未验证 提交 059a5a91 编写于 作者: L Liu Jicong 提交者: GitHub

Merge pull request #11994 from taosdata/feature/tq

feat(stream): support perf schema
......@@ -132,6 +132,7 @@ extern const int32_t TYPE_BYTES[15];
#define TSDB_PERFS_TABLE_CONSUMERS "consumers"
#define TSDB_PERFS_TABLE_SUBSCRIPTIONS "subscriptions"
#define TSDB_PERFS_TABLE_OFFSETS "offsets"
#define TSDB_PERFS_TABLE_STREAMS "streams"
#define TSDB_INDEX_TYPE_SMA "SMA"
#define TSDB_INDEX_TYPE_FULLTEXT "FULLTEXT"
......
......@@ -188,7 +188,6 @@ typedef struct SRequestSendRecvBody {
typedef struct {
int8_t resType;
int32_t code;
char topic[TSDB_TOPIC_FNAME_LEN];
int32_t vgId;
SSchemaWrapper schema;
......
......@@ -582,8 +582,9 @@ typedef struct {
typedef struct {
char name[TSDB_TOPIC_FNAME_LEN];
char db[TSDB_DB_FNAME_LEN];
char outputSTbName[TSDB_TABLE_FNAME_LEN];
char sourceDb[TSDB_DB_FNAME_LEN];
char targetDb[TSDB_DB_FNAME_LEN];
char targetSTbName[TSDB_TABLE_FNAME_LEN];
int64_t createTime;
int64_t updateTime;
int64_t uid;
......
......@@ -812,6 +812,7 @@ static int32_t mndRetrieveConsumer(SNodeMsg *pReq, SShowObj *pShow, SSDataBlock
colDataAppend(pColInfo, numOfRows, (const char *)status, false);
// subscribed topics
// TODO: split into multiple rows
char topics[TSDB_SHOW_LIST_LEN + VARSTR_HEADER_SIZE] = {0};
char *showStr = taosShowStrArray(pConsumer->assignedTopics);
tstrncpy(varDataVal(topics), showStr, TSDB_SHOW_LIST_LEN);
......
......@@ -410,7 +410,7 @@ int32_t tEncodeSStreamObj(SCoder *pEncoder, const SStreamObj *pObj) {
int32_t sz = 0;
/*int32_t outputNameSz = 0;*/
if (tEncodeCStr(pEncoder, pObj->name) < 0) return -1;
if (tEncodeCStr(pEncoder, pObj->db) < 0) return -1;
if (tEncodeCStr(pEncoder, pObj->sourceDb) < 0) return -1;
if (tEncodeI64(pEncoder, pObj->createTime) < 0) return -1;
if (tEncodeI64(pEncoder, pObj->updateTime) < 0) return -1;
if (tEncodeI64(pEncoder, pObj->uid) < 0) return -1;
......@@ -456,7 +456,7 @@ int32_t tEncodeSStreamObj(SCoder *pEncoder, const SStreamObj *pObj) {
int32_t tDecodeSStreamObj(SCoder *pDecoder, SStreamObj *pObj) {
if (tDecodeCStrTo(pDecoder, pObj->name) < 0) return -1;
if (tDecodeCStrTo(pDecoder, pObj->db) < 0) return -1;
if (tDecodeCStrTo(pDecoder, pObj->sourceDb) < 0) return -1;
if (tDecodeI64(pDecoder, &pObj->createTime) < 0) return -1;
if (tDecodeI64(pDecoder, &pObj->updateTime) < 0) return -1;
if (tDecodeI64(pDecoder, &pObj->uid) < 0) return -1;
......
