbLock); + ctgReleaseTbMetaToCache(pCtg, dbCache, tbCache); - ctgReleaseDBCache(pCtg, dbCache); - - ctgDebug("Got sver %d tver %d from cache, type:%d, dbFName:%s, tbName:%s", *sver, *tver, *tbType, dbFName, pTableName->tname); + ctgDebug("Got tb %s sver %d tver %d from cache, type:%d, dbFName:%s", pTableName->tname, *sver, *tver, *tbType, dbFName); return TSDB_CODE_SUCCESS; } -int32_t ctgReadTbTypeFromCache(SCatalog* pCtg, const char* dbFName, const char *tableName, int32_t *tbType) { - if (NULL == pCtg->dbCache) { - ctgWarn("empty db cache, dbFName:%s, tbName:%s", dbFName, tableName); - return TSDB_CODE_SUCCESS; - } - +int32_t ctgReadTbTypeFromCache(SCatalog* pCtg, char* dbFName, char *tableName, int32_t *tbType) { SCtgDBCache *dbCache = NULL; - ctgAcquireDBCache(pCtg, dbFName, &dbCache); - if (NULL == dbCache) { + SCtgTbCache *tbCache = NULL; + CTG_ERR_RET(ctgAcquireTbMetaFromCache(pCtg, dbFName, tableName, &dbCache, &tbCache)); + if (NULL == tbCache) { + ctgReleaseTbMetaToCache(pCtg, dbCache, tbCache); return TSDB_CODE_SUCCESS; } - CTG_LOCK(CTG_READ, &dbCache->tbCache.metaLock); - STableMeta *pTableMeta = (STableMeta *)taosHashAcquire(dbCache->tbCache.metaCache, tableName, strlen(tableName)); + *tbType = tbCache->pMeta->tableType; + ctgReleaseTbMetaToCache(pCtg, dbCache, tbCache); - if (NULL == pTableMeta) { - CTG_UNLOCK(CTG_READ, &dbCache->tbCache.metaLock); - ctgWarn("tbl not in cache, dbFName:%s, tbName:%s", dbFName, tableName); - ctgReleaseDBCache(pCtg, dbCache); - - return TSDB_CODE_SUCCESS; - } - - *tbType = atomic_load_8(&pTableMeta->tableType); - - taosHashRelease(dbCache->tbCache.metaCache, pTableMeta); - - CTG_UNLOCK(CTG_READ, &dbCache->tbCache.metaLock); - - ctgReleaseDBCache(pCtg, dbCache); - - ctgDebug("Got tbtype from cache, dbFName:%s, tbName:%s, type:%d", dbFName, tableName, *tbType); + ctgDebug("Got tb %s tbType %d from cache, dbFName:%s", tableName, *tbType, dbFName); return TSDB_CODE_SUCCESS; } -int32_t ctgChkAuthFromCache(SCatalog* pCtg, const char* user, const char* dbFName, AUTH_TYPE type, bool *inCache, bool *pass) { +int32_t ctgReadTbIndexFromCache(SCatalog* pCtg, SName* pTableName, SArray** pRes) { + int32_t code = 0; + SCtgDBCache *dbCache = NULL; + SCtgTbCache *tbCache = NULL; + char dbFName[TSDB_DB_FNAME_LEN] = {0}; + tNameGetFullDbName(pTableName, dbFName); + + *pRes = NULL; + + ctgAcquireTbIndexFromCache(pCtg, dbFName, pTableName->tname, &dbCache, &tbCache); + if (NULL == tbCache) { + ctgReleaseTbIndexToCache(pCtg, dbCache, tbCache); + return TSDB_CODE_SUCCESS; + } + + CTG_ERR_JRET(ctgCloneTableIndex(tbCache->pIndex->pIndex, pRes)); + +_return: + + ctgReleaseTbIndexToCache(pCtg, dbCache, tbCache); + + CTG_RET(code); +} + +int32_t ctgChkAuthFromCache(SCatalog* pCtg, char* user, char* dbFName, AUTH_TYPE type, bool *inCache, bool *pass) { if (NULL == pCtg->userCache) { ctgDebug("empty user auth cache, user:%s", user); goto _return; @@ -718,9 +842,9 @@ int32_t ctgUpdateTbMetaEnqueue(SCatalog* pCtg, STableMetaOutput *output, bool sy op->opId = CTG_OP_UPDATE_TB_META; op->syncOp = syncOp; - SCtgUpdateTblMsg *msg = taosMemoryMalloc(sizeof(SCtgUpdateTblMsg)); + SCtgUpdateTbMetaMsg *msg = taosMemoryMalloc(sizeof(SCtgUpdateTbMetaMsg)); if (NULL == msg) { - ctgError("malloc %d failed", (int32_t)sizeof(SCtgUpdateTblMsg)); + ctgError("malloc %d failed", (int32_t)sizeof(SCtgUpdateTbMetaMsg)); CTG_ERR_RET(TSDB_CODE_CTG_MEM_ERROR); } @@ -730,7 +854,7 @@ int32_t ctgUpdateTbMetaEnqueue(SCatalog* pCtg, STableMetaOutput *output, bool sy } msg->pCtg = pCtg; - msg->output = output; + msg->pMeta = output; op->data = msg; @@ -805,6 +929,67 @@ _return: CTG_RET(code); } +int32_t ctgUpdateTbIndexEnqueue(SCatalog* pCtg, STableIndex **pIndex, bool syncOp) { + int32_t code = 0; + SCtgCacheOperation *op = taosMemoryCalloc(1, sizeof(SCtgCacheOperation)); + op->opId = CTG_OP_UPDATE_TB_INDEX; + op->syncOp = syncOp; + + SCtgUpdateTbIndexMsg *msg = taosMemoryMalloc(sizeof(SCtgUpdateTbIndexMsg)); + if (NULL == msg) { + ctgError("malloc %d failed", (int32_t)sizeof(SCtgUpdateTbIndexMsg)); + CTG_ERR_JRET(TSDB_CODE_CTG_MEM_ERROR); + } + + msg->pCtg = pCtg; + msg->pIndex = *pIndex; + + op->data = msg; + + CTG_ERR_JRET(ctgEnqueue(pCtg, op)); + + *pIndex = NULL; + return TSDB_CODE_SUCCESS; + +_return: + + taosArrayDestroyEx((*pIndex)->pIndex, tFreeSTableIndexInfo); + taosMemoryFreeClear(*pIndex); + taosMemoryFreeClear(msg); + + CTG_RET(code); +} + +int32_t ctgDropTbIndexEnqueue(SCatalog* pCtg, SName* pName, bool syncOp) { + int32_t code = 0; + SCtgCacheOperation *op = taosMemoryCalloc(1, sizeof(SCtgCacheOperation)); + op->opId = CTG_OP_DROP_TB_INDEX; + op->syncOp = syncOp; + + SCtgDropTbIndexMsg *msg = taosMemoryMalloc(sizeof(SCtgDropTbIndexMsg)); + if (NULL == msg) { + ctgError("malloc %d failed", (int32_t)sizeof(SCtgDropTbIndexMsg)); + CTG_ERR_RET(TSDB_CODE_CTG_MEM_ERROR); + } + + msg->pCtg = pCtg; + tNameGetFullDbName(pName, msg->dbFName); + strcpy(msg->tbName, pName->tname); + + op->data = msg; + + CTG_ERR_JRET(ctgEnqueue(pCtg, op)); + + return TSDB_CODE_SUCCESS; + +_return: + + taosMemoryFreeClear(msg); + + CTG_RET(code); +} + + int32_t ctgMetaRentInit(SCtgRentMgmt *mgmt, uint32_t rentSec, int8_t type) { mgmt->slotRIdx = 0; mgmt->slotNum = rentSec / CTG_RENT_SLOT_SECOND; @@ -875,7 +1060,7 @@ int32_t ctgMetaRentUpdate(SCtgRentMgmt *mgmt, void *meta, int64_t id, int32_t si void *orig = taosArraySearch(slot->meta, &id, searchCompare, TD_EQ); if (NULL == orig) { - qError("meta not found in slot, id:%"PRIx64", slot idx:%d, type:%d, size:%d", id, widx, mgmt->type, (int32_t)taosArrayGetSize(slot->meta)); + qDebug("meta not found in slot, id:%"PRIx64", slot idx:%d, type:%d, size:%d", id, widx, mgmt->type, (int32_t)taosArrayGetSize(slot->meta)); CTG_ERR_JRET(TSDB_CODE_CTG_INTERNAL_ERROR); } @@ -888,7 +1073,7 @@ _return: CTG_UNLOCK(CTG_WRITE, &slot->lock); if (code) { - qWarn("meta in rent update failed, will try to add it, code:%x, id:%"PRIx64", slot idx:%d, type:%d", code, id, widx, mgmt->type); + qDebug("meta in rent update failed, will try to add it, code:%x, id:%"PRIx64", slot idx:%d, type:%d", code, id, widx, mgmt->type); CTG_RET(ctgMetaRentAdd(mgmt, meta, id, size)); } @@ -1006,14 +1191,14 @@ int32_t ctgAddNewDBCache(SCatalog *pCtg, const char *dbFName, uint64_t dbId) { SCtgDBCache newDBCache = {0}; newDBCache.dbId = dbId; - newDBCache.tbCache.metaCache = taosHashInit(gCtgMgmt.cfg.maxTblCacheNum, taosGetDefaultHashFunction(TSDB_DATA_TYPE_BINARY), true, HASH_ENTRY_LOCK); - if (NULL == newDBCache.tbCache.metaCache) { + newDBCache.tbCache = taosHashInit(gCtgMgmt.cfg.maxTblCacheNum, taosGetDefaultHashFunction(TSDB_DATA_TYPE_BINARY), true, HASH_ENTRY_LOCK); + if (NULL == newDBCache.tbCache) { ctgError("taosHashInit %d metaCache failed", gCtgMgmt.cfg.maxTblCacheNum); CTG_ERR_RET(TSDB_CODE_CTG_MEM_ERROR); } - newDBCache.tbCache.stbCache = taosHashInit(gCtgMgmt.cfg.maxTblCacheNum, taosGetDefaultHashFunction(TSDB_DATA_TYPE_UBIGINT), true, HASH_ENTRY_LOCK); - if (NULL == newDBCache.tbCache.stbCache) { + newDBCache.stbCache = taosHashInit(gCtgMgmt.cfg.maxTblCacheNum, taosGetDefaultHashFunction(TSDB_DATA_TYPE_UBIGINT), true, HASH_ENTRY_LOCK); + if (NULL == newDBCache.stbCache) { ctgError("taosHashInit %d stbCache failed", gCtgMgmt.cfg.maxTblCacheNum); CTG_ERR_JRET(TSDB_CODE_CTG_MEM_ERROR); } @@ -1050,22 +1235,22 @@ _return: } -void ctgRemoveStbRent(SCatalog* pCtg, SCtgTbMetaCache *cache) { - CTG_LOCK(CTG_WRITE, &cache->stbLock); - if (cache->stbCache) { - void *pIter = taosHashIterate(cache->stbCache, NULL); - while (pIter) { - uint64_t *suid = NULL; - suid = taosHashGetKey(pIter, NULL); - - if (TSDB_CODE_SUCCESS == ctgMetaRentRemove(&pCtg->stbRent, *suid, ctgStbVersionSortCompare, ctgStbVersionSearchCompare)) { - ctgDebug("stb removed from rent, suid:%"PRIx64, *suid); - } - - pIter = taosHashIterate(cache->stbCache, pIter); - } +void ctgRemoveStbRent(SCatalog* pCtg, SCtgDBCache *dbCache) { + if (NULL == dbCache->stbCache) { + return; + } + + void *pIter = taosHashIterate(dbCache->stbCache, NULL); + while (pIter) { + uint64_t *suid = NULL; + suid = taosHashGetKey(pIter, NULL); + + if (TSDB_CODE_SUCCESS == ctgMetaRentRemove(&pCtg->stbRent, *suid, ctgStbVersionSortCompare, ctgStbVersionSearchCompare)) { + ctgDebug("stb removed from rent, suid:%"PRIx64, *suid); + } + + pIter = taosHashIterate(dbCache->stbCache, pIter); } - CTG_UNLOCK(CTG_WRITE, &cache->stbLock); } @@ -1074,15 +1259,16 @@ int32_t ctgRemoveDBFromCache(SCatalog* pCtg, SCtgDBCache *dbCache, const char* d ctgInfo("start to remove db from cache, dbFName:%s, dbId:%"PRIx64, dbFName, dbCache->dbId); + CTG_LOCK(CTG_WRITE, &dbCache->dbLock); + atomic_store_8(&dbCache->deleted, 1); - - ctgRemoveStbRent(pCtg, &dbCache->tbCache); - + ctgRemoveStbRent(pCtg, dbCache); ctgFreeDbCache(dbCache); - CTG_ERR_RET(ctgMetaRentRemove(&pCtg->dbRent, dbCache->dbId, ctgDbVgVersionSortCompare, ctgDbVgVersionSearchCompare)); - - ctgDebug("db removed from rent, dbFName:%s, dbId:%"PRIx64, dbFName, dbCache->dbId); + CTG_UNLOCK(CTG_WRITE, &dbCache->dbLock); + + CTG_ERR_RET(ctgMetaRentRemove(&pCtg->dbRent, dbId, ctgDbVgVersionSortCompare, ctgDbVgVersionSearchCompare)); + ctgDebug("db removed from rent, dbFName:%s, dbId:%"PRIx64, dbFName, dbId); if (taosHashRemove(pCtg->dbCache, dbFName, strlen(dbFName))) { ctgInfo("taosHashRemove from dbCache failed, may be removed, dbFName:%s", dbFName); @@ -1090,7 +1276,6 @@ int32_t ctgRemoveDBFromCache(SCatalog* pCtg, SCtgDBCache *dbCache, const char* d } CTG_CACHE_STAT_DEC(dbNum, 1); - ctgInfo("db removed from cache, dbFName:%s, dbId:%"PRIx64, dbFName, dbId); return TSDB_CODE_SUCCESS; @@ -1138,120 +1323,76 @@ int32_t ctgGetAddDBCache(SCatalog* pCtg, const char *dbFName, uint64_t dbId, SCt return TSDB_CODE_SUCCESS; } +int32_t ctgUpdateRentStbVersion(SCatalog *pCtg, char* dbFName, char* tbName, uint64_t dbId, uint64_t suid, SCtgTbCache* pCache) { + SSTableVersion metaRent = {.dbId = dbId, .suid = suid}; + if (pCache->pMeta) { + metaRent.sversion = pCache->pMeta->sversion; + metaRent.tversion = pCache->pMeta->tversion; + } -int32_t ctgWriteDBVgInfoToCache(SCatalog* pCtg, const char* dbFName, uint64_t dbId, SDBVgInfo** pDbInfo) { - int32_t code = 0; - SDBVgInfo* dbInfo = *pDbInfo; - - if (NULL == dbInfo->vgHash) { - return TSDB_CODE_SUCCESS; + if (pCache->pIndex) { + metaRent.smaVer = pCache->pIndex->version; } - if (dbInfo->vgVersion < 0 || taosHashGetSize(dbInfo->vgHash) <= 0) { - ctgError("invalid db vgInfo, dbFName:%s, vgHash:%p, vgVersion:%d, vgHashSize:%d", - dbFName, dbInfo->vgHash, dbInfo->vgVersion, taosHashGetSize(dbInfo->vgHash)); - CTG_ERR_RET(TSDB_CODE_CTG_MEM_ERROR); - } - - bool newAdded = false; - SDbVgVersion vgVersion = {.dbId = dbId, .vgVersion = dbInfo->vgVersion, .numOfTable = dbInfo->numOfTable}; - - SCtgDBCache *dbCache = NULL; - CTG_ERR_RET(ctgGetAddDBCache(pCtg, dbFName, dbId, &dbCache)); - if (NULL == dbCache) { - ctgInfo("conflict db update, ignore this update, dbFName:%s, dbId:%"PRIx64, dbFName, dbId); - CTG_ERR_RET(TSDB_CODE_CTG_INTERNAL_ERROR); - } - - SDBVgInfo *vgInfo = NULL; - CTG_ERR_RET(ctgWAcquireVgInfo(pCtg, dbCache)); + strcpy(metaRent.dbFName, dbFName); + strcpy(metaRent.stbName, tbName); - if (dbCache->vgInfo) { - if (dbInfo->vgVersion < dbCache->vgInfo->vgVersion) { - ctgDebug("db vgVersion is old, dbFName:%s, vgVersion:%d, currentVersion:%d", dbFName, dbInfo->vgVersion, dbCache->vgInfo->vgVersion); - ctgWReleaseVgInfo(dbCache); - - return TSDB_CODE_SUCCESS; - } + CTG_ERR_RET(ctgMetaRentUpdate(&pCtg->stbRent, &metaRent, metaRent.suid, sizeof(SSTableVersion), ctgStbVersionSortCompare, ctgStbVersionSearchCompare)); - if (dbInfo->vgVersion == dbCache->vgInfo->vgVersion && dbInfo->numOfTable == dbCache->vgInfo->numOfTable) { - ctgDebug("no new db vgVersion or numOfTable, dbFName:%s, vgVersion:%d, numOfTable:%d", dbFName, dbInfo->vgVersion, dbInfo->numOfTable); - ctgWReleaseVgInfo(dbCache); - - return TSDB_CODE_SUCCESS; - } + ctgDebug("db %s,%" PRIx64 " stb %s,%" PRIx64 " sver %d tver %d smaVer %d updated to stbRent", + dbFName, dbId, tbName, suid, metaRent.sversion, metaRent.tversion, metaRent.smaVer); - ctgFreeVgInfo(dbCache->vgInfo); - } - - dbCache->vgInfo = dbInfo; - - *pDbInfo = NULL; - - ctgDebug("db vgInfo updated, dbFName:%s, vgVersion:%d, dbId:%"PRIx64, dbFName, vgVersion.vgVersion, vgVersion.dbId); - - ctgWReleaseVgInfo(dbCache); - - dbCache = NULL; - - strncpy(vgVersion.dbFName, dbFName, sizeof(vgVersion.dbFName)); - CTG_ERR_RET(ctgMetaRentUpdate(&pCtg->dbRent, &vgVersion, vgVersion.dbId, sizeof(SDbVgVersion), ctgDbVgVersionSortCompare, ctgDbVgVersionSearchCompare)); - - CTG_RET(code); + return TSDB_CODE_SUCCESS; } int32_t ctgWriteTbMetaToCache(SCatalog *pCtg, SCtgDBCache *dbCache, char *dbFName, uint64_t dbId, char *tbName, STableMeta *meta, int32_t metaSize) { - SCtgTbMetaCache *tbCache = &dbCache->tbCache; - - CTG_LOCK(CTG_READ, &tbCache->metaLock); - if (dbCache->deleted || NULL == tbCache->metaCache || NULL == tbCache->stbCache) { - CTG_UNLOCK(CTG_READ, &tbCache->metaLock); + if (NULL == dbCache->tbCache || NULL == dbCache->stbCache) { + taosMemoryFree(meta); ctgError("db is dropping, dbId:%"PRIx64, dbCache->dbId); CTG_ERR_RET(TSDB_CODE_CTG_DB_DROPPED); } + bool isStb = meta->tableType == TSDB_SUPER_TABLE; + SCtgTbCache* pCache = taosHashGet(dbCache->tbCache, tbName, strlen(tbName)); + STableMeta *orig = (pCache ? pCache->pMeta : NULL); int8_t origType = 0; uint64_t origSuid = 0; - bool isStb = meta->tableType == TSDB_SUPER_TABLE; - STableMeta *orig = taosHashGet(tbCache->metaCache, tbName, strlen(tbName)); + if (orig) { origType = orig->tableType; if (origType == meta->tableType && orig->uid == meta->uid && orig->sversion >= meta->sversion && orig->tversion >= meta->tversion) { - CTG_UNLOCK(CTG_READ, &tbCache->metaLock); + taosMemoryFree(meta); + ctgDebug("ignore table %s meta update", tbName); return TSDB_CODE_SUCCESS; } if (origType == TSDB_SUPER_TABLE) { - CTG_LOCK(CTG_WRITE, &tbCache->stbLock); - if (taosHashRemove(tbCache->stbCache, &orig->suid, sizeof(orig->suid))) { + if (taosHashRemove(dbCache->stbCache, &orig->suid, sizeof(orig->suid))) { ctgError("stb not exist in stbCache, dbFName:%s, stb:%s, suid:%"PRIx64, dbFName, tbName, orig->suid); } else { CTG_CACHE_STAT_DEC(stblNum, 1); + ctgDebug("stb removed from stbCache, dbFName:%s, stb:%s, suid:%"PRIx64, dbFName, tbName, orig->suid); } - CTG_UNLOCK(CTG_WRITE, &tbCache->stbLock); - - ctgDebug("stb removed from stbCache, dbFName:%s, stb:%s, suid:%"PRIx64, dbFName, tbName, orig->suid); - ctgMetaRentRemove(&pCtg->stbRent, orig->suid, ctgStbVersionSortCompare, ctgStbVersionSearchCompare); - origSuid = orig->suid; } } - if (isStb) { - CTG_LOCK(CTG_WRITE, &tbCache->stbLock); - } - - if (taosHashPut(tbCache->metaCache, tbName, strlen(tbName), meta, metaSize) != 0) { - if (isStb) { - CTG_UNLOCK(CTG_WRITE, &tbCache->stbLock); + if (NULL == pCache) { + SCtgTbCache cache = {0}; + cache.pMeta = meta; + if (taosHashPut(dbCache->tbCache, tbName, strlen(tbName), &cache, sizeof(SCtgTbCache)) != 0) { + taosMemoryFree(meta); + ctgError("taosHashPut new tbCache failed, dbFName:%s, tbName:%s, tbType:%d", dbFName, tbName, meta->tableType); + CTG_ERR_RET(TSDB_CODE_CTG_MEM_ERROR); } - CTG_UNLOCK(CTG_READ, &tbCache->metaLock); - ctgError("taosHashPut tbmeta to cache failed, dbFName:%s, tbName:%s, tbType:%d", dbFName, tbName, meta->tableType); - CTG_ERR_RET(TSDB_CODE_CTG_MEM_ERROR); + pCache = taosHashGet(dbCache->tbCache, tbName, strlen(tbName)); + } else { + taosMemoryFree(pCache->pMeta); + pCache->pMeta = meta; } if (NULL == orig) { @@ -1262,30 +1403,71 @@ int32_t ctgWriteTbMetaToCache(SCatalog *pCtg, SCtgDBCache *dbCache, char *dbFNam ctgdShowTableMeta(pCtg, tbName, meta); if (!isStb) { - CTG_UNLOCK(CTG_READ, &tbCache->metaLock); return TSDB_CODE_SUCCESS; } - STableMeta *tbMeta = taosHashGet(tbCache->metaCache, tbName, strlen(tbName)); - if (taosHashPut(tbCache->stbCache, &meta->suid, sizeof(meta->suid), &tbMeta, POINTER_BYTES) != 0) { - CTG_UNLOCK(CTG_WRITE, &tbCache->stbLock); - CTG_UNLOCK(CTG_READ, &tbCache->metaLock); - ctgError("taosHashPut stable to stable cache failed, suid:%"PRIx64, meta->suid); + if (origSuid != meta->suid && taosHashPut(dbCache->stbCache, &meta->suid, sizeof(meta->suid), tbName, strlen(tbName) + 1) != 0) { + ctgError("taosHashPut to stable cache failed, suid:%"PRIx64, meta->suid); CTG_ERR_RET(TSDB_CODE_CTG_MEM_ERROR); } CTG_CACHE_STAT_INC(stblNum, 1); + + ctgDebug("stb %" PRIx64 " updated to cache, dbFName:%s, tbName:%s, tbType:%d", meta->suid, dbFName, tbName, meta->tableType); + + CTG_ERR_RET(ctgUpdateRentStbVersion(pCtg, dbFName, tbName, dbId, meta->suid, pCache)); - CTG_UNLOCK(CTG_WRITE, &tbCache->stbLock); + return TSDB_CODE_SUCCESS; +} - CTG_UNLOCK(CTG_READ, &tbCache->metaLock); +int32_t ctgWriteTbIndexToCache(SCatalog *pCtg, SCtgDBCache *dbCache, char* dbFName, char *tbName, STableIndex **index) { + if (NULL == dbCache->tbCache) { + ctgFreeSTableIndex(*index); + taosMemoryFreeClear(*index); + ctgError("db is dropping, dbId:%"PRIx64, dbCache->dbId); + CTG_ERR_RET(TSDB_CODE_CTG_DB_DROPPED); + } - ctgDebug("stb updated to stbCache, dbFName:%s, tbName:%s, tbType:%d", dbFName, tbName, meta->tableType); + STableIndex* pIndex = *index; + uint64_t suid = pIndex->suid; + SCtgTbCache* pCache = taosHashGet(dbCache->tbCache, tbName, strlen(tbName)); + if (NULL == pCache) { + SCtgTbCache cache = {0}; + cache.pIndex = pIndex; + + if (taosHashPut(dbCache->tbCache, tbName, strlen(tbName), &cache, sizeof(cache)) != 0) { + ctgFreeSTableIndex(*index); + taosMemoryFreeClear(*index); + ctgError("taosHashPut new tbCache failed, tbName:%s", tbName); + CTG_ERR_RET(TSDB_CODE_CTG_MEM_ERROR); + } - SSTableMetaVersion metaRent = {.dbId = dbId, .suid = meta->suid, .sversion = meta->sversion, .tversion = meta->tversion}; - strcpy(metaRent.dbFName, dbFName); - strcpy(metaRent.stbName, tbName); - CTG_ERR_RET(ctgMetaRentAdd(&pCtg->stbRent, &metaRent, metaRent.suid, sizeof(SSTableMetaVersion))); + *index = NULL; + ctgDebug("table %s index updated to cache, ver:%d, num:%d", tbName, pIndex->version, (int32_t)taosArrayGetSize(pIndex->pIndex)); + + if (suid) { + CTG_ERR_RET(ctgUpdateRentStbVersion(pCtg, dbFName, tbName, dbCache->dbId, pIndex->suid, &cache)); + } + + return TSDB_CODE_SUCCESS; + } + + if (pCache->pIndex) { + if (0 == suid) { + suid = pCache->pIndex->suid; + } + taosArrayDestroyEx(pCache->pIndex->pIndex, tFreeSTableIndexInfo); + taosMemoryFreeClear(pCache->pIndex); + } + + pCache->pIndex = pIndex; + *index = NULL; + + ctgDebug("table %s index updated to cache, ver:%d, num:%d", tbName, pIndex->version, (int32_t)taosArrayGetSize(pIndex->pIndex)); + + if (suid) { + CTG_ERR_RET(ctgUpdateRentStbVersion(pCtg, dbFName, tbName, dbCache->dbId, suid, pCache)); + } return TSDB_CODE_SUCCESS; } @@ -1305,12 +1487,67 @@ _return: CTG_RET(code); } - int32_t ctgOpUpdateVgroup(SCtgCacheOperation *operation) { int32_t code = 0; SCtgUpdateVgMsg *msg = operation->data; + SDBVgInfo* dbInfo = msg->dbInfo; + char* dbFName = msg->dbFName; + SCatalog* pCtg = msg->pCtg; - CTG_ERR_JRET(ctgWriteDBVgInfoToCache(msg->pCtg, msg->dbFName, msg->dbId, &msg->dbInfo)); + if (NULL == dbInfo->vgHash) { + return TSDB_CODE_SUCCESS; + } + + if (dbInfo->vgVersion < 0 || taosHashGetSize(dbInfo->vgHash) <= 0) { + ctgError("invalid db vgInfo, dbFName:%s, vgHash:%p, vgVersion:%d, vgHashSize:%d", + dbFName, dbInfo->vgHash, dbInfo->vgVersion, taosHashGetSize(dbInfo->vgHash)); + CTG_ERR_RET(TSDB_CODE_CTG_MEM_ERROR); + } + + bool newAdded = false; + SDbVgVersion vgVersion = {.dbId = msg->dbId, .vgVersion = dbInfo->vgVersion, .numOfTable = dbInfo->numOfTable}; + + SCtgDBCache *dbCache = NULL; + CTG_ERR_RET(ctgGetAddDBCache(msg->pCtg, dbFName, msg->dbId, &dbCache)); + if (NULL == dbCache) { + ctgInfo("conflict db update, ignore this update, dbFName:%s, dbId:%"PRIx64, dbFName, msg->dbId); + CTG_ERR_RET(TSDB_CODE_CTG_INTERNAL_ERROR); + } + + SCtgVgCache *vgCache = &dbCache->vgCache; + CTG_ERR_RET(ctgWLockVgInfo(msg->pCtg, dbCache)); + + if (vgCache->vgInfo) { + SDBVgInfo *vgInfo = vgCache->vgInfo; + + if (dbInfo->vgVersion < vgInfo->vgVersion) { + ctgDebug("db vgVer is old, dbFName:%s, vgVer:%d, curVer:%d", dbFName, dbInfo->vgVersion, vgInfo->vgVersion); + ctgWUnlockVgInfo(dbCache); + + return TSDB_CODE_SUCCESS; + } + + if (dbInfo->vgVersion == vgInfo->vgVersion && dbInfo->numOfTable == vgInfo->numOfTable) { + ctgDebug("no new db vgVer or numOfTable, dbFName:%s, vgVer:%d, numOfTable:%d", dbFName, dbInfo->vgVersion, dbInfo->numOfTable); + ctgWUnlockVgInfo(dbCache); + + return TSDB_CODE_SUCCESS; + } + + ctgFreeVgInfo(vgInfo); + } + + vgCache->vgInfo = dbInfo; + msg->dbInfo = NULL; + + ctgDebug("db vgInfo updated, dbFName:%s, vgVer:%d, dbId:%"PRIx64, dbFName, vgVersion.vgVersion, vgVersion.dbId); + + ctgWUnlockVgInfo(dbCache); + + dbCache = NULL; + + strncpy(vgVersion.dbFName, dbFName, sizeof(vgVersion.dbFName)); + CTG_ERR_RET(ctgMetaRentUpdate(&msg->pCtg->dbRent, &vgVersion, vgVersion.dbId, sizeof(SDbVgVersion), ctgDbVgVersionSortCompare, ctgDbVgVersionSearchCompare)); _return: @@ -1356,14 +1593,14 @@ int32_t ctgOpDropDbVgroup(SCtgCacheOperation *operation) { goto _return; } - CTG_ERR_RET(ctgWAcquireVgInfo(pCtg, dbCache)); + CTG_ERR_RET(ctgWLockVgInfo(pCtg, dbCache)); - ctgFreeVgInfo(dbCache->vgInfo); - dbCache->vgInfo = NULL; + ctgFreeVgInfo(dbCache->vgCache.vgInfo); + dbCache->vgCache.vgInfo = NULL; ctgDebug("db vgInfo removed, dbFName:%s", msg->dbFName); - ctgWReleaseVgInfo(dbCache); + ctgWUnlockVgInfo(dbCache); _return: @@ -1375,42 +1612,47 @@ _return: int32_t ctgOpUpdateTbMeta(SCtgCacheOperation *operation) { int32_t code = 0; - SCtgUpdateTblMsg *msg = operation->data; + SCtgUpdateTbMetaMsg *msg = operation->data; SCatalog* pCtg = msg->pCtg; - STableMetaOutput* output = msg->output; + STableMetaOutput* pMeta = msg->pMeta; SCtgDBCache *dbCache = NULL; - if ((!CTG_IS_META_CTABLE(output->metaType)) && NULL == output->tbMeta) { - ctgError("no valid tbmeta got from meta rsp, dbFName:%s, tbName:%s", output->dbFName, output->tbName); + if ((!CTG_IS_META_CTABLE(pMeta->metaType)) && NULL == pMeta->tbMeta) { + ctgError("no valid tbmeta got from meta rsp, dbFName:%s, tbName:%s", pMeta->dbFName, pMeta->tbName); CTG_ERR_JRET(TSDB_CODE_CTG_INTERNAL_ERROR); } - if (CTG_IS_META_BOTH(output->metaType) && TSDB_SUPER_TABLE != output->tbMeta->tableType) { - ctgError("table type error, expected:%d, actual:%d", TSDB_SUPER_TABLE, output->tbMeta->tableType); + if (CTG_IS_META_BOTH(pMeta->metaType) && TSDB_SUPER_TABLE != pMeta->tbMeta->tableType) { + ctgError("table type error, expected:%d, actual:%d", TSDB_SUPER_TABLE, pMeta->tbMeta->tableType); CTG_ERR_JRET(TSDB_CODE_CTG_INTERNAL_ERROR); } - CTG_ERR_JRET(ctgGetAddDBCache(pCtg, output->dbFName, output->dbId, &dbCache)); + CTG_ERR_JRET(ctgGetAddDBCache(pCtg, pMeta->dbFName, pMeta->dbId, &dbCache)); if (NULL == dbCache) { - ctgInfo("conflict db update, ignore this update, dbFName:%s, dbId:%"PRIx64, output->dbFName, output->dbId); + ctgInfo("conflict db update, ignore this update, dbFName:%s, dbId:%" PRIx64, pMeta->dbFName, pMeta->dbId); CTG_ERR_JRET(TSDB_CODE_CTG_INTERNAL_ERROR); } - if (CTG_IS_META_TABLE(output->metaType) || CTG_IS_META_BOTH(output->metaType)) { - int32_t metaSize = CTG_META_SIZE(output->tbMeta); - - CTG_ERR_JRET(ctgWriteTbMetaToCache(pCtg, dbCache, output->dbFName, output->dbId, output->tbName, output->tbMeta, metaSize)); + if (CTG_IS_META_TABLE(pMeta->metaType) || CTG_IS_META_BOTH(pMeta->metaType)) { + int32_t metaSize = CTG_META_SIZE(pMeta->tbMeta); + CTG_ERR_JRET(ctgWriteTbMetaToCache(pCtg, dbCache, pMeta->dbFName, pMeta->dbId, pMeta->tbName, pMeta->tbMeta, metaSize)); + pMeta->tbMeta = NULL; } - if (CTG_IS_META_CTABLE(output->metaType) || CTG_IS_META_BOTH(output->metaType)) { - CTG_ERR_JRET(ctgWriteTbMetaToCache(pCtg, dbCache, output->dbFName, output->dbId, output->ctbName, (STableMeta *)&output->ctbMeta, sizeof(output->ctbMeta))); + if (CTG_IS_META_CTABLE(pMeta->metaType) || CTG_IS_META_BOTH(pMeta->metaType)) { + SCTableMeta* ctbMeta = taosMemoryMalloc(sizeof(SCTableMeta)); + if (NULL == ctbMeta) { + CTG_ERR_JRET(TSDB_CODE_OUT_OF_MEMORY); + } + memcpy(ctbMeta, &pMeta->ctbMeta, sizeof(SCTableMeta)); + CTG_ERR_JRET(ctgWriteTbMetaToCache(pCtg, dbCache, pMeta->dbFName, pMeta->dbId, pMeta->ctbName, (STableMeta *)ctbMeta, sizeof(SCTableMeta))); } _return: - if (output) { - taosMemoryFreeClear(output->tbMeta); - taosMemoryFreeClear(output); + if (pMeta) { + taosMemoryFreeClear(pMeta->tbMeta); + taosMemoryFreeClear(pMeta); } taosMemoryFreeClear(msg); @@ -1435,22 +1677,17 @@ int32_t ctgOpDropStbMeta(SCtgCacheOperation *operation) { return TSDB_CODE_SUCCESS; } - CTG_LOCK(CTG_WRITE, &dbCache->tbCache.stbLock); - if (taosHashRemove(dbCache->tbCache.stbCache, &msg->suid, sizeof(msg->suid))) { + if (taosHashRemove(dbCache->stbCache, &msg->suid, sizeof(msg->suid))) { ctgDebug("stb not exist in stbCache, may be removed, dbFName:%s, stb:%s, suid:%"PRIx64, msg->dbFName, msg->stbName, msg->suid); } else { CTG_CACHE_STAT_DEC(stblNum, 1); } - CTG_LOCK(CTG_READ, &dbCache->tbCache.metaLock); - if (taosHashRemove(dbCache->tbCache.metaCache, msg->stbName, strlen(msg->stbName))) { + if (taosHashRemove(dbCache->tbCache, msg->stbName, strlen(msg->stbName))) { ctgError("stb not exist in cache, dbFName:%s, stb:%s, suid:%"PRIx64, msg->dbFName, msg->stbName, msg->suid); } else { CTG_CACHE_STAT_DEC(tblNum, 1); } - CTG_UNLOCK(CTG_READ, &dbCache->tbCache.metaLock); - - CTG_UNLOCK(CTG_WRITE, &dbCache->tbCache.stbLock); ctgInfo("stb removed from cache, dbFName:%s, stbName:%s, suid:%"PRIx64, msg->dbFName, msg->stbName, msg->suid); @@ -1477,21 +1714,18 @@ int32_t ctgOpDropTbMeta(SCtgCacheOperation *operation) { } if (dbCache->dbId != msg->dbId) { - ctgDebug("dbId already modified, dbFName:%s, current:%"PRIx64", dbId:%"PRIx64", tbName:%s", msg->dbFName, dbCache->dbId, msg->dbId, msg->tbName); + ctgDebug("dbId %" PRIx64 " not match with curId %"PRIx64", dbFName:%s, tbName:%s", msg->dbId, dbCache->dbId, msg->dbFName, msg->tbName); return TSDB_CODE_SUCCESS; } - CTG_LOCK(CTG_READ, &dbCache->tbCache.metaLock); - if (taosHashRemove(dbCache->tbCache.metaCache, msg->tbName, strlen(msg->tbName))) { - CTG_UNLOCK(CTG_READ, &dbCache->tbCache.metaLock); - ctgError("stb not exist in cache, dbFName:%s, tbName:%s", msg->dbFName, msg->tbName); - CTG_ERR_RET(TSDB_CODE_CTG_INTERNAL_ERROR); + if (taosHashRemove(dbCache->tbCache, msg->tbName, strlen(msg->tbName))) { + ctgError("tb %s not exist in cache, dbFName:%s", msg->tbName, msg->dbFName); + CTG_ERR_JRET(TSDB_CODE_CTG_INTERNAL_ERROR); } else { CTG_CACHE_STAT_DEC(tblNum, 1); } - CTG_UNLOCK(CTG_READ, &dbCache->tbCache.metaLock); - ctgInfo("table removed from cache, dbFName:%s, tbName:%s", msg->dbFName, msg->tbName); + ctgDebug("table %s removed from cache, dbFName:%s", msg->tbName, msg->dbFName); _return: @@ -1553,7 +1787,6 @@ int32_t ctgOpUpdateUser(SCtgCacheOperation *operation) { _return: - taosHashCleanup(msg->userAuth.createdDbs); taosHashCleanup(msg->userAuth.readDbs); taosHashCleanup(msg->userAuth.writeDbs); @@ -1569,24 +1802,22 @@ int32_t ctgOpUpdateEpset(SCtgCacheOperation *operation) { SCatalog* pCtg = msg->pCtg; SCtgDBCache *dbCache = NULL; - CTG_ERR_RET(ctgAcquireDBCache(pCtg, msg->dbFName, &dbCache)); + CTG_ERR_JRET(ctgGetDBCache(pCtg, msg->dbFName, &dbCache)); if (NULL == dbCache) { ctgDebug("db %s not exist, ignore epset update", msg->dbFName); goto _return; } - SDBVgInfo *vgInfo = NULL; - CTG_ERR_RET(ctgWAcquireVgInfo(pCtg, dbCache)); - - if (NULL == dbCache->vgInfo) { - ctgWReleaseVgInfo(dbCache); + CTG_ERR_JRET(ctgWLockVgInfo(pCtg, dbCache)); + + SDBVgInfo *vgInfo = dbCache->vgCache.vgInfo; + if (NULL == vgInfo) { ctgDebug("vgroup in db %s not cached, ignore epset update", msg->dbFName); goto _return; } - SVgroupInfo* pInfo = taosHashGet(dbCache->vgInfo->vgHash, &msg->vgId, sizeof(msg->vgId)); + SVgroupInfo* pInfo = taosHashGet(vgInfo->vgHash, &msg->vgId, sizeof(msg->vgId)); if (NULL == pInfo) { - ctgWReleaseVgInfo(dbCache); ctgDebug("no vgroup %d in db %s, ignore epset update", msg->vgId, msg->dbFName); goto _return; } @@ -1599,12 +1830,10 @@ int32_t ctgOpUpdateEpset(SCtgCacheOperation *operation) { pInfo->epSet = msg->epSet; - ctgWReleaseVgInfo(dbCache); - _return: if (dbCache) { - ctgReleaseDBCache(msg->pCtg, dbCache); + ctgWUnlockVgInfo(dbCache); } taosMemoryFreeClear(msg); @@ -1612,6 +1841,62 @@ _return: CTG_RET(code); } +int32_t ctgOpUpdateTbIndex(SCtgCacheOperation *operation) { + int32_t code = 0; + SCtgUpdateTbIndexMsg *msg = operation->data; + SCatalog* pCtg = msg->pCtg; + STableIndex* pIndex = msg->pIndex; + SCtgDBCache *dbCache = NULL; + + CTG_ERR_JRET(ctgGetAddDBCache(pCtg, pIndex->dbFName, 0, &dbCache)); + + CTG_ERR_JRET(ctgWriteTbIndexToCache(pCtg, dbCache, pIndex->dbFName, pIndex->tbName, &pIndex)); + +_return: + + if (pIndex) { + taosArrayDestroyEx(pIndex->pIndex, tFreeSTableIndexInfo); + taosMemoryFreeClear(pIndex); + } + + taosMemoryFreeClear(msg); + + CTG_RET(code); +} + +int32_t ctgOpDropTbIndex(SCtgCacheOperation *operation) { + int32_t code = 0; + SCtgDropTbIndexMsg *msg = operation->data; + SCatalog* pCtg = msg->pCtg; + SCtgDBCache *dbCache = NULL; + + CTG_ERR_JRET(ctgGetDBCache(pCtg, msg->dbFName, &dbCache)); + if (NULL == dbCache) { + return TSDB_CODE_SUCCESS; + } + + STableIndex* pIndex = taosMemoryCalloc(1, sizeof(STableIndex)); + if (NULL == pIndex) { + CTG_ERR_JRET(TSDB_CODE_OUT_OF_MEMORY); + } + strcpy(pIndex->tbName, msg->tbName); + strcpy(pIndex->dbFName, msg->dbFName); + pIndex->version = -1; + + CTG_ERR_JRET(ctgWriteTbIndexToCache(pCtg, dbCache, pIndex->dbFName, pIndex->tbName, &pIndex)); + +_return: + + if (pIndex) { + taosArrayDestroyEx(pIndex->pIndex, tFreeSTableIndexInfo); + taosMemoryFreeClear(pIndex); + } + + taosMemoryFreeClear(msg); + + CTG_RET(code); +} + void ctgUpdateThreadUnexpectedStopped(void) { if (CTG_IS_LOCKED(&gCtgMgmt.lock) > 0) CTG_UNLOCK(CTG_READ, &gCtgMgmt.lock); diff --git a/source/libs/catalog/src/ctgDbg.c b/source/libs/catalog/src/ctgDbg.c index d70a176631..9de1ea22be 100644 --- a/source/libs/catalog/src/ctgDbg.c +++ b/source/libs/catalog/src/ctgDbg.c @@ -19,7 +19,7 @@ #include "catalogInt.h" extern SCatalogMgmt gCtgMgmt; -SCtgDebug gCTGDebug = {.apiEnable = true}; +SCtgDebug gCTGDebug = {0}; void ctgdUserCallback(SMetaData* pResult, void* param, int32_t code) { ASSERT(*(int32_t*)param == 1); @@ -266,11 +266,11 @@ int32_t ctgdGetStatNum(char *option, void *res) { } int32_t ctgdGetTbMetaNum(SCtgDBCache *dbCache) { - return dbCache->tbCache.metaCache ? (int32_t)taosHashGetSize(dbCache->tbCache.metaCache) : 0; + return dbCache->tbCache ? (int32_t)taosHashGetSize(dbCache->tbCache) : 0; } int32_t ctgdGetStbNum(SCtgDBCache *dbCache) { - return dbCache->tbCache.stbCache ? (int32_t)taosHashGetSize(dbCache->tbCache.stbCache) : 0; + return dbCache->stbCache ? (int32_t)taosHashGetSize(dbCache->stbCache) : 0; } int32_t ctgdGetRentNum(SCtgRentMgmt *rent) { @@ -363,17 +363,17 @@ void ctgdShowDBCache(SCatalog* pCtg, SHashObj *dbHash) { dbFName = taosHashGetKey(pIter, &len); - int32_t metaNum = dbCache->tbCache.metaCache ? taosHashGetSize(dbCache->tbCache.metaCache) : 0; - int32_t stbNum = dbCache->tbCache.stbCache ? taosHashGetSize(dbCache->tbCache.stbCache) : 0; + int32_t metaNum = dbCache->tbCache ? taosHashGetSize(dbCache->tbCache) : 0; + int32_t stbNum = dbCache->stbCache ? taosHashGetSize(dbCache->stbCache) : 0; int32_t vgVersion = CTG_DEFAULT_INVALID_VERSION; int32_t hashMethod = -1; int32_t vgNum = 0; - if (dbCache->vgInfo) { - vgVersion = dbCache->vgInfo->vgVersion; - hashMethod = dbCache->vgInfo->hashMethod; - if (dbCache->vgInfo->vgHash) { - vgNum = taosHashGetSize(dbCache->vgInfo->vgHash); + if (dbCache->vgCache.vgInfo) { + vgVersion = dbCache->vgCache.vgInfo->vgVersion; + hashMethod = dbCache->vgCache.vgInfo->hashMethod; + if (dbCache->vgCache.vgInfo->vgHash) { + vgNum = taosHashGetSize(dbCache->vgCache.vgInfo->vgHash); } } diff --git a/source/libs/catalog/src/ctgRemote.c b/source/libs/catalog/src/ctgRemote.c index bfd5057c69..178025704f 100644 --- a/source/libs/catalog/src/ctgRemote.c +++ b/source/libs/catalog/src/ctgRemote.c @@ -431,7 +431,7 @@ int32_t ctgGetIndexInfoFromMnode(SCatalog* pCtg, SRequestConnInfo *pConn, const return TSDB_CODE_SUCCESS; } -int32_t ctgGetTbIndexFromMnode(SCatalog* pCtg, SRequestConnInfo *pConn, SName *name, SArray** out, SCtgTask* pTask) { +int32_t ctgGetTbIndexFromMnode(SCatalog* pCtg, SRequestConnInfo *pConn, SName *name, STableIndex* out, SCtgTask* pTask) { char *msg = NULL; int32_t msgLen = 0; int32_t reqType = TDMT_MND_GET_TABLE_INDEX; @@ -448,10 +448,11 @@ int32_t ctgGetTbIndexFromMnode(SCatalog* pCtg, SRequestConnInfo *pConn, SName *n } if (pTask) { - void* pOut = taosMemoryCalloc(1, POINTER_BYTES); + void* pOut = taosMemoryCalloc(1, sizeof(STableIndex)); if (NULL == pOut) { CTG_ERR_RET(TSDB_CODE_OUT_OF_MEMORY); } + CTG_ERR_RET(ctgUpdateMsgCtx(&pTask->msgCtx, reqType, pOut, (char*)tbFName)); CTG_RET(ctgAsyncSendMsg(pCtg, pConn, pTask, reqType, msg, msgLen)); diff --git a/source/libs/catalog/src/ctgUtil.c b/source/libs/catalog/src/ctgUtil.c index a792c555ee..1cb2b99a67 100644 --- a/source/libs/catalog/src/ctgUtil.c +++ b/source/libs/catalog/src/ctgUtil.c @@ -44,6 +44,16 @@ char *ctgTaskTypeStr(CTG_TASK_TYPE type) { } } +void ctgFreeSTableIndex(void *info) { + if (NULL == info) { + return; + } + + STableIndex *pInfo = (STableIndex *)info; + + taosArrayDestroyEx(pInfo->pIndex, tFreeSTableIndexInfo); +} + void ctgFreeSMetaData(SMetaData* pData) { taosArrayDestroy(pData->pTableMeta); pData->pTableMeta = NULL; @@ -110,25 +120,39 @@ void ctgFreeMetaRent(SCtgRentMgmt *mgmt) { taosMemoryFreeClear(mgmt->slots); } - -void ctgFreeTbMetaCache(SCtgTbMetaCache *cache) { - CTG_LOCK(CTG_WRITE, &cache->stbLock); - if (cache->stbCache) { - int32_t stblNum = taosHashGetSize(cache->stbCache); - taosHashCleanup(cache->stbCache); - cache->stbCache = NULL; - CTG_CACHE_STAT_DEC(stblNum, stblNum); +void ctgFreeStbMetaCache(SCtgDBCache *dbCache) { + if (NULL == dbCache->stbCache) { + return; } - CTG_UNLOCK(CTG_WRITE, &cache->stbLock); - CTG_LOCK(CTG_WRITE, &cache->metaLock); - if (cache->metaCache) { - int32_t tblNum = taosHashGetSize(cache->metaCache); - taosHashCleanup(cache->metaCache); - cache->metaCache = NULL; - CTG_CACHE_STAT_DEC(tblNum, tblNum); + int32_t stblNum = taosHashGetSize(dbCache->stbCache); + taosHashCleanup(dbCache->stbCache); + dbCache->stbCache = NULL; + CTG_CACHE_STAT_DEC(stblNum, stblNum); +} + +void ctgFreeTbCacheImpl(SCtgTbCache *pCache) { + taosMemoryFreeClear(pCache->pMeta); + if (pCache->pIndex) { + taosArrayDestroyEx(pCache->pIndex->pIndex, tFreeSTableIndexInfo); + taosMemoryFreeClear(pCache->pIndex); } - CTG_UNLOCK(CTG_WRITE, &cache->metaLock); +} + +void ctgFreeTbCache(SCtgDBCache *dbCache) { + if (NULL == dbCache->tbCache) { + return; + } + + int32_t tblNum = taosHashGetSize(dbCache->tbCache); + SCtgTbCache *pCache = taosHashIterate(dbCache->tbCache, NULL); + while (NULL != pCache) { + ctgFreeTbCacheImpl(pCache); + pCache = taosHashIterate(dbCache->tbCache, pCache); + } + taosHashCleanup(dbCache->tbCache); + dbCache->tbCache = NULL; + CTG_CACHE_STAT_DEC(tblNum, tblNum); } void ctgFreeVgInfo(SDBVgInfo *vgInfo) { @@ -144,16 +168,18 @@ void ctgFreeVgInfo(SDBVgInfo *vgInfo) { taosMemoryFreeClear(vgInfo); } +void ctgFreeVgInfoCache(SCtgDBCache *dbCache) { + ctgFreeVgInfo(dbCache->vgCache.vgInfo); +} + void ctgFreeDbCache(SCtgDBCache *dbCache) { if (NULL == dbCache) { return; } - CTG_LOCK(CTG_WRITE, &dbCache->vgLock); - ctgFreeVgInfo (dbCache->vgInfo); - CTG_UNLOCK(CTG_WRITE, &dbCache->vgLock); - - ctgFreeTbMetaCache(&dbCache->tbCache); + ctgFreeVgInfoCache(dbCache); + ctgFreeStbMetaCache(dbCache); + ctgFreeTbCache(dbCache); } @@ -167,16 +193,13 @@ void ctgFreeHandle(SCatalog* pCtg) { void *pIter = taosHashIterate(pCtg->dbCache, NULL); while (pIter) { SCtgDBCache *dbCache = pIter; - atomic_store_8(&dbCache->deleted, 1); - ctgFreeDbCache(dbCache); pIter = taosHashIterate(pCtg->dbCache, pIter); } taosHashCleanup(pCtg->dbCache); - CTG_CACHE_STAT_DEC(dbNum, dbNum); } @@ -186,14 +209,12 @@ void ctgFreeHandle(SCatalog* pCtg) { void *pIter = taosHashIterate(pCtg->userCache, NULL); while (pIter) { SCtgUserAuth *userCache = pIter; - ctgFreeSCtgUserAuth(userCache); pIter = taosHashIterate(pCtg->userCache, pIter); } taosHashCleanup(pCtg->userCache); - CTG_CACHE_STAT_DEC(userNum, userNum); } @@ -252,9 +273,9 @@ void ctgFreeMsgCtx(SCtgMsgCtx* pCtx) { break; } case TDMT_MND_GET_TABLE_INDEX: { - SArray** pOut = (SArray**)pCtx->out; + STableIndex* pOut = (STableIndex*)pCtx->out; if (pOut) { - taosArrayDestroyEx(*pOut, tFreeSTableIndexInfo); + taosArrayDestroyEx(pOut->pIndex, tFreeSTableIndexInfo); taosMemoryFreeClear(pCtx->out); } break; @@ -535,9 +556,9 @@ int32_t ctgGetVgInfoFromHashValue(SCatalog *pCtg, SDBVgInfo *dbInfo, const SName } int32_t ctgStbVersionSearchCompare(const void* key1, const void* key2) { - if (*(uint64_t *)key1 < ((SSTableMetaVersion*)key2)->suid) { + if (*(uint64_t *)key1 < ((SSTableVersion*)key2)->suid) { return -1; - } else if (*(uint64_t *)key1 > ((SSTableMetaVersion*)key2)->suid) { + } else if (*(uint64_t *)key1 > ((SSTableVersion*)key2)->suid) { return 1; } else { return 0; @@ -555,9 +576,9 @@ int32_t ctgDbVgVersionSearchCompare(const void* key1, const void* key2) { } int32_t ctgStbVersionSortCompare(const void* key1, const void* key2) { - if (((SSTableMetaVersion*)key1)->suid < ((SSTableMetaVersion*)key2)->suid) { + if (((SSTableVersion*)key1)->suid < ((SSTableVersion*)key2)->suid) { return -1; - } else if (((SSTableMetaVersion*)key1)->suid > ((SSTableMetaVersion*)key2)->suid) { + } else if (((SSTableVersion*)key1)->suid > ((SSTableVersion*)key2)->suid) { return 1; } else { return 0; @@ -640,6 +661,27 @@ int32_t ctgCloneMetaOutput(STableMetaOutput *output, STableMetaOutput **pOutput) return TSDB_CODE_SUCCESS; } +int32_t ctgCloneTableIndex(SArray* pIndex, SArray** pRes) { + if (NULL == pIndex) { + *pRes = NULL; + return TSDB_CODE_SUCCESS; + } + + int32_t num = taosArrayGetSize(pIndex); + *pRes = taosArrayInit(num, sizeof(STableIndexInfo)); + if (NULL == *pRes) { + CTG_ERR_RET(TSDB_CODE_OUT_OF_MEMORY); + } + + for (int32_t i = 0; i < num; ++i) { + STableIndexInfo *pInfo = taosArrayGet(pIndex, i); + taosArrayPush(*pRes, pInfo); + } + + return TSDB_CODE_SUCCESS; +} + + int32_t ctgUpdateSendTargetInfo(SMsgSendInfo *pMsgSendInfo, int32_t msgType, SCtgTask* pTask) { if (msgType == TDMT_VND_TABLE_META) { SCtgTbMetaCtx* ctx = (SCtgTbMetaCtx*)pTask->taskCtx; diff --git a/source/libs/catalog/test/catalogTests.cpp b/source/libs/catalog/test/catalogTests.cpp index 04dbc26774..aab341544f 100644 --- a/source/libs/catalog/test/catalogTests.cpp +++ b/source/libs/catalog/test/catalogTests.cpp @@ -895,9 +895,9 @@ void *ctgTestSetCtableMetaThread(void *param) { output = (STableMetaOutput *)taosMemoryMalloc(sizeof(STableMetaOutput)); ctgTestBuildCTableMetaOutput(output); - SCtgUpdateTblMsg *msg = (SCtgUpdateTblMsg *)taosMemoryMalloc(sizeof(SCtgUpdateTblMsg)); + SCtgUpdateTbMetaMsg *msg = (SCtgUpdateTbMetaMsg *)taosMemoryMalloc(sizeof(SCtgUpdateTbMetaMsg)); msg->pCtg = pCtg; - msg->output = output; + msg->pMeta = output; operation.data = msg; code = ctgOpUpdateTbMeta(&operation); @@ -989,7 +989,7 @@ TEST(tableMeta, normalTable) { ASSERT_EQ(tableMeta->tableInfo.rowSize, 12); SDbVgVersion *dbs = NULL; - SSTableMetaVersion *stb = NULL; + SSTableVersion *stb = NULL; uint32_t dbNum = 0, stbNum = 0, allDbNum = 0, allStbNum = 0; int32_t i = 0; while (i < 5) { @@ -1098,7 +1098,7 @@ TEST(tableMeta, childTableCase) { ASSERT_EQ(tableMeta->tableInfo.rowSize, 12); SDbVgVersion *dbs = NULL; - SSTableMetaVersion *stb = NULL; + SSTableVersion *stb = NULL; uint32_t dbNum = 0, stbNum = 0, allDbNum = 0, allStbNum = 0; int32_t i = 0; while (i < 5) { @@ -1220,7 +1220,7 @@ TEST(tableMeta, superTableCase) { ASSERT_EQ(tableMeta->tableInfo.rowSize, 12); SDbVgVersion *dbs = NULL; - SSTableMetaVersion *stb = NULL; + SSTableVersion *stb = NULL; uint32_t dbNum = 0, stbNum = 0, allDbNum = 0, allStbNum = 0; int32_t i = 0; while (i < 5) { @@ -2299,7 +2299,7 @@ TEST(rentTest, allRent) { SArray *vgList = NULL; ctgTestStop = false; SDbVgVersion *dbs = NULL; - SSTableMetaVersion *stable = NULL; + SSTableVersion *stable = NULL; uint32_t num = 0; ctgTestInitLogFile(); diff --git a/source/libs/executor/inc/executorInt.h b/source/libs/executor/inc/executorInt.h index 88f308710e..df54161720 100644 --- a/source/libs/executor/inc/executorInt.h +++ b/source/libs/executor/inc/executorInt.h @@ -27,10 +27,7 @@ typedef struct { int32_t bytes; } SGroupKeys, SStateKeys; -int32_t initGroupOptrInfo(SArray** pGroupColVals, int32_t* keyLen, char** keyBuf, const SArray* pGroupColList); uint64_t calcGroupId(char* pData, int32_t len); -void recordNewGroupKeys(SArray* pGroupCols, SArray* pGroupColVals, SSDataBlock* pBlock, int32_t rowIndex); -int32_t buildGroupKeys(void* pKey, const SArray* pGroupColVals); #ifdef __cplusplus } #endif diff --git a/source/libs/executor/inc/executorimpl.h b/source/libs/executor/inc/executorimpl.h index 94c6512e77..ab60acab53 100644 --- a/source/libs/executor/inc/executorimpl.h +++ b/source/libs/executor/inc/executorimpl.h @@ -336,12 +336,6 @@ typedef struct STableScanInfo { int32_t dataBlockLoadFlag; SInterval interval; // if the upstream is an interval operator, the interval info is also kept here to get the time window to check if current data block needs to be loaded. - SArray* pGroupCols; - SArray* pGroupColVals; // current group column values, SArray - char* keyBuf; // group by keys for hash - int32_t groupKeyLen; // total group by column width - SHashObj* pGroupSet; // quick locate the window object for each result - SSampleExecInfo sample; // sample execution info int32_t curTWinIdx; } STableScanInfo; @@ -789,7 +783,7 @@ SResultRow* doSetResultOutBufByKey(SDiskbasedBuf* pResultBuf, SResultRowInfo* pR SOperatorInfo* createExchangeOperatorInfo(void* pTransporter, SExchangePhysiNode* pExNode, SExecTaskInfo* pTaskInfo); -SOperatorInfo* createTableScanOperatorInfo(STableScanPhysiNode* pTableScanNode, tsdbReaderT pDataReader, SReadHandle* pHandle, SArray* groupKyes, SExecTaskInfo* pTaskInfo); +SOperatorInfo* createTableScanOperatorInfo(STableScanPhysiNode* pTableScanNode, tsdbReaderT pDataReader, SReadHandle* pHandle, SExecTaskInfo* pTaskInfo); SOperatorInfo* createTagScanOperatorInfo(SReadHandle* pReadHandle, STagScanPhysiNode* pPhyNode, STableListInfo* pTableListInfo, SExecTaskInfo* pTaskInfo); SOperatorInfo* createSysTableScanOperatorInfo(void* readHandle, SSystemTableScanPhysiNode *pScanPhyNode, SExecTaskInfo* pTaskInfo); @@ -827,7 +821,7 @@ SOperatorInfo* createGroupOperatorInfo(SOperatorInfo* downstream, SExprInfo* pEx SExprInfo* pScalarExprInfo, int32_t numOfScalarExpr, SExecTaskInfo* pTaskInfo); SOperatorInfo* createDataBlockInfoScanOperator(void* dataReader, SExecTaskInfo* pTaskInfo); -SOperatorInfo* createStreamScanOperatorInfo(void* pDataReader, SReadHandle* pHandle, SArray* pTableIdList, +SOperatorInfo* createStreamScanOperatorInfo(void* pDataReader, SReadHandle* pHandle, STableScanPhysiNode* pTableScanNode, SExecTaskInfo* pTaskInfo, STimeWindowAggSupp* pTwSup); diff --git a/source/libs/executor/src/executorimpl.c b/source/libs/executor/src/executorimpl.c index 474b700868..99a9dcb6af 100644 --- a/source/libs/executor/src/executorimpl.c +++ b/source/libs/executor/src/executorimpl.c @@ -4558,7 +4558,6 @@ static tsdbReaderT doCreateDataReader(STableScanPhysiNode* pTableScanNode, SRead static int32_t getTableList(void* metaHandle, int32_t tableType, uint64_t tableUid, STableListInfo* pListInfo, SNode* pTagCond); -static SArray* extractTableIdList(const STableListInfo* pTableGroupInfo); static SArray* extractColumnInfo(SNodeList* pNodeList); static SArray* createSortInfo(SNodeList* pNodeList); @@ -4592,6 +4591,85 @@ int32_t extractTableSchemaVersion(SReadHandle* pHandle, uint64_t uid, SExecTaskI return TSDB_CODE_SUCCESS; } +int32_t generateGroupIdMap(STableListInfo* pTableListInfo, SReadHandle* pHandle, SArray* groupKey){ + if(groupKey == NULL) { + return TDB_CODE_SUCCESS; + } + + pTableListInfo->map = taosHashInit(32, taosGetDefaultHashFunction(TSDB_DATA_TYPE_BINARY), false, HASH_NO_LOCK); + if (pTableListInfo->map == NULL) { + return TSDB_CODE_OUT_OF_MEMORY; + } + int32_t keyLen = 0; + void *keyBuf = NULL; + int32_t numOfGroupCols = taosArrayGetSize(groupKey); + for (int32_t j = 0; j < numOfGroupCols; ++j) { + SColumn* pCol = taosArrayGet(groupKey, j); + keyLen += pCol->bytes; // actual data + null_flag + } + + int32_t nullFlagSize = sizeof(int8_t) * numOfGroupCols; + keyLen += nullFlagSize; + + keyBuf = taosMemoryCalloc(1, keyLen); + if (keyBuf == NULL) { + return TSDB_CODE_OUT_OF_MEMORY; + } + + for(int32_t i = 0; i < taosArrayGetSize(pTableListInfo->pTableList); i++){ + STableKeyInfo *info = taosArrayGet(pTableListInfo->pTableList, i); + SMetaReader mr = {0}; + metaReaderInit(&mr, pHandle->meta, 0); + metaGetTableEntryByUid(&mr, info->uid); + + char* isNull = (char*)keyBuf; + char* pStart = (char*)keyBuf + sizeof(int8_t) * numOfGroupCols; + for (int32_t j = 0; j < numOfGroupCols; ++j) { + SColumn* pCol = taosArrayGet(groupKey, j); + + if(strcmp(pCol->name, "tbname") == 0){ + isNull[i] = 0; + memcpy(pStart, mr.me.name, strlen(mr.me.name)); + pStart += strlen(mr.me.name); + }else{ + STagVal tagVal = {0}; + tagVal.cid = pCol->colId; + const char* p = metaGetTableTagVal(&mr.me, pCol->type, &tagVal); + if(p == NULL){ + isNull[j] = 1; + continue; + } + isNull[i] = 0; + if (pCol->type == TSDB_DATA_TYPE_JSON) { +// int32_t dataLen = getJsonValueLen(pkey->pData); +// memcpy(pStart, (pkey->pData), dataLen); +// pStart += dataLen; + } else if (IS_VAR_DATA_TYPE(pCol->type)) { + memcpy(pStart, tagVal.pData, tagVal.nData); + pStart += tagVal.nData; + ASSERT(tagVal.nData <= pCol->bytes); + } else { + memcpy(pStart, &(tagVal.i64), pCol->bytes); + pStart += pCol->bytes; + } + } + } + + int32_t len = (int32_t) (pStart - (char*)keyBuf); + uint64_t* groupId = taosHashGet(pTableListInfo->map, keyBuf, len); + if (groupId) { + taosHashPut(pTableListInfo->map, &(info->uid), sizeof(uint64_t), groupId, sizeof(uint64_t)); + } else { + uint64_t tmpId = calcGroupId(keyBuf, len); + taosHashPut(pTableListInfo->map, &(info->uid), sizeof(uint64_t), &tmpId, sizeof(uint64_t)); + } + + metaReaderClear(&mr); + } + taosMemoryFree(keyBuf); + return TDB_CODE_SUCCESS; +} + SOperatorInfo* createOperatorTree(SPhysiNode* pPhyNode, SExecTaskInfo* pTaskInfo, SReadHandle* pHandle, uint64_t queryId, uint64_t taskId, STableListInfo* pTableListInfo, SNode* pTagCond) { int32_t type = nodeType(pPhyNode); @@ -4605,15 +4683,23 @@ SOperatorInfo* createOperatorTree(SPhysiNode* pPhyNode, SExecTaskInfo* pTaskInfo if (pDataReader == NULL && terrno != 0) { return NULL; } - SArray* groupKyes = extractPartitionColInfo(pTableScanNode->pPartitionKeys); + int32_t code = extractTableSchemaVersion(pHandle, pTableScanNode->scan.uid, pTaskInfo); if (code) { tsdbCleanupReadHandle(pDataReader); return NULL; } - + + SArray* groupKeys = extractPartitionColInfo(pTableScanNode->pPartitionKeys); + code = generateGroupIdMap(pTableListInfo, pHandle, groupKeys); //todo for json + taosArrayDestroy(groupKeys); + if (code){ + tsdbCleanupReadHandle(pDataReader); + return NULL; + } + SOperatorInfo* pOperator = - createTableScanOperatorInfo(pTableScanNode, pDataReader, pHandle, groupKyes, pTaskInfo); + createTableScanOperatorInfo(pTableScanNode, pDataReader, pHandle, pTaskInfo); STableScanInfo* pScanInfo = pOperator->info; pTaskInfo->cost.pRecoder = &pScanInfo->readRecorder; @@ -4639,12 +4725,18 @@ SOperatorInfo* createOperatorTree(SPhysiNode* pPhyNode, SExecTaskInfo* pTaskInfo } else { qDebug("%s pDataReader is not NULL", GET_TASKID(pTaskInfo)); } - SArray* tableIdList = extractTableIdList(pTableListInfo); + + SArray* groupKeys = extractPartitionColInfo(pTableScanNode->pPartitionKeys); + int32_t code = generateGroupIdMap(pTableListInfo, pHandle, groupKeys); //todo for json + taosArrayDestroy(groupKeys); + if (code){ + tsdbCleanupReadHandle(pDataReader); + return NULL; + } SOperatorInfo* pOperator = - createStreamScanOperatorInfo(pDataReader, pHandle, tableIdList, pTableScanNode, pTaskInfo, &twSup); + createStreamScanOperatorInfo(pDataReader, pHandle, pTableScanNode, pTaskInfo, &twSup); - taosArrayDestroy(tableIdList); return pOperator; } else if (QUERY_NODE_PHYSICAL_PLAN_SYSTABLE_SCAN == type) { SSystemTableScanPhysiNode* pSysScanPhyNode = (SSystemTableScanPhysiNode*)pPhyNode; @@ -4969,6 +5061,7 @@ SArray* extractColumnInfo(SNodeList* pNodeList) { } SArray* extractPartitionColInfo(SNodeList* pNodeList) { + if(!pNodeList) return NULL; size_t numOfCols = LIST_LENGTH(pNodeList); SArray* pList = taosArrayInit(numOfCols, sizeof(SColumn)); if (pList == NULL) { @@ -5073,7 +5166,9 @@ int32_t getTableList(void* metaHandle, int32_t tableType, uint64_t tableUid, STa SArray* res = taosArrayInit(8, sizeof(uint64_t)); code = doFilterTag(pTagCond, &metaArg, res); - if (code != TSDB_CODE_SUCCESS) { + if (code == TSDB_CODE_INDEX_REBUILDING){ // todo + // doFilter(); + } else if (code != TSDB_CODE_SUCCESS) { qError("failed to get tableIds, reason: %s, suid: %" PRIu64 "", tstrerror(code), tableUid); taosArrayDestroy(res); terrno = code; @@ -5081,6 +5176,7 @@ int32_t getTableList(void* metaHandle, int32_t tableType, uint64_t tableUid, STa } else { qDebug("sucess to get tableIds, size: %d, suid: %" PRIu64 "", (int)taosArrayGetSize(res), tableUid); } + for (int i = 0; i < taosArrayGetSize(res); i++) { STableKeyInfo info = {.lastKey = TSKEY_INITIAL_VAL, .uid = *(uint64_t*)taosArrayGet(res, i)}; taosArrayPush(pListInfo->pTableList, &info); @@ -5097,18 +5193,6 @@ int32_t getTableList(void* metaHandle, int32_t tableType, uint64_t tableUid, STa return code; } -SArray* extractTableIdList(const STableListInfo* pTableGroupInfo) { - SArray* tableIdList = taosArrayInit(4, sizeof(uint64_t)); - - // Transfer the Array of STableKeyInfo into uid list. - for (int32_t i = 0; i < taosArrayGetSize(pTableGroupInfo->pTableList); ++i) { - STableKeyInfo* pkeyInfo = taosArrayGet(pTableGroupInfo->pTableList, i); - taosArrayPush(tableIdList, &pkeyInfo->uid); - } - - return tableIdList; -} - tsdbReaderT doCreateDataReader(STableScanPhysiNode* pTableScanNode, SReadHandle* pHandle, STableListInfo* pTableListInfo, uint64_t queryId, uint64_t taskId, SNode* pTagCond) { int32_t code = diff --git a/source/libs/executor/src/groupoperator.c b/source/libs/executor/src/groupoperator.c index f663f5798a..a48b4080a3 100644 --- a/source/libs/executor/src/groupoperator.c +++ b/source/libs/executor/src/groupoperator.c @@ -37,7 +37,7 @@ static void destroyGroupOperatorInfo(void* param, int32_t numOfOutput) { taosArrayDestroy(pInfo->pGroupColVals); } -int32_t initGroupOptrInfo(SArray** pGroupColVals, int32_t* keyLen, char** keyBuf, const SArray* pGroupColList) { +static int32_t initGroupOptrInfo(SArray** pGroupColVals, int32_t* keyLen, char** keyBuf, const SArray* pGroupColList) { *pGroupColVals = taosArrayInit(4, sizeof(SGroupKeys)); if ((*pGroupColVals) == NULL) { return TSDB_CODE_OUT_OF_MEMORY; @@ -118,7 +118,7 @@ static bool groupKeyCompare(SArray* pGroupCols, SArray* pGroupColVals, SSDataBlo return true; } -void recordNewGroupKeys(SArray* pGroupCols, SArray* pGroupColVals, SSDataBlock* pBlock, int32_t rowIndex) { +static void recordNewGroupKeys(SArray* pGroupCols, SArray* pGroupColVals, SSDataBlock* pBlock, int32_t rowIndex) { SColumnDataAgg* pColAgg = NULL; size_t numOfGroupCols = taosArrayGetSize(pGroupCols); @@ -150,7 +150,7 @@ void recordNewGroupKeys(SArray* pGroupCols, SArray* pGroupColVals, SSDataBlock* } } -int32_t buildGroupKeys(void* pKey, const SArray* pGroupColVals) { +static int32_t buildGroupKeys(void* pKey, const SArray* pGroupColVals) { ASSERT(pKey != NULL); size_t numOfGroupCols = taosArrayGetSize(pGroupColVals); diff --git a/source/libs/executor/src/scanoperator.c b/source/libs/executor/src/scanoperator.c index 8fb4878cd9..d30e4ef6db 100644 --- a/source/libs/executor/src/scanoperator.c +++ b/source/libs/executor/src/scanoperator.c @@ -391,22 +391,16 @@ static