提交 86f8bf6c 编写于 作者: S Shengliang Guan

refactor: make trans support multi steps

上级 53def5b7
......@@ -130,7 +130,7 @@ static void dmProcessRpcMsg(SDnode *pDnode, SRpcMsg *pRpc, SEpSet *pEpSet) {
_OVER:
if (code != 0) {
dError("msg:%p, failed to process since %s", pMsg, terrstr());
dTrace("msg:%p, failed to process since %s, type:%s", pMsg, terrstr(), TMSG_INFO(pRpc->msgType));
if (terrno != 0) code = terrno;
if (IsReq(pRpc)) {
......
......@@ -31,17 +31,18 @@ typedef enum {
typedef struct {
int32_t id;
tmsg_t msgType;
int8_t msgSent;
int8_t msgReceived;
int32_t errCode;
int32_t acceptableCode;
int8_t stage;
int8_t isRaw;
int8_t rawWritten;
SSdbRaw *pRaw;
int8_t msgSent;
int8_t msgReceived;
tmsg_t msgType;
SEpSet epSet;
int32_t errCode;
int32_t acceptableCode;
int32_t contLen;
void *pCont;
SSdbRaw *pRaw;
} STransAction;
typedef void (*TransCbFp)(SMnode *pMnode, void *param, int32_t paramLen);
......
......@@ -78,10 +78,8 @@ static int32_t mndCreateDefaultAcct(SMnode *pMnode) {
if (pRaw == NULL) return -1;
sdbSetRawStatus(pRaw, SDB_STATUS_READY);
mDebug("acct:%s, will be created while deploy sdb, raw:%p", acctObj.acct, pRaw);
#if 0
return sdbWrite(pMnode->pSdb, pRaw);
#else
mDebug("acct:%s, will be created when deploying, raw:%p", acctObj.acct, pRaw);
STrans *pTrans = mndTransCreate(pMnode, TRN_POLICY_RETRY, TRN_TYPE_CREATE_ACCT, NULL);
if (pTrans == NULL) {
mError("acct:%s, failed to create since %s", acctObj.acct, terrstr());
......@@ -94,7 +92,6 @@ static int32_t mndCreateDefaultAcct(SMnode *pMnode) {
mndTransDrop(pTrans);
return -1;
}
sdbSetRawStatus(pRaw, SDB_STATUS_READY);
if (mndTransPrepare(pMnode, pTrans) != 0) {
mError("trans:%d, failed to prepare since %s", pTrans->id, terrstr());
......@@ -104,7 +101,6 @@ static int32_t mndCreateDefaultAcct(SMnode *pMnode) {
mndTransDrop(pTrans);
return 0;
#endif
}
static SSdbRaw *mndAcctActionEncode(SAcctObj *pAcct) {
......
......@@ -172,13 +172,13 @@ static int32_t mndCreateDefaultCluster(SMnode *pMnode) {
clusterObj.id = mndGenerateUid(clusterObj.name, TSDB_CLUSTER_ID_LEN);
clusterObj.id = (clusterObj.id >= 0 ? clusterObj.id : -clusterObj.id);
pMnode->clusterId = clusterObj.id;
mDebug("cluster:%" PRId64 ", name is %s", clusterObj.id, clusterObj.name);
mInfo("cluster:%" PRId64 ", name is %s", clusterObj.id, clusterObj.name);
SSdbRaw *pRaw = mndClusterActionEncode(&clusterObj);
if (pRaw == NULL) return -1;
sdbSetRawStatus(pRaw, SDB_STATUS_READY);
mDebug("cluster:%" PRId64 ", will be created while deploy sdb, raw:%p", clusterObj.id, pRaw);
mDebug("cluster:%" PRId64 ", will be created when deploying, raw:%p", clusterObj.id, pRaw);
#if 0
return sdbWrite(pMnode->pSdb, pRaw);
#else
......
......@@ -1314,7 +1314,7 @@ int32_t mndValidateDbInfo(SMnode *pMnode, SDbVgVersion *pDbs, int32_t numOfDbs,
SDbObj *pDb = mndAcquireDb(pMnode, pDbVgVersion->dbFName);
if (pDb == NULL) {
mDebug("db:%s, no exist", pDbVgVersion->dbFName);
mTrace("db:%s, no exist", pDbVgVersion->dbFName);
memcpy(usedbRsp.db, pDbVgVersion->dbFName, TSDB_DB_FNAME_LEN);
usedbRsp.uid = pDbVgVersion->dbId;
usedbRsp.vgVersion = -1;
......