......@@ -124,14 +124,6 @@ static const SInfosTableSchema userStbsSchema[] = {
{.name = "table_comment", .bytes = 1024 + VARSTR_HEADER_SIZE, .type = TSDB_DATA_TYPE_VARCHAR},
};
static const SInfosTableSchema userStreamsSchema[] = {
{.name = "stream_name", .bytes = SYSTABLE_SCH_DB_NAME_LEN, .type = TSDB_DATA_TYPE_VARCHAR},
{.name = "user_name", .bytes = 23, .type = TSDB_DATA_TYPE_VARCHAR},
{.name = "dest_table", .bytes = SYSTABLE_SCH_DB_NAME_LEN, .type = TSDB_DATA_TYPE_VARCHAR},
{.name = "create_time", .bytes = 8, .type = TSDB_DATA_TYPE_TIMESTAMP},
{.name = "sql", .bytes = 1024, .type = TSDB_DATA_TYPE_VARCHAR},
};
static const SInfosTableSchema userTblsSchema[] = {
{.name = "table_name", .bytes = SYSTABLE_SCH_TABLE_NAME_LEN, .type = TSDB_DATA_TYPE_VARCHAR},
{.name = "db_name", .bytes = SYSTABLE_SCH_DB_NAME_LEN, .type = TSDB_DATA_TYPE_VARCHAR},
......@@ -259,7 +251,6 @@ static const SInfosTableMeta infosMeta[] = {
{TSDB_INS_TABLE_USER_FUNCTIONS, userFuncSchema, tListLen(userFuncSchema)},
{TSDB_INS_TABLE_USER_INDEXES, userIdxSchema, tListLen(userIdxSchema)},
{TSDB_INS_TABLE_USER_STABLES, userStbsSchema, tListLen(userStbsSchema)},
{TSDB_INS_TABLE_USER_STREAMS, userStreamsSchema, tListLen(userStreamsSchema)},
{TSDB_INS_TABLE_USER_TABLES, userTblsSchema, tListLen(userTblsSchema)},
{TSDB_INS_TABLE_USER_TABLE_DISTRIBUTED, userTblDistSchema, tListLen(userTblDistSchema)},
{TSDB_INS_TABLE_USER_USERS, userUsersSchema, tListLen(userUsersSchema)},
......
......@@ -76,6 +76,18 @@ static const SPerfsTableSchema offsetSchema[] = {
{.name = "skip_log_cnt", .bytes = 8, .type = TSDB_DATA_TYPE_BIGINT},
};
static const SPerfsTableSchema streamSchema[] = {
{.name = "stream_name", .bytes = SYSTABLE_SCH_DB_NAME_LEN, .type = TSDB_DATA_TYPE_VARCHAR},
{.name = "create_time", .bytes = 8, .type = TSDB_DATA_TYPE_TIMESTAMP},
{.name = "sql", .bytes = TSDB_SHOW_SQL_LEN, .type = TSDB_DATA_TYPE_VARCHAR},
{.name = "status", .bytes = 20 + VARSTR_HEADER_SIZE, .type = TSDB_DATA_TYPE_BINARY},
{.name = "source_db", .bytes = SYSTABLE_SCH_DB_NAME_LEN, .type = TSDB_DATA_TYPE_VARCHAR},
{.name = "target_db", .bytes = SYSTABLE_SCH_DB_NAME_LEN, .type = TSDB_DATA_TYPE_VARCHAR},
{.name = "target_table", .bytes = SYSTABLE_SCH_TABLE_NAME_LEN, .type = TSDB_DATA_TYPE_VARCHAR},
{.name = "watermark", .bytes = 8, .type = TSDB_DATA_TYPE_BIGINT},
{.name = "trigger", .bytes = 4, .type = TSDB_DATA_TYPE_INT},
};
static const SPerfsTableMeta perfsMeta[] = {
{TSDB_PERFS_TABLE_CONNECTIONS, connectionsSchema, tListLen(connectionsSchema)},
{TSDB_PERFS_TABLE_QUERIES, queriesSchema, tListLen(queriesSchema)},
......@@ -83,6 +95,7 @@ static const SPerfsTableMeta perfsMeta[] = {
{TSDB_PERFS_TABLE_CONSUMERS, consumerSchema, tListLen(consumerSchema)},
{TSDB_PERFS_TABLE_SUBSCRIPTIONS, subscriptionSchema, tListLen(subscriptionSchema)},
{TSDB_PERFS_TABLE_OFFSETS, offsetSchema, tListLen(offsetSchema)},
{TSDB_PERFS_TABLE_STREAMS, streamSchema, tListLen(streamSchema)},
};
// connection/application/
......