SSDataBlock* doTableScanImpl(SOperatorInfo* pOperator) { longjmp(pOperator->pTaskInfo->env, code); } - recordNewGroupKeys(pTableScanInfo->pGroupCols, pTableScanInfo->pGroupColVals, pBlock, 0); - int32_t len = buildGroupKeys(pTableScanInfo->keyBuf, pTableScanInfo->pGroupColVals); - - uint64_t* groupId = taosHashGet(pTableScanInfo->pGroupSet, pTableScanInfo->keyBuf, len); - if (groupId) { - pBlock->info.groupId = *groupId; - } else if (len != 0) { - pBlock->info.groupId = calcGroupId(pTableScanInfo->keyBuf, len); - taosHashPut(pTableScanInfo->pGroupSet, pTableScanInfo->keyBuf, len, &pBlock->info.groupId, sizeof(uint64_t)); - } - // current block is filter out according to filter condition, continue load the next block if (status == FUNC_DATA_REQUIRED_FILTEROUT || pBlock->info.rows == 0) { continue; } + uint64_t* groupId = taosHashGet(pOperator->pTaskInfo->tableqinfoList.map, &pBlock->info.uid, sizeof(int64_t)); + if (groupId) { + pBlock->info.groupId = *groupId; + } + pOperator->resultInfo.totalRows = pTableScanInfo->readRecorder.totalRows; pTableScanInfo->readRecorder.elapsedTime += (taosGetTimestampUs() - st) / 1000.0; @@ -530,21 +524,13 @@ static void destroyTableScanOperatorInfo(void* param, int32_t numOfOutput) { tsdbCleanupReadHandle(pTableScanInfo->dataReader); - taosArrayDestroy(pTableScanInfo->pGroupCols); - for (int i = 0; i < taosArrayGetSize(pTableScanInfo->pGroupColVals); i++) { - SGroupKeys key = *(SGroupKeys*)taosArrayGet(pTableScanInfo->pGroupColVals, i); - taosMemoryFree(key.pData); - } - taosArrayDestroy(pTableScanInfo->pGroupColVals); - taosMemoryFree(pTableScanInfo->keyBuf); - taosHashCleanup(pTableScanInfo->pGroupSet); if (pTableScanInfo->pColMatchInfo != NULL) { taosArrayDestroy(pTableScanInfo->pColMatchInfo); } } SOperatorInfo* createTableScanOperatorInfo(STableScanPhysiNode* pTableScanNode, tsdbReaderT pDataReader, - SReadHandle* readHandle, SArray* groupKyes, SExecTaskInfo* pTaskInfo) { + SReadHandle* readHandle, SExecTaskInfo* pTaskInfo) { STableScanInfo* pInfo = taosMemoryCalloc(1, sizeof(STableScanInfo)); SOperatorInfo* pOperator = taosMemoryCalloc(1, sizeof(SOperatorInfo)); if (pInfo == NULL || pOperator == NULL) { @@ -591,18 +577,6 @@ SOperatorInfo* createTableScanOperatorInfo(STableScanPhysiNode* pTableScanNode, pOperator->numOfExprs = numOfCols; pOperator->pTaskInfo = pTaskInfo; - // for table group - pInfo->pGroupCols = groupKyes; - _hash_fn_t hashFn = taosGetDefaultHashFunction(TSDB_DATA_TYPE_BINARY); - pInfo->pGroupSet = taosHashInit(100, hashFn, false, HASH_NO_LOCK); - if (pInfo->pGroupSet == NULL) { - goto _error; - } - code = initGroupOptrInfo(&pInfo->pGroupColVals, &pInfo->groupKeyLen, &pInfo->keyBuf, groupKyes); - if (code != TSDB_CODE_SUCCESS) { - goto _error; - } - pOperator->fpSet = createOperatorFpSet(operatorDummyOpenFn, doTableScan, NULL, NULL, destroyTableScanOperatorInfo, NULL, NULL, getTableScannerExecInfo); @@ -912,6 +886,11 @@ static SSDataBlock* doStreamBlockScan(SOperatorInfo* pOperator) { pInfo->pRes->info.groupId = groupId; } + uint64_t* groupIdPre = taosHashGet(pOperator->pTaskInfo->tableqinfoList.map, &uid, sizeof(int64_t)); + if (groupIdPre) { + pInfo->pRes->info.groupId = *groupIdPre; + } + for (int32_t i = 0; i < taosArrayGetSize(pInfo->pColMatchInfo); ++i) { SColMatchInfo* pColMatchInfo = taosArrayGet(pInfo->pColMatchInfo, i); if (!pColMatchInfo->output) { @@ -979,11 +958,24 @@ static SSDataBlock* doStreamBlockScan(SOperatorInfo* pOperator) { } } -SOperatorInfo* createStreamScanOperatorInfo(void* pDataReader, SReadHandle* pHandle, SArray* pTableIdList, +static SArray* extractTableIdList(const STableListInfo* pTableGroupInfo) { + SArray* tableIdList = taosArrayInit(4, sizeof(uint64_t)); + + // Transfer the Array of STableKeyInfo into uid list. + for (int32_t i = 0; i < taosArrayGetSize(pTableGroupInfo->pTableList); ++i) { + STableKeyInfo* pkeyInfo = taosArrayGet(pTableGroupInfo->pTableList, i); + taosArrayPush(tableIdList, &pkeyInfo->uid); + } + + return tableIdList; +} + +SOperatorInfo* createStreamScanOperatorInfo(void* pDataReader, SReadHandle* pHandle, STableScanPhysiNode* pTableScanNode, SExecTaskInfo* pTaskInfo, STimeWindowAggSupp* pTwSup) { SStreamBlockScanInfo* pInfo = taosMemoryCalloc(1, sizeof(SStreamBlockScanInfo)); SOperatorInfo* pOperator = taosMemoryCalloc(1, sizeof(SOperatorInfo)); + if (pInfo == NULL || pOperator == NULL) { terrno = TSDB_CODE_QRY_OUT_OF_MEMORY; goto _error; @@ -992,7 +984,7 @@ SOperatorInfo* createStreamScanOperatorInfo(void* pDataReader, SReadHandle* pHan SScanPhysiNode* pScanPhyNode = &pTableScanNode->scan; SDataBlockDescNode* pDescNode = pScanPhyNode->node.pOutputDataBlockDesc; - SOperatorInfo* pTableScanDummy = createTableScanOperatorInfo(pTableScanNode, pDataReader, pHandle, NULL, pTaskInfo); + SOperatorInfo* pTableScanDummy = createTableScanOperatorInfo(pTableScanNode, pDataReader, pHandle, pTaskInfo); STableScanInfo* pSTInfo = (STableScanInfo*)pTableScanDummy->info; @@ -1014,10 +1006,13 @@ SOperatorInfo* createStreamScanOperatorInfo(void* pDataReader, SReadHandle* pHan // set the extract column id to streamHandle tqReadHandleSetColIdList((STqReadHandle*)pHandle->reader, pColIds); - int32_t code = tqReadHandleSetTbUidList(pHandle->reader, pTableIdList); + SArray* tableIdList = extractTableIdList(&pTaskInfo->tableqinfoList); + int32_t code = tqReadHandleSetTbUidList(pHandle->reader, tableIdList); if (code != 0) { + taosArrayDestroy(tableIdList); goto _error; } + taosArrayDestroy(tableIdList); pInfo->pBlockLists = taosArrayInit(4, POINTER_BYTES); if (pInfo->pBlockLists == NULL) { diff --git a/source/libs/function/inc/builtinsimpl.h b/source/libs/function/inc/builtinsimpl.h index 7db1396603..1bb30934c6 100644 --- a/source/libs/function/inc/builtinsimpl.h +++ b/source/libs/function/inc/builtinsimpl.h @@ -55,9 +55,12 @@ int32_t maxCombine(SqlFunctionCtx* pDestCtx, SqlFunctionCtx* pSourceCtx); bool getAvgFuncEnv(struct SFunctionNode* pFunc, SFuncExecEnv* pEnv); bool avgFunctionSetup(SqlFunctionCtx *pCtx, SResultRowEntryInfo* pResultInfo); int32_t avgFunction(SqlFunctionCtx* pCtx); +int32_t avgFunctionMerge(SqlFunctionCtx* pCtx); int32_t avgFinalize(SqlFunctionCtx* pCtx, SSDataBlock* pBlock); +int32_t avgPartialFinalize(SqlFunctionCtx* pCtx, SSDataBlock* pBlock); int32_t avgInvertFunction(SqlFunctionCtx* pCtx); int32_t avgCombine(SqlFunctionCtx* pDestCtx, SqlFunctionCtx* pSourceCtx); +int32_t getAvgInfoSize(); bool getStddevFuncEnv(struct SFunctionNode* pFunc, SFuncExecEnv* pEnv); bool stddevFunctionSetup(SqlFunctionCtx *pCtx, SResultRowEntryInfo* pResultInfo); diff --git a/source/libs/function/src/builtins.c b/source/libs/function/src/builtins.c index 1c3f274ee9..4dae296f05 100644 --- a/source/libs/function/src/builtins.c +++ b/source/libs/function/src/builtins.c @@ -155,6 +155,35 @@ static int32_t translateSum(SFunctionNode* pFunc, char* pErrBuf, int32_t len) { return TSDB_CODE_SUCCESS; } +static int32_t translateAvgPartial(SFunctionNode* pFunc, char* pErrBuf, int32_t len) { + if (1 != LIST_LENGTH(pFunc->pParameterList)) { + return invaildFuncParaNumErrMsg(pErrBuf, len, pFunc->functionName); + } + + uint8_t paraType = ((SExprNode*)nodesListGetNode(pFunc->pParameterList, 0))->resType.type; + if (!IS_NUMERIC_TYPE(paraType) && !IS_NULL_TYPE(paraType)) { + return invaildFuncParaTypeErrMsg(pErrBuf, len, pFunc->functionName); + } + + pFunc->node.resType = (SDataType){.bytes = getAvgInfoSize() + VARSTR_HEADER_SIZE, .type = TSDB_DATA_TYPE_BINARY}; + return TSDB_CODE_SUCCESS; +} + +static int32_t translateAvgMerge(SFunctionNode* pFunc, char* pErrBuf, int32_t len) { + if (1 != LIST_LENGTH(pFunc->pParameterList)) { + return invaildFuncParaNumErrMsg(pErrBuf, len, pFunc->functionName); + } + + uint8_t paraType = ((SExprNode*)nodesListGetNode(pFunc->pParameterList, 0))->resType.type; + if (TSDB_DATA_TYPE_BINARY != paraType) { + return invaildFuncParaTypeErrMsg(pErrBuf, len, pFunc->functionName); + } + + pFunc->node.resType = (SDataType){.bytes = tDataTypes[TSDB_DATA_TYPE_DOUBLE].bytes, .type = TSDB_DATA_TYPE_DOUBLE}; + + return TSDB_CODE_SUCCESS; +} + static int32_t translateWduration(SFunctionNode* pFunc, char* pErrBuf, int32_t len) { // pseudo column do not need to check parameters pFunc->node.resType = (SDataType){.bytes = sizeof(int64_t), .type = TSDB_DATA_TYPE_BIGINT}; @@ -1517,6 +1546,32 @@ const SBuiltinFuncDefinition funcMgtBuiltins[] = { .finalizeFunc = avgFinalize, .invertFunc = avgInvertFunction, .combineFunc = avgCombine, + .pPartialFunc = "_avg_partial", + .pMergeFunc = "_avg_merge" + }, + { + .name = "_avg_partial", + .type = FUNCTION_TYPE_AVG_PARTIAL, + .classification = FUNC_MGT_AGG_FUNC, + .translateFunc = translateAvgPartial, + .getEnvFunc = getAvgFuncEnv, + .initFunc = avgFunctionSetup, + .processFunc = avgFunction, + .finalizeFunc = avgPartialFinalize, + .invertFunc = avgInvertFunction, + .combineFunc = avgCombine, + }, + { + .name = "_avg_merge", + .type = FUNCTION_TYPE_AVG_MERGE, + .classification = FUNC_MGT_AGG_FUNC, + .translateFunc = translateAvgMerge, + .getEnvFunc = getAvgFuncEnv, + .initFunc = avgFunctionSetup, + .processFunc = avgFunctionMerge, + .finalizeFunc = avgFinalize, + .invertFunc = avgInvertFunction, + .combineFunc = avgCombine, }, { .name = "percentile", diff --git a/source/libs/function/src/builtinsimpl.c b/source/libs/function/src/builtinsimpl.c index b5009b7d41..5ee9dbdee6 100644 --- a/source/libs/function/src/builtinsimpl.c +++ b/source/libs/function/src/builtinsimpl.c @@ -51,6 +51,7 @@ typedef struct SAvgRes { double result; SSumRes sum; int64_t count; + int16_t type; // store the original input type, used in merge function } SAvgRes; typedef struct STuplePos { @@ -624,6 +625,8 @@ bool getSumFuncEnv(SFunctionNode* UNUSED_PARAM(pFunc), SFuncExecEnv* pEnv) { return true; } +int32_t getAvgInfoSize() { return (int32_t)sizeof(SAvgRes); } + bool getAvgFuncEnv(SFunctionNode* UNUSED_PARAM(pFunc), SFuncExecEnv* pEnv) { pEnv->calcMemSize = sizeof(SAvgRes); return true; @@ -642,11 +645,12 @@ bool avgFunctionSetup(SqlFunctionCtx* pCtx, SResultRowEntryInfo* pResultInfo) { int32_t avgFunction(SqlFunctionCtx* pCtx) { int32_t numOfElem = 0; - // Only the pre-computing information loaded and actual data does not loaded SInputColumnInfoData* pInput = &pCtx->input; + SColumnDataAgg* pAgg = pInput->pColumnDataAgg[0]; int32_t type = pInput->pData[0]->info.type; SAvgRes* pAvgRes = GET_ROWCELL_INTERBUF(GET_RES_INFO(pCtx)); + pAvgRes->type = type; // computing based on the true data block SColumnInfoData* pCol = pInput->pData[0]; @@ -660,95 +664,107 @@ int32_t avgFunction(SqlFunctionCtx* pCtx) { goto _avg_over; } - switch (type) { - case TSDB_DATA_TYPE_TINYINT: { - int8_t* plist = (int8_t*)pCol->pData; - for (int32_t i = start; i < numOfRows + pInput->startRowIndex; ++i) { - if (pCol->hasNull && colDataIsNull_f(pCol->nullbitmap, i)) { - continue; + if (pInput->colDataAggIsSet) { + numOfElem = numOfRows - pAgg->numOfNull; + ASSERT(numOfElem >= 0); + + pAvgRes->count += numOfElem; + if (IS_INTEGER_TYPE(type)) { + pAvgRes->sum.isum += pAgg->sum; + } else if (IS_FLOAT_TYPE(type)) { + pAvgRes->sum.dsum += GET_DOUBLE_VAL((const char*)&(pAgg->sum)); + } + } else { // computing based on the true data block + switch (type) { + case TSDB_DATA_TYPE_TINYINT: { + int8_t* plist = (int8_t*)pCol->pData; + for (int32_t i = start; i < numOfRows + pInput->startRowIndex; ++i) { + if (pCol->hasNull && colDataIsNull_f(pCol->nullbitmap, i)) { + continue; + } + + numOfElem += 1; + pAvgRes->count += 1; + pAvgRes->sum.isum += plist[i]; } - numOfElem += 1; - pAvgRes->count += 1; - pAvgRes->sum.isum += plist[i]; + break; } - break; - } + case TSDB_DATA_TYPE_SMALLINT: { + int16_t* plist = (int16_t*)pCol->pData; + for (int32_t i = start; i < numOfRows + pInput->startRowIndex; ++i) { + if (pCol->hasNull && colDataIsNull_f(pCol->nullbitmap, i)) { + continue; + } - case TSDB_DATA_TYPE_SMALLINT: { - int16_t* plist = (int16_t*)pCol->pData; - for (int32_t i = start; i < numOfRows + pInput->startRowIndex; ++i) { - if (pCol->hasNull && colDataIsNull_f(pCol->nullbitmap, i)) { - continue; + numOfElem += 1; + pAvgRes->count += 1; + pAvgRes->sum.isum += plist[i]; + } + break; + } + + case TSDB_DATA_TYPE_INT: { + int32_t* plist = (int32_t*)pCol->pData; + for (int32_t i = start; i < numOfRows + pInput->startRowIndex; ++i) { + if (pCol->hasNull && colDataIsNull_f(pCol->nullbitmap, i)) { + continue; + } + + numOfElem += 1; + pAvgRes->count += 1; + pAvgRes->sum.isum += plist[i]; } - numOfElem += 1; - pAvgRes->count += 1; - pAvgRes->sum.isum += plist[i]; + break; } - break; - } - case TSDB_DATA_TYPE_INT: { - int32_t* plist = (int32_t*)pCol->pData; - for (int32_t i = start; i < numOfRows + pInput->startRowIndex; ++i) { - if (pCol->hasNull && colDataIsNull_f(pCol->nullbitmap, i)) { - continue; + case TSDB_DATA_TYPE_BIGINT: { + int64_t* plist = (int64_t*)pCol->pData; + for (int32_t i = start; i < numOfRows + pInput->startRowIndex; ++i) { + if (pCol->hasNull && colDataIsNull_f(pCol->nullbitmap, i)) { + continue; + } + + numOfElem += 1; + pAvgRes->count += 1; + pAvgRes->sum.isum += plist[i]; } - - numOfElem += 1; - pAvgRes->count += 