......@@ -98,7 +98,7 @@ static int32_t mndCreateDefaultDnode(SMnode *pMnode) {
if (pRaw == NULL) return -1;
if (sdbSetRawStatus(pRaw, SDB_STATUS_READY) != 0) return -1;
mDebug("dnode:%d, will be created while deploy sdb, raw:%p", dnodeObj.id, pRaw);
mDebug("dnode:%d, will be created when deploying, raw:%p", dnodeObj.id, pRaw);
#if 0
return sdbWrite(pMnode->pSdb, pRaw);
......@@ -388,9 +388,10 @@ static int32_t mndProcessStatusReq(SRpcMsg *pReq) {
mndReleaseMnode(pMnode, pObj);
}
int64_t dnodeVer = sdbGetTableVer(pMnode->pSdb, SDB_DNODE) + sdbGetTableVer(pMnode->pSdb, SDB_MNODE);
int64_t curMs = taosGetTimestampMs();
bool online = mndIsDnodeOnline(pMnode, pDnode, curMs);
bool dnodeChanged = (statusReq.dnodeVer != sdbGetTableVer(pMnode->pSdb, SDB_DNODE));
bool dnodeChanged = (statusReq.dnodeVer != dnodeVer);
bool reboot = (pDnode->rebootTime != statusReq.rebootTime);
bool needCheck = !online || dnodeChanged || reboot;
......@@ -433,7 +434,8 @@ static int32_t mndProcessStatusReq(SRpcMsg *pReq) {
if (!online) {
mInfo("dnode:%d, from offline to online", pDnode->id);
} else {
mDebug("dnode:%d, send dnode eps", pDnode->id);
mDebug("dnode:%d, send dnode epset, online:%d ver:% " PRId64 ":%" PRId64 " reboot:%d", pDnode->id, online,
statusReq.dnodeVer, dnodeVer, reboot);
}
pDnode->rebootTime = statusReq.rebootTime;
......@@ -441,7 +443,7 @@ static int32_t mndProcessStatusReq(SRpcMsg *pReq) {
pDnode->numOfSupportVnodes = statusReq.numOfSupportVnodes;
SStatusRsp statusRsp = {0};
statusRsp.dnodeVer = sdbGetTableVer(pMnode->pSdb, SDB_DNODE) + sdbGetTableVer(pMnode->pSdb, SDB_MNODE);
statusRsp.dnodeVer = dnodeVer;
statusRsp.dnodeCfg.dnodeId = pDnode->id;
statusRsp.dnodeCfg.clusterId = pMnode->clusterId;
statusRsp.pDnodeEps = taosArrayInit(mndGetDnodeSize(pMnode), sizeof(SDnodeEp));
......
......@@ -472,7 +472,7 @@ int32_t mndProcessRpcMsg(SRpcMsg *pMsg) {
} else if (code == 0) {
mTrace("msg:%p, successfully processed and response", pMsg);
} else {
mError("msg:%p, failed to process since %s, app:%p type:%s", pMsg, terrstr(), pMsg->info.ahandle,
mDebug("msg:%p, failed to process since %s, app:%p type:%s", pMsg, terrstr(), pMsg->info.ahandle,
TMSG_INFO(pMsg->msgType));
}
......
......@@ -90,7 +90,7 @@ static int32_t mndCreateDefaultMnode(SMnode *pMnode) {
if (pRaw == NULL) return -1;
sdbSetRawStatus(pRaw, SDB_STATUS_READY);
mDebug("mnode:%d, will be created while deploy sdb, raw:%p", mnodeObj.id, pRaw);
mDebug("mnode:%d, will be created when deploying, raw:%p", mnodeObj.id, pRaw);
#if 0
return sdbWrite(pMnode->pSdb, pRaw);
......
......@@ -1597,7 +1597,7 @@ static int32_t mndProcessTableMetaReq(SRpcMsg *pReq) {
pReq->info.rspLen = rspLen;
code = 0;
mDebug("stb:%s.%s, meta is retrieved", infoReq.dbFName, infoReq.tbName);
mTrace("%s.%s, meta is retrieved", infoReq.dbFName, infoReq.tbName);
_OVER:
if (code != 0) {
......