......@@ -382,7 +382,7 @@ int32_t mndScheduleStream(SMnode* pMnode, STrans* pTrans, SStreamObj* pStream) {
pTask->dispatchType = TASK_DISPATCH__SHUFFLE;
pTask->dispatchMsgType = TDMT_VND_TASK_WRITE_EXEC;
SDbObj* pDb = mndAcquireDb(pMnode, pStream->db);
SDbObj* pDb = mndAcquireDb(pMnode, pStream->sourceDb);
ASSERT(pDb);
if (mndExtractDbInfo(pMnode, pDb, &pTask->shuffleDispatcher.dbInfo, NULL) < 0) {
sdbRelease(pSdb, pDb);
......
......@@ -40,7 +40,7 @@ static int32_t mndProcessMCreateSmaReq(SNodeMsg *pReq);
static int32_t mndProcessMDropSmaReq(SNodeMsg *pReq);
static int32_t mndProcessVCreateSmaRsp(SNodeMsg *pRsp);
static int32_t mndProcessVDropSmaRsp(SNodeMsg *pRsp);
static int32_t mndRetrieveSma(SNodeMsg *pReq, SShowObj *pShow, SSDataBlock* pBlock, int32_t rows);
static int32_t mndRetrieveSma(SNodeMsg *pReq, SShowObj *pShow, SSDataBlock *pBlock, int32_t rows);
static void mndCancelGetNextSma(SMnode *pMnode, void *pIter);
int32_t mndInitSma(SMnode *pMnode) {
......@@ -406,7 +406,7 @@ static int32_t mndCreateSma(SMnode *pMnode, SNodeMsg *pReq, SMCreateSmaReq *pCre
SStreamObj streamObj = {0};
tstrncpy(streamObj.name, pCreate->name, TSDB_STREAM_FNAME_LEN);
tstrncpy(streamObj.db, pDb->name, TSDB_DB_FNAME_LEN);
tstrncpy(streamObj.sourceDb, pDb->name, TSDB_DB_FNAME_LEN);
streamObj.createTime = taosGetTimestampMs();
streamObj.updateTime = streamObj.createTime;
streamObj.uid = mndGenerateUid(pCreate->name, strlen(pCreate->name));
......@@ -707,7 +707,8 @@ int32_t mndProcessGetSmaReq(SMnode *pMnode, SUserIndexReq *indexReq, SUserI
SNode *node = NULL;
FOREACH(node, pList) {
SFunctionNode *pFunc = (SFunctionNode *)node;
extOffset += snprintf(rsp->indexExts + extOffset, sizeof(rsp->indexExts) - extOffset - 1, "%s%s", (extOffset ? ",":""), pFunc->functionName);
extOffset += snprintf(rsp->indexExts + extOffset, sizeof(rsp->indexExts) - extOffset - 1, "%s%s",
(extOffset ? "," : ""), pFunc->functionName);
}
*exist = true;
......@@ -718,13 +719,12 @@ int32_t mndProcessGetSmaReq(SMnode *pMnode, SUserIndexReq *indexReq, SUserI
return code;
}
static int32_t mndProcessVDropSmaRsp(SNodeMsg *pRsp) {
mndTransProcessRsp(pRsp);
return 0;
}
static int32_t mndRetrieveSma(SNodeMsg *pReq, SShowObj *pShow, SSDataBlock* pBlock, int32_t rows) {
static int32_t mndRetrieveSma(SNodeMsg *pReq, SShowObj *pShow, SSDataBlock *pBlock, int32_t rows) {
SMnode *pMnode = pReq->pNode;
SSdb *pSdb = pMnode->pSdb;
int32_t numOfRows = 0;
......@@ -758,8 +758,8 @@ static int32_t mndRetrieveSma(SNodeMsg *pReq, SShowObj *pShow, SSDataBlock* pBlo
char n1[TSDB_TABLE_FNAME_LEN + VARSTR_HEADER_SIZE] = {0};
STR_TO_VARSTR(n1, (char *)tNameGetTableName(&stbName));
SColumnInfoData* pColInfo = taosArrayGet(pBlock->pDataBlock, cols++);
colDataAppend(pColInfo, numOfRows, (const char*) n, false);
SColumnInfoData *pColInfo = taosArrayGet(pBlock->pDataBlock, cols++);
colDataAppend(pColInfo, numOfRows, (const char *)n, false);
pColInfo = taosArrayGet(pBlock->pDataBlock, cols++);
colDataAppend(pColInfo, numOfRows, (const char *)&pSma->createdTime, false);
......