1; - pAvgRes->sum.isum += plist[i]; + break; } - break; - } + case TSDB_DATA_TYPE_FLOAT: { + float* plist = (float*)pCol->pData; + for (int32_t i = start; i < numOfRows + pInput->startRowIndex; ++i) { + if (pCol->hasNull && colDataIsNull_f(pCol->nullbitmap, i)) { + continue; + } - case TSDB_DATA_TYPE_BIGINT: { - int64_t* plist = (int64_t*)pCol->pData; - for (int32_t i = start; i < numOfRows + pInput->startRowIndex; ++i) { - if (pCol->hasNull && colDataIsNull_f(pCol->nullbitmap, i)) { - continue; + numOfElem += 1; + pAvgRes->count += 1; + pAvgRes->sum.dsum += plist[i]; } - - numOfElem += 1; - pAvgRes->count += 1; - pAvgRes->sum.isum += plist[i]; + break; } - break; - } - case TSDB_DATA_TYPE_FLOAT: { - float* plist = (float*)pCol->pData; - for (int32_t i = start; i < numOfRows + pInput->startRowIndex; ++i) { - if (pCol->hasNull && colDataIsNull_f(pCol->nullbitmap, i)) { - continue; + case TSDB_DATA_TYPE_DOUBLE: { + double* plist = (double*)pCol->pData; + for (int32_t i = start; i < numOfRows + pInput->startRowIndex; ++i) { + if (pCol->hasNull && colDataIsNull_f(pCol->nullbitmap, i)) { + continue; + } + + numOfElem += 1; + pAvgRes->count += 1; + pAvgRes->sum.dsum += plist[i]; } - - numOfElem += 1; - pAvgRes->count += 1; - pAvgRes->sum.dsum += plist[i]; + break; } - break; + + default: + break; } - - case TSDB_DATA_TYPE_DOUBLE: { - double* plist = (double*)pCol->pData; - for (int32_t i = start; i < numOfRows + pInput->startRowIndex; ++i) { - if (pCol->hasNull && colDataIsNull_f(pCol->nullbitmap, i)) { - continue; - } - - numOfElem += 1; - pAvgRes->count += 1; - pAvgRes->sum.dsum += plist[i]; - } - break; - } - - default: - break; } _avg_over: @@ -757,6 +773,37 @@ _avg_over: return TSDB_CODE_SUCCESS; } +static void avgTransferInfo(SAvgRes* pInput, SAvgRes* pOutput) { + pOutput->type = pInput->type; + if (IS_INTEGER_TYPE(pOutput->type)) { + pOutput->sum.isum += pInput->sum.isum; + } else { + pOutput->sum.dsum += pInput->sum.dsum; + } + + pOutput->count += pInput->count; + + return; +} + +int32_t avgFunctionMerge(SqlFunctionCtx* pCtx) { + SInputColumnInfoData* pInput = &pCtx->input; + SColumnInfoData* pCol = pInput->pData[0]; + ASSERT(pCol->info.type == TSDB_DATA_TYPE_BINARY); + + SAvgRes* pInfo = GET_ROWCELL_INTERBUF(GET_RES_INFO(pCtx)); + + int32_t start = pInput->startRowIndex; + char* data = colDataGetData(pCol, start); + SAvgRes* pInputInfo = (SAvgRes*)varDataVal(data); + + avgTransferInfo(pInputInfo, pInfo); + + SET_VAL(GET_RES_INFO(pCtx), 1, 1); + + return TSDB_CODE_SUCCESS; +} + #define LIST_AVG_N(sumT, T) \ do { \ T* plist = (T*)pCol->pData; \ @@ -841,8 +888,8 @@ int32_t avgCombine(SqlFunctionCtx* pDestCtx, SqlFunctionCtx* pSourceCtx) { int32_t avgFinalize(SqlFunctionCtx* pCtx, SSDataBlock* pBlock) { SInputColumnInfoData* pInput = &pCtx->input; - int32_t type = pInput->pData[0]->info.type; SAvgRes* pAvgRes = GET_ROWCELL_INTERBUF(GET_RES_INFO(pCtx)); + int32_t type = pAvgRes->type; if (IS_INTEGER_TYPE(type)) { pAvgRes->result = pAvgRes->sum.isum / ((double)pAvgRes->count); @@ -858,6 +905,24 @@ int32_t avgFinalize(SqlFunctionCtx* pCtx, SSDataBlock* pBlock) { return functionFinalize(pCtx, pBlock); } +int32_t avgPartialFinalize(SqlFunctionCtx* pCtx, SSDataBlock* pBlock) { + SResultRowEntryInfo* pResInfo = GET_RES_INFO(pCtx); + SAvgRes* pInfo = GET_ROWCELL_INTERBUF(GET_RES_INFO(pCtx)); + int32_t resultBytes = getAvgInfoSize(); + char* res = taosMemoryCalloc(resultBytes + VARSTR_HEADER_SIZE, sizeof(char)); + + memcpy(varDataVal(res), pInfo, resultBytes); + varDataSetLen(res, resultBytes); + + int32_t slotId = pCtx->pExpr->base.resSchema.slotId; + SColumnInfoData* pCol = taosArrayGet(pBlock->pDataBlock, slotId); + + colDataAppend(pCol, pBlock->info.rows, res, false); + + taosMemoryFree(res); + return pResInfo->numOfRes; +} + EFuncDataRequired statisDataRequired(SFunctionNode* pFunc, STimeWindow* pTimeWindow) { return FUNC_DATA_REQUIRED_STATIS_LOAD; } @@ -3226,11 +3291,10 @@ int32_t spreadFunctionMerge(SqlFunctionCtx* pCtx) { ASSERT(pCol->info.type == TSDB_DATA_TYPE_BINARY); SSpreadInfo* pInfo = GET_ROWCELL_INTERBUF(GET_RES_INFO(pCtx)); - SSpreadInfo* pInputInfo; int32_t start = pInput->startRowIndex; char* data = colDataGetData(pCol, start); - pInputInfo = (SSpreadInfo*)varDataVal(data); + SSpreadInfo* pInputInfo = (SSpreadInfo*)varDataVal(data); spreadTransferInfo(pInputInfo, pInfo); diff --git a/source/libs/index/src/indexFilter.c b/source/libs/index/src/indexFilter.c index 6197edd474..90aafb1097 100644 --- a/source/libs/index/src/indexFilter.c +++ b/source/libs/index/src/indexFilter.c @@ -97,7 +97,14 @@ static int32_t sifGetOperParamNum(EOperatorType ty) { } return 2; } -static int32_t sifValidateColumn(SColumnNode *cn) { +static int32_t sifValidOp(EOperatorType ty) { + if ((ty >= OP_TYPE_ADD && ty <= OP_TYPE_BIT_OR) || (ty == OP_TYPE_IN || ty == OP_TYPE_NOT_IN) || + (ty == OP_TYPE_LIKE || ty == OP_TYPE_NOT_LIKE || ty == OP_TYPE_MATCH || ty == OP_TYPE_NMATCH)) { + return -1; + } + return 0; +} +static int32_t sifValidColumn(SColumnNode *cn) { // add more check if (cn == NULL) { return TSDB_CODE_QRY_INVALID_INPUT; @@ -176,6 +183,7 @@ static int32_t sifInitJsonParam(SNode *node, SIFParam *param, SIFCtx *ctx) { memcpy(param->colName, r->literal, strlen(r->literal)); // sprintf(param->colName, "%s_%s", l->colName, r->literal); param->colValType = r->typeData; + param->status = SFLT_COARSE_INDEX; return 0; // memcpy(param->colName, l->colName, sizeof(l->colName)); } @@ -197,7 +205,7 @@ static int32_t sifInitParam(SNode *node, SIFParam *param, SIFCtx *ctx) { case QUERY_NODE_COLUMN: { SColumnNode *cn = (SColumnNode *)node; /*only support tag column*/ - SIF_ERR_RET(sifValidateColumn(cn)); + SIF_ERR_RET(sifValidColumn(cn)); param->colId = cn->colId; param->colValType = cn->node.resType.type; @@ -247,11 +255,13 @@ static int32_t sifInitOperParams(SIFParam **params, SOperatorNode *node, SIFCtx return code; } SIFParam *paramList = taosMemoryCalloc(nParam, sizeof(SIFParam)); + if (NULL == paramList) { SIF_ERR_RET(TSDB_CODE_QRY_OUT_OF_MEMORY); } - if (nodeType(node->pLeft) == QUERY_NODE_OPERATOR) { + if (nodeType(node->pLeft) == QUERY_NODE_OPERATOR && + (((SOperatorNode *)(node->pLeft))->opType == OP_TYPE_JSON_GET_VALUE)) { SNode *interNode = (node->pLeft); SIF_ERR_JRET(sifInitJsonParam(interNode, ¶mList[0], ctx)); if (nParam > 1) { @@ -505,6 +515,11 @@ static int32_t sifGetOperFn(int32_t funcId, sif_func_t *func, SIdxFltStatus *sta static int32_t sifExecOper(SOperatorNode *node, SIFCtx *ctx, SIFParam *output) { int32_t code = 0; + if (sifValidOp(node->opType) < 0) { + output->status = SFLT_NOT_INDEX; + return code; + } + int32_t nParam = sifGetOperParamNum(node->opType); if (nParam <= 1) { output->status = SFLT_NOT_INDEX; @@ -513,9 +528,15 @@ static int32_t sifExecOper(SOperatorNode *node, SIFCtx *ctx, SIFParam *output) { if (node->opType == OP_TYPE_JSON_GET_VALUE) { return code; } - SIFParam *params = NULL; + SIFParam *params = NULL; SIF_ERR_RET(sifInitOperParams(¶ms, node, ctx)); + + if (params[0].status == SFLT_NOT_INDEX || (nParam > 1 && params[1].status == SFLT_NOT_INDEX)) { + output->status = SFLT_NOT_INDEX; + return code; + } + // ugly code, refactor later output->arg = ctx->arg; sif_func_t operFn = sifNullFunc; diff --git a/source/libs/qcom/src/querymsg.c b/source/libs/qcom/src/querymsg.c index 10d0f4a44a..b89c220519 100644 --- a/source/libs/qcom/src/querymsg.c +++ b/source/libs/qcom/src/querymsg.c @@ -484,13 +484,11 @@ int32_t queryProcessGetTbIndexRsp(void *output, char *msg, int32_t msgSize) { return TSDB_CODE_TSC_INVALID_INPUT; } - STableIndexRsp out = {0}; - if (tDeserializeSTableIndexRsp(msg, msgSize, &out) != 0) { + STableIndexRsp *out = (STableIndexRsp*)output; + if (tDeserializeSTableIndexRsp(msg, msgSize, out) != 0) { qError("tDeserializeSTableIndexRsp failed, msgSize:%d", msgSize); return TSDB_CODE_INVALID_MSG; } - - *(void **)output = out.pIndex; return TSDB_CODE_SUCCESS; } diff --git a/source/libs/qworker/src/qworker.c b/source/libs/qworker/src/qworker.c index ae0eda5449..451607e7d0 100644 --- a/source/libs/qworker/src/qworker.c +++ b/source/libs/qworker/src/qworker.c @@ -837,7 +837,6 @@ void qwProcessHbTimerEvent(void *param, void *tmrId) { SQWorker *mgmt = qwAcquire(refId); if (NULL == mgmt) { QW_DLOG("qwAcquire %" PRIx64 "failed", refId); - taosMemoryFree(param); return; } diff --git a/source/libs/scalar/src/filter.c b/source/libs/scalar/src/filter.c index 6704bfdde3..a446aa7bdb 100644 --- a/source/libs/scalar/src/filter.c +++ b/source/libs/scalar/src/filter.c @@ -1476,6 +1476,11 @@ void filterDumpInfoToString(SFilterInfo *info, const char *msg, int32_t options) for (uint32_t i = 0; i < info->fields[FLD_TYPE_VALUE].num; ++i) { SFilterField *field = &info->fields[FLD_TYPE_VALUE].fields[i]; if (field->desc) { + if (QUERY_NODE_VALUE != nodeType(field->desc)) { + qDebug("VAL%d => [type:not value node][val:NIL]", i); //TODO + continue; + } + SValueNode *var = (SValueNode *)field->desc; SDataType *dType = &var->node.resType; if (dType->type == TSDB_DATA_TYPE_VALUE_ARRAY) { diff --git a/source/libs/sync/inc/syncInt.h b/source/libs/sync/inc/syncInt.h index e7777af749..8a44dcd9bc 100644 --- a/source/libs/sync/inc/syncInt.h +++ b/source/libs/sync/inc/syncInt.h @@ -168,6 +168,7 @@ SSyncNode* syncNodeOpen(const SSyncInfo* pSyncInfo); void syncNodeStart(SSyncNode* pSyncNode); void syncNodeStartStandBy(SSyncNode* pSyncNode); void syncNodeClose(SSyncNode* pSyncNode); +int32_t syncNodePropose(SSyncNode* pSyncNode, const SRpcMsg* pMsg, bool isWeak); // option bool syncNodeSnapshotEnable(SSyncNode* pSyncNode); @@ -232,6 +233,9 @@ int32_t syncNodeCommit(SSyncNode* ths, SyncIndex beginIndex, SyncIndex endIndex, bool syncNodeInRaftGroup(SSyncNode* ths, SRaftId* pRaftId); SSyncSnapshotSender* syncNodeGetSnapshotSender(SSyncNode* ths, SRaftId* pDestId); +void syncStartNormal(int64_t rid); +void syncStartStandBy(int64_t rid); + // for debug -------------- void syncNodePrint(SSyncNode* pObj); void syncNodePrint2(char* s, SSyncNode* pObj); diff --git a/source/libs/sync/src/syncAppendEntries.c b/source/libs/sync/src/syncAppendEntries.c index 29ee263756..08a4081ad3 100644 --- a/source/libs/sync/src/syncAppendEntries.c +++ b/source/libs/sync/src/syncAppendEntries.c @@ -995,7 +995,7 @@ int32_t syncNodeOnAppendEntriesSnapshotCb(SSyncNode* ths, SyncAppendEntries* pMs ths->commitIndex = snapshot.lastApplyIndex; sDebug("vgId:%d sync event commit by snapshot from index:%ld to index:%ld, %s", ths->vgId, commitBegin, - commitEnd, syncUtilState2String(ths->state)); + commitEnd, syncUtilState2String(ths->state)); } SyncIndex beginIndex = ths->commitIndex + 1; diff --git a/source/libs/sync/src/syncCommit.c b/source/libs/sync/src/syncCommit.c index eac9c1fbd8..3424fac5e7 100644 --- a/source/libs/sync/src/syncCommit.c +++ b/source/libs/sync/src/syncCommit.c @@ -57,7 +57,7 @@ void syncMaybeAdvanceCommitIndex(SSyncNode* pSyncNode) { pSyncNode->commitIndex = snapshot.lastApplyIndex; sDebug("vgId:%d sync event commit by snapshot from index:%ld to index:%ld, %s", pSyncNode->vgId, - pSyncNode->commitIndex, snapshot.lastApplyIndex, syncUtilState2String(pSyncNode->state)); + pSyncNode->commitIndex, snapshot.lastApplyIndex, syncUtilState2String(pSyncNode->state)); } // update commit index diff --git a/source/libs/sync/src/syncMain.c b/source/libs/sync/src/syncMain.c index e7f11b9a7d..d2e82d25f6 100644 --- a/source/libs/sync/src/syncMain.c +++ b/source/libs/sync/src/syncMain.c @@ -149,12 +149,12 @@ void syncStop(int64_t rid) { int32_t syncSetStandby(int64_t rid) { SSyncNode* pSyncNode = (SSyncNode*)taosAcquireRef(tsNodeRefId, rid); if (pSyncNode == NULL) { - return -1; + return TAOS_SYNC_OTHER_ERROR; } if (pSyncNode->state == TAOS_SYNC_STATE_LEADER) { taosReleaseRef(tsNodeRefId, pSyncNode->rid); - return -1; + return TAOS_SYNC_OTHER_ERROR; } // state change @@ -174,14 +174,88 @@ int32_t syncSetStandby(int64_t rid) { return 0; } -int32_t syncReconfig(int64_t rid, const SSyncCfg* pSyncCfg) { - int32_t ret = 0; - char* newconfig = syncCfg2Str((SSyncCfg*)pSyncCfg); +int32_t syncReconfigBuild(int64_t rid, const SSyncCfg* pNewCfg, SRpcMsg* pRpcMsg) { + SSyncNode* pSyncNode = (SSyncNode*)taosAcquireRef(tsNodeRefId, rid); + if (pSyncNode == NULL) { + return TAOS_SYNC_OTHER_ERROR; + } + ASSERT(rid == pSyncNode->rid); + int32_t ret = 0; + bool IamInNew = false; + for (int i = 0; i < pNewCfg->replicaNum; ++i) { + if (strcmp((pNewCfg->nodeInfo)[i].nodeFqdn, pSyncNode->myNodeInfo.nodeFqdn) == 0 && + (pNewCfg->nodeInfo)[i].nodePort == pSyncNode->myNodeInfo.nodePort) { + IamInNew = true; + } + + /* + SRaftId newId; + newId.addr = syncUtilAddr2U64((pNewCfg->nodeInfo)[i].nodeFqdn, (pNewCfg->nodeInfo)[i].nodePort); + newId.vgId = pSyncNode->vgId; + if (syncUtilSameId(&(pSyncNode->myRaftId), &newId)) { + IamInNew = true; + } + */ + } + + if (!IamInNew) { + taosReleaseRef(tsNodeRefId, pSyncNode->rid); + return TAOS_SYNC_NOT_IN_NEW_CONFIG; + } + + 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); + + taosReleaseRef(tsNodeRefId, pSyncNode->rid); + return ret; +} + +int32_t