......@@ -65,7 +65,7 @@ int32_t mndSyncGetSnapshot(struct SSyncFSM *pFsm, SSnapshot *pSnapshot) {
void mndRestoreFinish(struct SSyncFSM *pFsm) {
SMnode *pMnode = pFsm->data;
if (!pMnode->deploy) {
mInfo("mnode sync restore finished");
mInfo("mnode sync restore finished, and will handle outstanding transactions");
mndTransPullup(pMnode);
mndSetRestore(pMnode, true);
} else {
......@@ -244,7 +244,7 @@ void mndSyncStart(SMnode *pMnode) {
} else {
syncStart(pMgmt->sync);
}
mDebug("sync:%" PRId64 " is started, standby:%d", pMgmt->sync, pMgmt->standby);
mDebug("mnode sync started, id:%" PRId64 " standby:%d", pMgmt->sync, pMgmt->standby);
}
void mndSyncStop(SMnode *pMnode) {}
......
......@@ -77,7 +77,7 @@ static int32_t mndCreateDefaultUser(SMnode *pMnode, char *acct, char *user, char
if (pRaw == NULL) return -1;
sdbSetRawStatus(pRaw, SDB_STATUS_READY);
mDebug("user:%s, will be created while deploy sdb, raw:%p", userObj.user, pRaw);
mDebug("user:%s, will be created when deploying, raw:%p", userObj.user, pRaw);
#if 0
return sdbWrite(pMnode->pSdb, pRaw);
......
......@@ -501,7 +501,7 @@ int32_t mndAllocVgroup(SMnode *pMnode, SDbObj *pDb, SVgObj **ppVgroups) {
*ppVgroups = pVgroups;
code = 0;
mInfo("db:%s, %d vgroups is alloced, replica:%d", pDb->name, pDb->cfg.numOfVgroups, pDb->cfg.replications);
mInfo("db:%s, total %d vgroups is alloced, replica:%d", pDb->name, pDb->cfg.numOfVgroups, pDb->cfg.replications);
_OVER:
if (code != 0) taosMemoryFree(pVgroups);
......@@ -539,7 +539,7 @@ int32_t mndAddVnodeToVgroup(SMnode *pMnode, SVgObj *pVgroup, SArray *pArray) {
pVgid->role = TAOS_SYNC_STATE_FOLLOWER;
pDnode->numOfVnodes++;
mInfo("db:%s, vgId:%d, vn:%d dnode:%d is added", pVgroup->dbName, pVgroup->vgId, maxPos, pVgid->dnodeId);
mInfo("db:%s, vgId:%d, vnode_index:%d dnode:%d is added", pVgroup->dbName, pVgroup->vgId, maxPos, pVgid->dnodeId);
maxPos++;
if (maxPos == 3) return 0;
}
......
......@@ -168,6 +168,7 @@ typedef struct SSdb {
char *currDir;
char *tmpDir;
int64_t lastCommitVer;
int64_t lastCommitTerm;
int64_t curVer;
int64_t curTerm;
int64_t tableVer[SDB_MAX];
......
......@@ -55,6 +55,7 @@ SSdb *sdbInit(SSdbOpt *pOption) {
pSdb->curVer = -1;
pSdb->curTerm = -1;
pSdb->lastCommitVer = -1;
pSdb->lastCommitTerm = -1;
pSdb->pMnode = pOption->pMnode;
taosThreadMutexInit(&pSdb->filelock, NULL);
mDebug("sdb init successfully");
......