......@@ -40,7 +40,7 @@ static int32_t mndProcessTaskDeployInternalRsp(SNodeMsg *pRsp);
/*static int32_t mndProcessDropStreamInRsp(SNodeMsg *pRsp);*/
static int32_t mndProcessStreamMetaReq(SNodeMsg *pReq);
static int32_t mndGetStreamMeta(SNodeMsg *pReq, SShowObj *pShow, STableMetaRsp *pMeta);
static int32_t mndRetrieveStream(SNodeMsg *pReq, SShowObj *pShow, char *data, int32_t rows);
static int32_t mndRetrieveStream(SNodeMsg *pReq, SShowObj *pShow, SSDataBlock *pBlock, int32_t rows);
static void mndCancelGetNextStream(SMnode *pMnode, void *pIter);
int32_t mndInitStream(SMnode *pMnode) {
......@@ -58,8 +58,8 @@ int32_t mndInitStream(SMnode *pMnode) {
/*mndSetMsgHandle(pMnode, TDMT_MND_DROP_STREAM, mndProcessDropStreamReq);*/
/*mndSetMsgHandle(pMnode, TDMT_MND_DROP_STREAM_RSP, mndProcessDropStreamInRsp);*/
// mndAddShowRetrieveHandle(pMnode, TSDB_MGMT_TABLE_TOPICS, mndRetrieveStream);
/*mndAddShowFreeIterHandle(pMnode, TSDB_MGMT_TABLE_TOPICS, mndCancelGetNextStream);*/
mndAddShowRetrieveHandle(pMnode, TSDB_MGMT_TABLE_STREAMS, mndRetrieveStream);
mndAddShowFreeIterHandle(pMnode, TSDB_MGMT_TABLE_STREAMS, mndCancelGetNextStream);
return sdbSetTable(pMnode->pSdb, table);
}
......@@ -294,8 +294,8 @@ static int32_t mndCreateStream(SMnode *pMnode, SNodeMsg *pReq, SCMCreateStreamRe
mDebug("stream:%s to create", pCreate->name);
SStreamObj streamObj = {0};
tstrncpy(streamObj.name, pCreate->name, TSDB_STREAM_FNAME_LEN);
tstrncpy(streamObj.db, pDb->name, TSDB_DB_FNAME_LEN);
tstrncpy(streamObj.outputSTbName, pCreate->outputSTbName, TSDB_TABLE_FNAME_LEN);
tstrncpy(streamObj.sourceDb, pDb->name, TSDB_DB_FNAME_LEN);
tstrncpy(streamObj.targetSTbName, pCreate->outputSTbName, TSDB_TABLE_FNAME_LEN);
streamObj.createTime = taosGetTimestampMs();
streamObj.updateTime = streamObj.createTime;
streamObj.uid = mndGenerateUid(pCreate->name, strlen(pCreate->name));
......@@ -424,58 +424,55 @@ static int32_t mndGetNumOfStreams(SMnode *pMnode, char *dbName, int32_t *pNumOfS
return 0;
}
static int32_t mndRetrieveStream(SNodeMsg *pReq, SShowObj *pShow, char *data, int32_t rows) {
static int32_t mndRetrieveStream(SNodeMsg *pReq, SShowObj *pShow, SSDataBlock *pBlock, int32_t rows) {
SMnode *pMnode = pReq->pNode;
SSdb *pSdb = pMnode->pSdb;
int32_t numOfRows = 0;
SStreamObj *pStream = NULL;
int32_t cols = 0;
char *pWrite;
char prefix[TSDB_DB_FNAME_LEN] = {0};
SDbObj *pDb = mndAcquireDb(pMnode, pShow->db);
if (pDb == NULL) return 0;
tstrncpy(prefix, pShow->db, TSDB_DB_FNAME_LEN);
strcat(prefix, TS_PATH_DELIMITER);
int32_t prefixLen = (int32_t)strlen(prefix);
while (numOfRows < rows) {
pShow->pIter = sdbFetch(pSdb, SDB_STREAM, pShow->pIter, (void **)&pStream);