syncReconfig(int64_t rid, const SSyncCfg* pNewCfg) { + SSyncNode* pSyncNode = (SSyncNode*)taosAcquireRef(tsNodeRefId, rid); + if (pSyncNode == NULL) { + return TAOS_SYNC_OTHER_ERROR; + } + ASSERT(rid == pSyncNode->rid); + + bool IamInNew = false; + for (int i = 0; i < pNewCfg->replicaNum; ++i) { + if (strcmp((pNewCfg->nodeInfo)[i].nodeFqdn, pSyncNode->myNodeInfo.nodeFqdn) == 0 && + (pNewCfg->nodeInfo)[i].nodePort == pSyncNode->myNodeInfo.nodePort) { + IamInNew = true; + } + + /* + // some problem in inet_addr + + SRaftId newId = EMPTY_RAFT_ID; + newId.addr = syncUtilAddr2U64((pNewCfg->nodeInfo)[i].nodeFqdn, (pNewCfg->nodeInfo)[i].nodePort); + newId.vgId = pSyncNode->vgId; + + if (syncUtilSameId(&(pSyncNode->myRaftId), &newId)) { + IamInNew = true; + } + */ + } + + if (!IamInNew) { + sError("sync reconfig error, not in new config"); + taosReleaseRef(tsNodeRefId, pSyncNode->rid); + return TAOS_SYNC_NOT_IN_NEW_CONFIG; + } + + char* newconfig = syncCfg2Str((SSyncCfg*)pNewCfg); if (gRaftDetailLog) { sInfo("==syncReconfig== newconfig:%s", newconfig); } + int32_t ret = 0; + SRpcMsg rpcMsg = {0}; rpcMsg.msgType = TDMT_SYNC_CONFIG_CHANGE; rpcMsg.info.noResp = 1; @@ -189,58 +263,59 @@ int32_t syncReconfig(int64_t rid, const SSyncCfg* pSyncCfg) { rpcMsg.pCont = rpcMallocCont(rpcMsg.contLen); snprintf(rpcMsg.pCont, rpcMsg.contLen, "%s", newconfig); taosMemoryFree(newconfig); - ret = syncPropose(rid, &rpcMsg, false); + ret = syncNodePropose(pSyncNode, &rpcMsg, false); + + taosReleaseRef(tsNodeRefId, pSyncNode->rid); return ret; } int32_t syncLeaderTransfer(int64_t rid) { - int32_t ret = 0; + SSyncNode* pSyncNode = (SSyncNode*)taosAcquireRef(tsNodeRefId, rid); + if (pSyncNode == NULL) { + return TAOS_SYNC_OTHER_ERROR; + } + ASSERT(rid == pSyncNode->rid); + if (pSyncNode->peersNum == 0) { + taosReleaseRef(tsNodeRefId, pSyncNode->rid); + return TAOS_SYNC_OTHER_ERROR; + } + + SNodeInfo newLeader = (pSyncNode->peersNodeInfo)[0]; + taosReleaseRef(tsNodeRefId, pSyncNode->rid); + + int32_t ret = syncLeaderTransferTo(rid, newLeader); return ret; } int32_t syncLeaderTransferTo(int64_t rid, SNodeInfo newLeader) { SSyncNode* pSyncNode = (SSyncNode*)taosAcquireRef(tsNodeRefId, rid); if (pSyncNode == NULL) { - return false; + return TAOS_SYNC_OTHER_ERROR; } - assert(rid == pSyncNode->rid); + ASSERT(rid == pSyncNode->rid); int32_t ret = 0; if (pSyncNode->replicaNum == 1) { - taosReleaseRef(tsNodeRefId, pSyncNode->rid); sError("only one replica, cannot drop leader"); + taosReleaseRef(tsNodeRefId, pSyncNode->rid); return TAOS_SYNC_ONLY_ONE_REPLICA; } SyncLeaderTransfer* pMsg = syncLeaderTransferBuild(pSyncNode->vgId); pMsg->newLeaderId.addr = syncUtilAddr2U64(newLeader.nodeFqdn, newLeader.nodePort); pMsg->newLeaderId.vgId = pSyncNode->vgId; + pMsg->newNodeInfo = newLeader; ASSERT(pMsg != NULL); SRpcMsg rpcMsg = {0}; syncLeaderTransfer2RpcMsg(pMsg, &rpcMsg); syncLeaderTransferDestroy(pMsg); - ret = syncPropose(rid, &rpcMsg, false); - + ret = syncNodePropose(pSyncNode, &rpcMsg, false); taosReleaseRef(tsNodeRefId, pSyncNode->rid); 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; -} - bool syncCanLeaderTransfer(int64_t rid) { SSyncNode* pSyncNode = (SSyncNode*)taosAcquireRef(tsNodeRefId, rid); if (pSyncNode == NULL) { @@ -273,8 +348,6 @@ bool syncCanLeaderTransfer(int64_t rid) { return matchOK; } -int32_t syncGiveUpLeader(int64_t rid) { return 0; } - int32_t syncForwardToPeer(int64_t rid, const SRpcMsg* pMsg, bool isWeak) { int32_t ret = syncPropose(rid, pMsg, isWeak); return ret; @@ -469,16 +542,26 @@ int32_t syncPropose(int64_t rid, const SRpcMsg* pMsg, bool isWeak) { SSyncNode* pSyncNode = taosAcquireRef(tsNodeRefId, rid); if (pSyncNode == NULL) { - return TAOS_SYNC_PROPOSE_OTHER_ERROR; + return TAOS_SYNC_OTHER_ERROR; } assert(rid == pSyncNode->rid); + sDebug("vgId:%d sync event propose msgType:%s", pSyncNode->vgId, TMSG_INFO(pMsg->msgType)); + ret = syncNodePropose(pSyncNode, pMsg, isWeak); + + taosReleaseRef(tsNodeRefId, pSyncNode->rid); + return ret; +} + +int32_t syncNodePropose(SSyncNode* pSyncNode, const SRpcMsg* pMsg, bool isWeak) { + int32_t ret = TAOS_SYNC_PROPOSE_SUCCESS; + sDebug("vgId:%d sync event propose msgType:%s", pSyncNode->vgId, TMSG_INFO(pMsg->msgType)); if (pSyncNode->state == TAOS_SYNC_STATE_LEADER) { SRespStub stub; stub.createTime = taosGetTimestampMs(); stub.rpcMsg = *pMsg; uint64_t seqNum = syncRespMgrAdd(pSyncNode->pSyncRespMgr, &stub); - sDebug("vgId:%d, sync event propose, type:%s seq:%" PRIu64 " handle:%p", pSyncNode->vgId, TMSG_INFO(pMsg->msgType), + sDebug("vgId:%d sync event propose, type:%s seq:%" PRIu64 " handle:%p", pSyncNode->vgId, TMSG_INFO(pMsg->msgType), seqNum, pMsg->info.handle); SyncClientRequest* pSyncMsg = syncClientRequestBuild2(pMsg, seqNum, isWeak, pSyncNode->vgId); @@ -488,16 +571,14 @@ int32_t syncPropose(int64_t rid, const SRpcMsg* pMsg, bool isWeak) { if (pSyncNode->FpEqMsg != NULL && (*pSyncNode->FpEqMsg)(pSyncNode->msgcb, &rpcMsg) == 0) { ret = TAOS_SYNC_PROPOSE_SUCCESS; } else { - sTrace("syncPropose pSyncNode->FpEqMsg is NULL"); + sError("syncPropose pSyncNode->FpEqMsg is NULL"); } syncClientRequestDestroy(pSyncMsg); } else { - sDebug("vgId:%d, failed to propose since not leader, type:%s handle:%p %s", pSyncNode->vgId, - TMSG_INFO(pMsg->msgType), pMsg->info.handle, syncUtilState2String(pSyncNode->state)); + sError("syncPropose not leader, %s", syncUtilState2String(pSyncNode->state)); ret = TAOS_SYNC_PROPOSE_NOT_LEADER; } - taosReleaseRef(tsNodeRefId, pSyncNode->rid); return ret; } @@ -1162,8 +1243,14 @@ void syncNodeUpdateConfig(SSyncNode* pSyncNode, SSyncCfg* newConfig, SyncIndex l int32_t oldReplicaNum = pSyncNode->replicaNum; SRaftId oldReplicasId[TSDB_MAX_REPLICA]; memcpy(oldReplicasId, pSyncNode->replicasId, sizeof(oldReplicasId)); - SSyncSnapshotSender* oldSenders[TSDB_MAX_REPLICA]; - memcpy(oldSenders, pSyncNode->senders, sizeof(oldSenders)); + SSyncSnapshotSender* oldSenders[TSDB_MAX_REPLICA]; + for (int i = 0; i < TSDB_MAX_REPLICA; ++i) { + oldSenders[i] = (pSyncNode->senders)[i]; + sDebug("vgId:%d sync event save senders %d, %p", pSyncNode->vgId, i, oldSenders[i]); + if (gRaftDetailLog) { + ; + } + } // init internal pSyncNode->myNodeInfo = pSyncNode->pRaftCfg->cfg.nodeInfo[pSyncNode->pRaftCfg->cfg.myIndex]; @@ -1195,24 +1282,51 @@ void syncNodeUpdateConfig(SSyncNode* pSyncNode, SSyncCfg* newConfig, SyncIndex l pSyncNode->quorum = syncUtilQuorum(pSyncNode->pRaftCfg->cfg.replicaNum); - // reset snapshot senders, memory leak + // reset snapshot senders + + // clear new for (int i = 0; i < TSDB_MAX_REPLICA; ++i) { (pSyncNode->senders)[i] = NULL; } + + // reset new for (int i = 0; i < pSyncNode->replicaNum; ++i) { + // reset sender + bool reset = false; for (int j = 0; j < TSDB_MAX_REPLICA; ++j) { if (syncUtilSameId(&(pSyncNode->replicasId)[i], &oldReplicasId[j])) { - char host[128]; + char host[128]; uint16_t port; syncUtilU642Addr((pSyncNode->replicasId)[i].addr, host, sizeof(host), &port); - sDebug("vgId:%d sync event reset sender for %lu, %s:%d", pSyncNode->vgId, (pSyncNode->replicasId)[i].addr, host, port); + sDebug("vgId:%d sync event reset sender for %lu, newIndex:%d, %s:%d, %p", pSyncNode->vgId, + (pSyncNode->replicasId)[i].addr, i, host, port, oldSenders[j]); (pSyncNode->senders)[i] = oldSenders[j]; + oldSenders[j] = NULL; + reset = true; + + // reset replicaIndex + int32_t oldreplicaIndex = (pSyncNode->senders)[i]->replicaIndex; + (pSyncNode->senders)[i]->replicaIndex = i; + sDebug("vgId:%d sync event udpate replicaIndex from %d to %d, %s:%d, %p, reset:%d", pSyncNode->vgId, + oldreplicaIndex, i, host, port, (pSyncNode->senders)[i], reset); } } } + + // create new for (int i = 0; i < TSDB_MAX_REPLICA; ++i) { if ((pSyncNode->senders)[i] == NULL) { (pSyncNode->senders)[i] = snapshotSenderCreate(pSyncNode, i); + sDebug("vgId:%d sync event create new sender %p replicaIndex:%d", pSyncNode->vgId, (pSyncNode->senders)[i], i); + } + } + + // free old + for (int i = 0; i < TSDB_MAX_REPLICA; ++i) { + if (oldSenders[i] != NULL) { + snapshotSenderDestroy(oldSenders[i]); + sDebug("vgId:%d sync event delete old sender %p replicaIndex:%d", pSyncNode->vgId, oldSenders[i], i); + oldSenders[i] = NULL; } } @@ -1272,8 +1386,8 @@ void syncNodeUpdateTerm(SSyncNode* pSyncNode, SyncTerm term) { } void syncNodeBecomeFollower(SSyncNode* pSyncNode, const char* debugStr) { - sDebug("vgId:%d sync event become follower, isStandBy:%d, %s", pSyncNode->vgId, pSyncNode->pRaftCfg->isStandBy, - debugStr); + sDebug("vgId:%d sync event become follower, isStandBy:%d, replicaNum:%d, %s", pSyncNode->vgId, + pSyncNode->pRaftCfg->isStandBy, pSyncNode->replicaNum, debugStr); // maybe clear leader cache if (pSyncNode->state == TAOS_SYNC_STATE_LEADER) { @@ -1307,8 +1421,8 @@ void syncNodeBecomeFollower(SSyncNode* pSyncNode, const char* debugStr) { // /\ UNCHANGED <> // void syncNodeBecomeLeader(SSyncNode* pSyncNode, const char* debugStr) { - sDebug("vgId:%d sync event become leader, isStandBy:%d, %s", pSyncNode->vgId, pSyncNode->pRaftCfg->isStandBy, - debugStr); + sDebug("vgId:%d sync event become leader, isStandBy:%d, replicaNum:%d %s", pSyncNode->vgId, + pSyncNode->pRaftCfg->isStandBy, pSyncNode->replicaNum, debugStr); // state change pSyncNode->state = TAOS_SYNC_STATE_LEADER; @@ -1857,10 +1971,52 @@ const char* syncStr(ESyncState state) { } static int32_t syncDoLeaderTransfer(SSyncNode* ths, SRpcMsg* pRpcMsg, SSyncRaftEntry* pEntry) { - SyncLeaderTransfer* pSyncLeaderTransfer; - if (syncUtilSameId(&(pSyncLeaderTransfer->newLeaderId), &(ths->myRaftId))) { + SyncLeaderTransfer* pSyncLeaderTransfer = syncLeaderTransferFromRpcMsg2(pRpcMsg); + + /* + char host[128]; + uint16_t port; + syncUtilU642Addr(pSyncLeaderTransfer->newLeaderId.addr, host, sizeof(host), &port); + sDebug("vgId:%d sync event, maybe leader transfer to %s:%d %lu", ths->vgId, host, port, + pSyncLeaderTransfer->newLeaderId.addr); + */ + + sDebug("vgId:%d sync event, begin leader transfer", ths->vgId); + + if (strcmp(pSyncLeaderTransfer->newNodeInfo.nodeFqdn, ths->myNodeInfo.nodeFqdn) == 0 && + pSyncLeaderTransfer->newNodeInfo.nodePort == ths->myNodeInfo.nodePort) { + sDebug("vgId:%d sync event, maybe leader transfer to %s:%d %lu", ths->vgId, + pSyncLeaderTransfer->newNodeInfo.nodeFqdn, pSyncLeaderTransfer->newNodeInfo.nodePort, + pSyncLeaderTransfer->newLeaderId.addr); + + // reset elect timer now! + int32_t electMS = 1; + int32_t ret = syncNodeRestartElectTimer(ths, electMS); + ASSERT(ret == 0); } + /* + if (syncUtilSameId(&(pSyncLeaderTransfer->newLeaderId), &(ths->myRaftId))) { + // reset elect timer now! + int32_t electMS = 1; + int32_t ret = syncNodeRestartElectTimer(ths, electMS); + ASSERT(ret == 0); + } + */ + if (ths->pFsm->FpLeaderTransferCb != NULL) { + SFsmCbMeta cbMeta; + cbMeta.code = 0; + cbMeta.currentTerm = ths->pRaftStore->currentTerm; + cbMeta.flag = 0; + cbMeta.index = pEntry->index; + cbMeta.isWeak = pEntry->isWeak; + cbMeta.seqNum = pEntry->seqNum; + cbMeta.state = ths->state; + cbMeta.term = pEntry->term; + ths->pFsm->FpLeaderTransferCb(ths->pFsm, pRpcMsg, cbMeta); + } + + syncLeaderTransferDestroy(pSyncLeaderTransfer); return 0; } @@ -1884,26 +2040,31 @@ static int32_t syncNodeConfigChange(SSyncNode* ths, SRpcMsg* pRpcMsg, SSyncRaftE bool isDrop; - //if (IamInNew || (!IamInNew && ths->state != TAOS_SYNC_STATE_LEADER)) { + // if (IamInNew || (!IamInNew && ths->state != TAOS_SYNC_STATE_LEADER)) { if (IamInNew) { syncNodeUpdateConfig(ths, &newSyncCfg, pEntry->index, &isDrop); // change isStandBy to normal if (!isDrop) { + char tmpbuf[128]; + snprintf(tmpbuf, sizeof(tmpbuf), "config change from %d to %d", oldSyncCfg.replicaNum, newSyncCfg.replicaNum); if (ths->state == TAOS_SYNC_STATE_LEADER) { - syncNodeBecomeLeader(ths, "config change"); + syncNodeBecomeLeader(ths, tmpbuf); } else { - syncNodeBecomeFollower(ths, "config change"); + syncNodeBecomeFollower(ths, tmpbuf); } } } else { - syncNodeBecomeFollower(ths, "config change2"); + char tmpbuf[128]; + snprintf(tmpbuf, sizeof(tmpbuf), "config change2 from %d to %d", oldSyncCfg.replicaNum, newSyncCfg.replicaNum); + syncNodeBecomeFollower(ths, tmpbuf); } - + if (gRaftDetailLog) { char* sOld = syncCfg2Str(&oldSyncCfg); char* sNew = syncCfg2Str(&newSyncCfg); - sInfo("==config change== 0x11 old:%s new:%s isDrop:%d index:%ld \n", sOld, sNew, isDrop, pEntry->index); + sInfo("==config change== 0x11 old:%s new:%s isDrop:%d index:%ld IamInNew:%d \n", sOld, sNew, isDrop, pEntry->index, + IamInNew); taosMemoryFree(sOld); taosMemoryFree(sNew); } @@ -1965,7 +2126,7 @@ int32_t syncNodeCommit(SSyncNode* ths, SyncIndex beginIndex, SyncIndex endIndex, ASSERT(code == 0); } - // config change + // leader transfer if (pEntry->originalRpcType == TDMT_SYNC_LEADER_TRANSFER) { code = syncDoLeaderTransfer(ths, &rpcMsg, pEntry); ASSERT(code == 0); diff --git a/source/libs/sync/src/syncSnapshot.c b/source/libs/sync/src/syncSnapshot.c index b8ed283170..afebfa798e 100644 --- a/source/libs/sync/src/syncSnapshot.c +++ b/source/libs/sync/src/syncSnapshot.c @@ -393,7 +393,7 @@ cJSON *snapshotSender2Json(SSyncSnapshotSender *pSender) { char *snapshotSender2Str(SSyncSnapshotSender *pSender) { cJSON *pJson = snapshotSender2Json(pSender); - char * serialized = cJSON_Print(pJson); + char *serialized = cJSON_Print(pJson); cJSON_Delete(pJson); return serialized; } @@ -514,7 +514,7 @@ cJSON *snapshotReceiver2Json(SSyncSnapshotReceiver *pReceiver) { cJSON_AddStringToObject(pFromId, "addr", u64buf); { uint64_t u64 = pReceiver->fromId.addr; - cJSON * pTmp = pFromId; + cJSON *pTmp = pFromId; char host[128] = {0}; uint16_t port; syncUtilU642Addr(u64, host, sizeof(host), &port); @@ -538,7 +538,7 @@ cJSON *snapshotReceiver2Json(SSyncSnapshotReceiver *pReceiver) { char *snapshotReceiver2Str(SSyncSnapshotReceiver *pReceiver) { cJSON *pJson = snapshotReceiver2Json(pReceiver); - char * serialized = cJSON_Print(pJson); + char *serialized = cJSON_Print(pJson); cJSON_Delete(pJson); return serialized; } @@ -588,6 +588,8 @@ int32_t syncNodeOnSnapshotSendCb(SSyncNode *pSyncNode, SyncSnapshotSend *pMsg) { // maybe update lastconfig if (pMsg->lastConfigIndex >= SYNC_INDEX_BEGIN) { + int32_t oldReplicaNum = pSyncNode->replicaNum; + // update new config myIndex bool IamInNew = false; SSyncCfg newSyncCfg = pMsg->lastConfig; @@ -614,10 +616,12 @@ int32_t syncNodeOnSnapshotSendCb(SSyncNode *pSyncNode, SyncSnapshotSend *pMsg) { // change isStandBy to normal if (!isDrop) { + char tmpbuf[128]; + snprintf(tmpbuf, sizeof(tmpbuf), "config change3 from %d to %d", oldReplicaNum, newSyncCfg.replicaNum); if (pSyncNode->state == TAOS_SYNC_STATE_LEADER) { - syncNodeBecomeLeader(pSyncNode, "config change"); + syncNodeBecomeLeader(pSyncNode, tmpbuf); } else { - syncNodeBecomeFollower(pSyncNode, "config change"); + syncNodeBecomeFollower(pSyncNode, tmpbuf); } } } diff --git a/source/libs/sync/test/CMakeLists.txt b/source/libs/sync/test/CMakeLists.txt index d39035ba53..725343e373 100644 --- a/source/libs/sync/test/CMakeLists.txt +++ b/source/libs/sync/test/CMakeLists.txt @@ -1,4 +1,5 @@ add_executable(syncTest "") +add_executable(syncRaftIdCheck "") add_executable(syncEnvTest "") add_executable(syncPingTimerTest "") add_executable(syncIOTickQTest "") @@ -54,6 +55,10 @@ target_sources(syncTest PRIVATE "syncTest.cpp" ) +target_sources(syncRaftIdCheck + PRIVATE + "syncRaftIdCheck.cpp" +) target_sources(syncEnvTest PRIVATE "syncEnvTest.cpp" @@ -257,6 +262,11 @@ target_include_directories(syncTest "${TD_SOURCE_DIR}/include/libs/sync" "${CMAKE_CURRENT_SOURCE_DIR}/../inc" ) +target_include_directories(syncRaftIdCheck + PUBLIC + "${TD_SOURCE_DIR}/include/libs/sync" + "${CMAKE_CURRENT_SOURCE_DIR}/../inc" +) target_include_directories(syncEnvTest PUBLIC "${TD_SOURCE_DIR}/include/libs/sync" @@ -508,6 +518,10 @@ target_link_libraries(syncTest sync gtest_main ) +target_link_libraries(syncRaftIdCheck + sync + gtest_main +) target_link_libraries(syncEnvTest sync gtest_main diff --git a/source/libs/sync/test/syncRaftIdCheck.cpp b/source/libs/sync/test/syncRaftIdCheck.cpp new file mode 100644 index 0000000000..90560e91e7 --- /dev/null +++ b/source/libs/sync/test/syncRaftIdCheck.cpp @@ -0,0 +1,32 @@ +#include +#include +#include "syncIO.h" +#include "syncInt.h" +#include "syncUtil.h" + +void usage(char* exe) { + printf("Usage: %s host port \n", exe); + printf("Usage: %s u64 \n", exe); +} + +int main(int argc, char** argv) { + if (argc == 2) { + uint64_t u64 = atoll(argv[1]); + char host[128]; + uint16_t port; + syncUtilU642Addr(u64, host, sizeof(host), &port); + printf("%lu -> %s:%d \n", u64, host, port); + + } else if (argc == 3) { + uint64_t u64; + char* host = argv[1]; + uint16_t port = atoi(argv[2]); + u64 = syncUtilAddr2U64(host, port); + printf("%s:%d -> %lu \n", host, port, u64); + } else { + usage(argv[0]); + exit(-1); + } + + return 0; +} diff --git a/source/libs/sync/test/syncTestTool.cpp b/source/libs/sync/test/syncTestTool.cpp index 0c8b26e9d9..ebcd7368cc 100644 --- a/source/libs/sync/test/syncTestTool.cpp +++ b/source/libs/sync/test/syncTestTool.cpp @@ -153,6 +153,16 @@ void ReConfigCb(struct SSyncFSM* pFsm, const SRpcMsg* pMsg, SReConfigCbMeta cbMe taosMemoryFree(s); } +void LeaderTransferCb(struct SSyncFSM* pFsm, const SRpcMsg* pMsg, SFsmCbMeta cbMeta) { + char logBuf[256] = {0}; + snprintf(logBuf, sizeof(logBuf), + "==callback== ==LeaderTransferCb== pFsm:%p, index:%ld, isWeak:%d, code:%d, state:%d %s, flag:%lu, term:%lu " + "currentTerm:%lu \n", + pFsm, cbMeta.index, cbMeta.isWeak, cbMeta.code, cbMeta.state, syncUtilState2String(cbMeta.state), + cbMeta.flag, cbMeta.term, cbMeta.currentTerm); + syncRpcMsgLog2(logBuf, (SRpcMsg*)pMsg); +} + SSyncFSM* createFsm() { SSyncFSM* pFsm = (SSyncFSM*)taosMemoryMalloc(sizeof(SSyncFSM)); memset(pFsm, 0, sizeof(*pFsm)); @@ -172,6 +182,8 @@ SSyncFSM* createFsm() { pFsm->FpSnapshotStopWrite = SnapshotStopWrite; pFsm->FpSnapshotDoWrite = SnapshotDoWrite; + pFsm->FpLeaderTransferCb = LeaderTransferCb; + return pFsm; } @@ -277,7 +289,8 @@ void usage(char* exe) { printf( "usage: %s replicaNum(1-5) myIndex(0-..) enableSnapshot(0/1) lastApplyIndex(>=-1) lastApplyTerm(>=0) " "writeRecordNum(>=0) " - "isStandBy(0/1) isConfigChange(0-5) iterTimes(>=0) finishLastApplyIndex(>=-1) finishLastApplyTerm(>=0) \n", + "isStandBy(0/1) isConfigChange(0-5) iterTimes(>=0) finishLastApplyIndex(>=-1) finishLastApplyTerm(>=0) " + "leaderTransfer(0/1) \n", exe); } @@ -294,9 +307,9 @@ SRpcMsg* createRpcMsg(int i, int count, int myIndex) { int main(int argc, char** argv) { sprintf(tsTempDir, "%s", "."); tsAsyncLog = 0; - sDebugFlag = DEBUG_SCREEN + DEBUG_FILE + DEBUG_TRACE + DEBUG_INFO + DEBUG_ERROR; + sDebugFlag = DEBUG_SCREEN + DEBUG_FILE + DEBUG_TRACE + DEBUG_INFO + DEBUG_ERROR + DEBUG_DEBUG; - if (argc != 12) { + if (argc != 13) { usage(argv[0]); exit(-1); } @@ -312,12 +325,14 @@ int main(int argc, char** argv) { int32_t iterTimes = atoi(argv[9]); int32_t finishLastApplyIndex = atoi(argv[10]); int32_t finishLastApplyTerm = atoi(argv[11]); + int32_t leaderTransfer = atoi(argv[12]); - sTrace( + sInfo( "args: replicaNum:%d, myIndex:%d, enableSnapshot:%d, lastApplyIndex:%d, lastApplyTerm:%d, writeRecordNum:%d, " - "isStandBy:%d, isConfigChange:%d, iterTimes:%d, finishLastApplyIndex:%d, finishLastApplyTerm:%d", + "isStandBy:%d, isConfigChange:%d, iterTimes:%d, finishLastApplyIndex:%d, finishLastApplyTerm:%d, " + "leaderTransfer:%d", replicaNum, myIndex, enableSnapshot, lastApplyIndex, lastApplyTerm, writeRecordNum, isStandBy, isConfigChange, - iterTimes, finishLastApplyIndex, finishLastApplyTerm); + iterTimes, finishLastApplyIndex, finishLastApplyTerm, leaderTransfer); // check parameter assert(replicaNum >= 1 && replicaNum <= 5); @@ -363,24 +378,31 @@ int main(int argc, char** argv) { //--------------------------- int32_t alreadySend = 0; + int32_t leaderTransferWait = 0; while (1) { char* simpleStr = syncNode2SimpleStr(pSyncNode); + leaderTransferWait++; + if (leaderTransferWait == 7) { + sTrace("begin leader transfer ..."); + int32_t ret = syncLeaderTransfer(rid); + } + if (alreadySend < writeRecordNum) { SRpcMsg* pRpcMsg = createRpcMsg(alreadySend, writeRecordNum, myIndex); int32_t ret = syncPropose(rid, pRpcMsg, false); if (ret == TAOS_SYNC_PROPOSE_NOT_LEADER) { - sTrace("%s value%d write not leader", simpleStr, alreadySend); + sTrace("%s value%d write not leader, leaderTransferWait:%d", simpleStr, alreadySend, leaderTransferWait); } else { assert(ret == 0); - sTrace("%s value%d write ok", simpleStr, alreadySend); + sTrace("%s value%d write ok, leaderTransferWait:%d", simpleStr, alreadySend, leaderTransferWait); } alreadySend++; rpcFreeCont(pRpcMsg->pCont); taosMemoryFree(pRpcMsg); } else { - sTrace("%s", simpleStr); + sTrace("%s, leaderTransferWait:%d", simpleStr, leaderTransferWait); } taosMsleep(1000); diff --git a/source/os/src/osLz4.c b/source/os/src/osLz4.c index f4e4951857..7f3c42fb6c 100644 --- a/source/os/src/osLz4.c +++ b/source/os/src/osLz4.c @@ -45,13 +45,13 @@ int32_t BUILDIN_CLZL(uint64_t val) { #else _MyBitScanReverse64(&r, val); #endif - return (int)(r >> 3); + return (int)(63 - r); } int32_t BUILDIN_CLZ(uint32_t val) { unsigned long r = 0; _BitScanReverse(&r, val); - return (int)(r >> 3); + return (int)(31 - r); } int32_t BUILDIN_CTZL(uint64_t val) { @@ -61,13 +61,13 @@ int32_t BUILDIN_CTZL(uint64_t val) { #else _MyBitScanForward64(&r, val); #endif - return (int)(r >> 3); + return (int)(r); } int32_t BUILDIN_CTZ(uint32_t val) { unsigned long r = 0; _BitScanForward(&r, val); - return (int)(r >> 3); + return (int)(r); } #endif diff --git a/source/util/src/tarray.c b/source/util/src/tarray.c index a34b332b9d..2357f760d1 100644 --- a/source/util/src/tarray.c +++ b/source/util/src/tarray.c @@ -458,6 +458,9 @@ static void taosArrayInsertSort(SArray* pArray, __ext_compar_fn_t fn, const void } SArray* taosArrayDeepCopy(const SArray* pSrc, FCopy deepCopy) { + if (NULL == pSrc) { + return NULL; + } ASSERT(pSrc->elemSize == sizeof(void*)); SArray* pArray = taosArrayInit(pSrc->size, sizeof(void*)); for (int32_t i = 0; i < pSrc->size; i++) { diff --git a/source/util/src/terror.c b/source/util/src/terror.c index e122ad0ab6..079d5ef590 100644 --- a/source/util/src/terror.c +++ b/source/util/src/terror.c @@ -566,6 +566,7 @@ TAOS_DEFINE_ERROR(TSDB_CODE_TSMA_RM_SKEY_IN_HASH, "Rm tsma skey in cac 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_INDEX_REBUILDING, "Index is rebuilding") #ifdef TAOS_ERROR_C }; diff --git a/tests/script/tsim/stable/tag_filter.sim b/tests/script/tsim/stable/tag_filter.sim index 04b8f21de9..1f400eb803 100644 --- a/tests/script/tsim/stable/tag_filter.sim +++ b/tests/script/tsim/stable/tag_filter.sim @@ -73,5 +73,80 @@ if $rows != 6 then endi +print ========== prepare stbBin and ctbBin +sql create table db.stbBin (ts timestamp, c1 int, c2 binary(4)) tags(t1 binary(16)) + +sql create table db.ctbBin using db.stbBin tags("a") +sql insert into db.ctbBin values(now, 1, "2") + +sql create table db.ctbBin1 using db.stbBin tags("b") +sql insert into db.ctbBin1 values(now, 2, "2") + +sql create table db.ctbBin2 using db.stbBin tags("c") +sql insert into db.ctbBin2 values(now, 3, "2") + +sql create table db.ctbBin3 using db.stbBin tags("d") +sql insert into db.ctbBin3 values(now, 4, "2") + +sql select * from db.stbBin where t1 = "a" +if $rows != 1 then + return -1 +endi + + +sql select * from db.stbBin where t1 < "a" +if $rows != 0 then + return -=1 +endi + +sql select * from db.stbBin where t1 < "b" +if $rows != 1 then + return -1 +endi + + +sql select * from db.stbBin where t1 between "a" and "e" +if $rows != 4 then + return -1 +endi + + +print ========== prepare stbNc and ctbNc +sql create table db.stbNc (ts timestamp, c1 int, c2 binary(4)) tags(t1 nchar(16)) + + +sql create table db.ctbNc using db.stbNc tags("a") +sql insert into db.ctbNc values(now, 1, "2") + +sql create table db.ctbNc1 using db.stbNc tags("b") +sql insert into db.ctbNc1 values(now, 2, "2") + +sql create table db.ctbNc2 using db.stbNc tags("c") +sql insert into db.ctbNc2 values(now, 3, "2") + +sql create table db.ctbNc3 using db.stbNc tags("d") +sql insert into db.ctbNc3 values(now, 4, "2") + +sql select * from db.stbNc where t1 = "a" +if $rows != 1 then + return -1 +endi + + +sql select * from db.stbNc where t1 < "a" +if $rows != 0 then + return -=1 +endi + +sql select * from db.stbNc where t1 < "b" +if $rows != 1 then + return -1 +endi + + +sql select * from db.stbNc where t1 between "a" and "e" +if $rows != 4 then + return -1 +endi system sh/exec.sh -n dnode1 -s stop -x SIGINT diff --git a/tests/system-test/7-tmq/subscribeDb1.py b/tests/system-test/7-tmq/subscribeDb1.py index 44c24eb616..ed92a429ae 100644 --- a/tests/system-test/7-tmq/subscribeDb1.py +++ b/tests/system-test/7-tmq/subscribeDb1.py @@ -437,12 +437,12 @@ class TDTestCase: event.wait() tdLog.info("start consume processor") - pollDelay = 100 + pollDelay = 20 showMsg = 1 showRow = 1 self.startTmqSimProcess(buildPath,cfgPath,pollDelay,parameterDict["dbName"],showMsg, showRow) - time.sleep(5) + time.sleep(3) tdLog.info("pkill consume processor") if (platform.system().lower() == 'windows'): os.system("TASKKILL /F /IM tmq_sim.exe") @@ -465,7 +465,7 @@ class TDTestCase: totalConsumeRows += resultList[i] if totalConsumeRows >= expectrowcnt or totalConsumeRows <= 0: - tdLog.info("act consume rows: %d, expect consume rows: %d"%(totalConsumeRows, expectrowcnt)) + tdLog.info("act consume rows: %d, expect consume rows between %d and 0"%(totalConsumeRows, expectrowcnt)) tdLog.exit("tmq consume rows error!") time.sleep(15) diff --git a/tests/system-test/fulltest.sh b/tests/system-test/fulltest.sh index cda0e37237..e04db9dae9 100755 --- a/tests/system-test/fulltest.sh +++ b/tests/system-test/fulltest.sh @@ -104,7 +104,7 @@ python3 ./test.py -f 6-cluster/5dnode3mnodeStop.py python3 ./test.py -f 7-tmq/basic5.py python3 ./test.py -f 7-tmq/subscribeDb.py python3 ./test.py -f 7-tmq/subscribeDb0.py -# python3 ./test.py -f 7-tmq/subscribeDb1.py +python3 ./test.py -f 7-tmq/subscribeDb1.py python3 ./test.py -f 7-tmq/subscribeStb.py python3 ./test.py -f 7-tmq/subscribeStb0.py python3 ./test.py -f 7-tmq/subscribeStb1.py