......@@ -70,6 +70,7 @@ static void sdbResetData(SSdb *pSdb) {
pSdb->curVer = -1;
pSdb->curTerm = -1;
pSdb->lastCommitVer = -1;
pSdb->lastCommitTerm = -1;
mDebug("sdb reset successfully");
}
......@@ -211,12 +212,12 @@ static int32_t sdbReadFileImp(SSdb *pSdb) {
char file[PATH_MAX] = {0};
snprintf(file, sizeof(file), "%s%ssdb.data", pSdb->currDir, TD_DIRSEP);
mDebug("start to read file:%s", file);
mDebug("start to read sdb file:%s", file);
SSdbRaw *pRaw = taosMemoryMalloc(WAL_MAX_SIZE + 100);
if (pRaw == NULL) {
terrno = TSDB_CODE_OUT_OF_MEMORY;
mError("failed read file since %s", terrstr());
mError("failed read sdb file since %s", terrstr());
return -1;
}
......@@ -224,12 +225,12 @@ static int32_t sdbReadFileImp(SSdb *pSdb) {
if (pFile == NULL) {
taosMemoryFree(pRaw);
terrno = TAOS_SYSTEM_ERROR(errno);
mError("failed to read file:%s since %s", file, terrstr());
mError("failed to read sdb file:%s since %s", file, terrstr());
return 0;
}
if (sdbReadFileHead(pSdb, pFile) != 0) {
mError("failed to read file:%s head since %s", file, terrstr());
mError("failed to read sdb file:%s head since %s", file, terrstr());
taosMemoryFree(pRaw);
taosCloseFile(&pFile);
return -1;
......@@ -245,13 +246,13 @@ static int32_t sdbReadFileImp(SSdb *pSdb) {
if (ret < 0) {
code = TAOS_SYSTEM_ERROR(errno);
mError("failed to read file:%s since %s", file, tstrerror(code));
mError("failed to read sdb file:%s since %s", file, tstrerror(code));
break;
}
if (ret != readLen) {
code = TSDB_CODE_FILE_CORRUPTED;
mError("failed to read file:%s since %s", file, tstrerror(code));
mError("failed to read sdb file:%s since %s", file, tstrerror(code));
break;
}
......@@ -259,34 +260,36 @@ static int32_t sdbReadFileImp(SSdb *pSdb) {
ret = taosReadFile(pFile, pRaw->pData, readLen);
if (ret < 0) {
code = TAOS_SYSTEM_ERROR(errno);
mError("failed to read file:%s since %s", file, tstrerror(code));
mError("failed to read sdb file:%s since %s", file, tstrerror(code));
break;
}
if (ret != readLen) {
code = TSDB_CODE_FILE_CORRUPTED;
mError("failed to read file:%s since %s", file, tstrerror(code));
mError("failed to read sdb file:%s since %s", file, tstrerror(code));
break;
}
int32_t totalLen = sizeof(SSdbRaw) + pRaw->dataLen + sizeof(int32_t);
if ((!taosCheckChecksumWhole((const uint8_t *)pRaw, totalLen)) != 0) {
code = TSDB_CODE_CHECKSUM_ERROR;
mError("failed to read file:%s since %s", file, tstrerror(code));
mError("failed to read sdb file:%s since %s", file, tstrerror(code));
break;
}
code = sdbWriteWithoutFree(pSdb, pRaw);
if (code != 0) {
mError("failed to read file:%s since %s", file, terrstr());
mError("failed to read sdb file:%s since %s", file, terrstr());
goto _OVER;
}
}
code = 0;
pSdb->lastCommitVer = pSdb->curVer;
pSdb->lastCommitTerm = pSdb->curTerm;
memcpy(pSdb->tableVer, tableVer, sizeof(tableVer));
mDebug("read file:%s successfully, ver:%" PRId64, file, pSdb->lastCommitVer);
mDebug("read sdb file:%s successfully, ver:%" PRId64 " term:%" PRId64, file, pSdb->lastCommitVer,
pSdb->lastCommitTerm);
_OVER:
taosCloseFile(&pFile);
......@@ -302,7 +305,7 @@ int32_t sdbReadFile(SSdb *pSdb) {
sdbResetData(pSdb);
int32_t code = sdbReadFileImp(pSdb);
if (code != 0) {
mError("failed to read sdb since %s", terrstr());
mError("failed to read sdb file since %s", terrstr());
sdbResetData(pSdb);
}
......@@ -318,18 +321,19 @@ static int32_t sdbWriteFileImp(SSdb *pSdb) {
char curfile[PATH_MAX] = {0};
snprintf(curfile, sizeof(curfile), "%s%ssdb.data", pSdb->currDir, TD_DIRSEP);
mDebug("start to write file:%s, current ver:%" PRId64 " term:%" PRId64 ", commit ver:%" PRId64, curfile, pSdb->curVer,
pSdb->curTerm, pSdb->lastCommitVer);
mDebug("start to write sdb file, current ver:%" PRId64 " term:%" PRId64 ", commit ver:%" PRId64 " term:%" PRId64