pShow->pIter = sdbFetch(pSdb, SDB_TOPIC, pShow->pIter, (void **)&pStream);
if (pShow->pIter == NULL) break;
if (pStream->dbUid != pDb->uid) {
if (strncmp(pStream->name, prefix, prefixLen) != 0) {
mError("Inconsistent stream data, name:%s, db:%s, dbUid:%" PRIu64, pStream->name, pDb->name, pDb->uid);
}
SColumnInfoData *pColInfo;
SName n;
int32_t cols = 0;
sdbRelease(pSdb, pStream);
continue;
}
char streamName[TSDB_TABLE_NAME_LEN + VARSTR_HEADER_SIZE] = {0};
tNameFromString(&n, pStream->name, T_NAME_ACCT | T_NAME_DB);
tNameGetDbName(&n, varDataVal(streamName));
varDataSetLen(streamName, strlen(varDataVal(streamName)));
pColInfo = taosArrayGet(pBlock->pDataBlock, cols++);
colDataAppend(pColInfo, numOfRows, (const char *)streamName, false);
cols = 0;
pColInfo = taosArrayGet(pBlock->pDataBlock, cols++);
colDataAppend(pColInfo, numOfRows, (const char *)&pStream->createTime, false);
char streamName[TSDB_TABLE_NAME_LEN] = {0};
tstrncpy(streamName, pStream->name + prefixLen, TSDB_TABLE_NAME_LEN);
pWrite = data + pShow->offset[cols] * rows + pShow->bytes[cols] * numOfRows;
STR_TO_VARSTR(pWrite, streamName);
cols++;
char sql[TSDB_SHOW_SQL_LEN + VARSTR_HEADER_SIZE] = {0};
tstrncpy(&sql[VARSTR_HEADER_SIZE], pStream->sql, TSDB_SHOW_SQL_LEN);
varDataSetLen(sql, strlen(&sql[VARSTR_HEADER_SIZE]));
pColInfo = taosArrayGet(pBlock->pDataBlock, cols++);
colDataAppend(pColInfo, numOfRows, (const char *)sql, false);
pWrite = data + pShow->offset[cols] * rows + pShow->bytes[cols] * numOfRows;
*(int64_t *)pWrite = pStream->createTime;
cols++;
pColInfo = taosArrayGet(pBlock->pDataBlock, cols++);
colDataAppend(pColInfo, numOfRows, (const char *)&pStream->status, true);
pWrite = data + pShow->offset[cols] * rows + pShow->bytes[cols] * numOfRows;
STR_WITH_MAXSIZE_TO_VARSTR(pWrite, pStream->sql, pShow->bytes[cols]);
cols++;
pColInfo = taosArrayGet(pBlock->pDataBlock, cols++);
colDataAppend(pColInfo, numOfRows, (const char *)&pStream->sourceDb, true);
numOfRows++;
sdbRelease(pSdb, pStream);
}
pColInfo = taosArrayGet(pBlock->pDataBlock, cols++);
colDataAppend(pColInfo, numOfRows, (const char *)&pStream->targetDb, true);
mndReleaseDb(pMnode, pDb);
pShow->numOfRows += numOfRows;
return numOfRows;
pColInfo = taosArrayGet(pBlock->pDataBlock, cols++);
colDataAppend(pColInfo, numOfRows, (const char *)&pStream->targetSTbName, true);
pColInfo = taosArrayGet(pBlock->pDataBlock, cols++);
colDataAppend(pColInfo, numOfRows, (const char *)&pStream->waterMark, false);
pColInfo = taosArrayGet(pBlock->pDataBlock, cols++);
colDataAppend(pColInfo, numOfRows, (const char *)&pStream->trigger, false);
}
return 0;
}
static void mndCancelGetNextStream(SMnode *pMnode, void *pIter) {
......