" file:%s",
pSdb->curVer, pSdb->curTerm, pSdb->lastCommitVer, pSdb->lastCommitTerm, curfile);
TdFilePtr pFile = taosOpenFile(tmpfile, TD_FILE_CREATE | TD_FILE_WRITE | TD_FILE_TRUNC);
if (pFile == NULL) {
terrno = TAOS_SYSTEM_ERROR(errno);
mError("failed to open file:%s for write since %s", tmpfile, terrstr());
mError("failed to open sdb file:%s for write since %s", tmpfile, terrstr());
return -1;
}
if (sdbWriteFileHead(pSdb, pFile) != 0) {
mError("failed to write file:%s head since %s", tmpfile, terrstr());
mError("failed to write sdb file:%s head since %s", tmpfile, terrstr());
taosCloseFile(&pFile);
return -1;
}
......@@ -338,7 +342,7 @@ static int32_t sdbWriteFileImp(SSdb *pSdb) {
SdbEncodeFp encodeFp = pSdb->encodeFps[i];
if (encodeFp == NULL) continue;
mTrace("write %s to file, total %d rows", sdbTableName(i), sdbGetSize(pSdb, i));
mTrace("write %s to sdb file, total %d rows", sdbTableName(i), sdbGetSize(pSdb, i));
SHashObj *hash = pSdb->hashObjs[i];
TdThreadRwlock *pLock = &pSdb->locks[i];
......@@ -394,7 +398,7 @@ static int32_t sdbWriteFileImp(SSdb *pSdb) {
code = taosFsyncFile(pFile);
if (code != 0) {
code = TAOS_SYSTEM_ERROR(errno);
mError("failed to sync file:%s since %s", tmpfile, tstrerror(code));
mError("failed to sync sdb file:%s since %s", tmpfile, tstrerror(code));
}
}
......@@ -404,15 +408,17 @@ static int32_t sdbWriteFileImp(SSdb *pSdb) {
code = taosRenameFile(tmpfile, curfile);
if (code != 0) {
code = TAOS_SYSTEM_ERROR(errno);
mError("failed to write file:%s since %s", curfile, tstrerror(code));
mError("failed to write sdb file:%s since %s", curfile, tstrerror(code));
}
}
if (code != 0) {
mError("failed to write file:%s since %s", curfile, tstrerror(code));
mError("failed to write sdb file:%s since %s", curfile, tstrerror(code));
} else {
pSdb->lastCommitVer = pSdb->curVer;
mDebug("write file:%s successfully, ver:%" PRId64 " term:%" PRId64, curfile, pSdb->lastCommitVer, pSdb->curTerm);
pSdb->lastCommitTerm = pSdb->curTerm;
mDebug("write sdb file successfully, ver:%" PRId64 " term:%" PRId64 " file:%s", pSdb->lastCommitVer,
pSdb->lastCommitTerm, curfile);
}
terrno = code;
......@@ -427,7 +433,7 @@ int32_t sdbWriteFile(SSdb *pSdb) {
taosThreadMutexLock(&pSdb->filelock);
int32_t code = sdbWriteFileImp(pSdb);
if (code != 0) {
mError("failed to write sdb since %s", terrstr());
mError("failed to write sdb file since %s", terrstr());
}
taosThreadMutexUnlock(&pSdb->filelock);
return code;
......@@ -493,7 +499,7 @@ int32_t sdbStartRead(SSdb *pSdb, SSdbIter **ppIter) {
if (taosCopyFile(datafile, pIter->name) < 0) {
taosThreadMutexUnlock(&pSdb->filelock);
terrno = TAOS_SYSTEM_ERROR(errno);
mError("failed to copy file %s to %s since %s", datafile, pIter->name, terrstr());
mError("failed to copy sdb file %s to %s since %s", datafile, pIter->name, terrstr());
sdbCloseIter(pIter);
return -1;
}
......@@ -502,7 +508,7 @@ int32_t sdbStartRead(SSdb *pSdb, SSdbIter **ppIter) {
pIter->file = taosOpenFile(pIter->name, TD_FILE_READ);
if (pIter->file == NULL) {
terrno = TAOS_SYSTEM_ERROR(errno);
mError("failed to open file:%s since %s", pIter->name, terrstr());
mError("failed to open sdb file:%s since %s", pIter->name, terrstr());
sdbCloseIter(pIter);
return -1;
}
......
......@@ -79,7 +79,11 @@ int32_t qwExecTask(QW_FPARAMS_DEF, SQWTaskCtx *ctx, bool *queryEnd) {
if (taskHandle) {
code = qExecTask(taskHandle, &pRes, &useconds);
if (code) {
QW_TASK_ELOG("qExecTask failed, code:%x - %s", code, tstrerror(code));
if (code != TSDB_CODE_OPS_NOT_SUPPORT) {
QW_TASK_ELOG("qExecTask failed, code:%x - %s", code, tstrerror(code));
} else {
QW_TASK_DLOG("qExecTask failed, code:%x - %s", code, tstrerror(code));
}
QW_ERR_RET(code);
}
}
......
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