......@@ -309,9 +309,6 @@ static int32_t mndDoRebalance(SMnode *pMnode, const SMqRebInputObj *pInput, SMqR
newConsumerEp.consumerId = consumerId;
newConsumerEp.vgs = taosArrayInit(0, sizeof(void *));
taosHashPut(pOutput->pSub->consumerHash, &consumerId, sizeof(int64_t), &newConsumerEp, sizeof(SMqConsumerEp));
/*SMqConsumer* pTestNew = taosHashGet(pOutput->pSub->consumerHash, &consumerId, sizeof(int64_t));*/
/*ASSERT(pTestNew->consumerId == consumerId);*/
/*ASSERT(pTestNew->vgs == newConsumerEp.vgs);*/
taosArrayPush(pOutput->newConsumers, &consumerId);
}
}
......@@ -369,7 +366,13 @@ static int32_t mndDoRebalance(SMnode *pMnode, const SMqRebInputObj *pInput, SMqR
}
}
// 8. generate logs
// 8. TODO generate logs
mInfo("rebalance calculation completed, rebalanced vg:");
for (int32_t i = 0; i < taosArrayGetSize(pOutput->rebVgs); i++) {
SMqRebOutputVg *pOutputRebVg = taosArrayGet(pOutput->rebVgs, i);
mInfo("vg: %d moved from consumer %ld to consumer %ld", pOutputRebVg->pVgEp->vgId, pOutputRebVg->oldConsumerId,
pOutputRebVg->newConsumerId);
}
// 9. clear
taosHashCleanup(pHash);
......@@ -447,7 +450,9 @@ static int32_t mndPersistRebResult(SMnode *pMnode, SNodeMsg *pMsg, const SMqRebO
goto REB_FAIL;
}
}
// 4. commit log: modification log
// 4. TODO commit log: modification log
// 5. execution
if (mndTransPrepare(pMnode, pTrans) != 0) goto REB_FAIL;
mndTransDrop(pTrans);
......
......@@ -61,7 +61,6 @@ int vnodeProcessWriteReq(SVnode *pVnode, SRpcMsg *pMsg, int64_t version, SRpcMsg
pReq = POINTER_SHIFT(pMsg->pCont, sizeof(SMsgHead));
len = pMsg->contLen - sizeof(SMsgHead);
// todo: change the interface here
if (tqPushMsg(pVnode->pTq, pMsg->pCont, pMsg->contLen, pMsg->msgType, version) < 0) {
vError("vgId: %d failed to push msg to TQ since %s", TD_VID(pVnode), tstrerror(terrno));
return -1;
......
......@@ -81,7 +81,7 @@
./test.sh -f tsim/insert/backquote.sim -m
./test.sh -f tsim/parser/fourArithmetic-basic.sim -m
./test.sh -f tsim/query/interval-offset.sim -m
#./test.sh -f tsim/tmq/basic1.sim -m
./test.sh -f tsim/tmq/basic3.sim -m
./test.sh -f tsim/stable/vnode3.sim -m
./test.sh -f tsim/qnode/basic1.sim -m
./test.sh -f tsim/mnode/basic1.sim -m
......
Markdown is supported
0% .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑!
想要评论请 注册