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

fix: ftrunc on appending WAL log failure for recovery in walWriteImpl

This commit is contained in:
Benguang Zhao 2022-10-18 01:37:55 +08:00
parent 23e18bff76
commit f0427c9322
1 changed files with 29 additions and 12 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

41
source/libs/wal/src/walWrite.c
View File

@ -408,25 +408,33 @@ END:
static int32_t walWriteIndex(SWal *pWal, int64_t ver, int64_t offset) { static int32_t walWriteIndex(SWal *pWal, int64_t ver, int64_t offset) {
SWalIdxEntry entry = {.ver = ver, .offset = offset}; SWalIdxEntry entry = {.ver = ver, .offset = offset};
int64_t idxOffset = taosLSeekFile(pWal->pIdxFile, 0, SEEK_END); SWalFileInfo *pFileInfo = walGetCurFileInfo(pWal);
ASSERT(pFileInfo != NULL);
ASSERT(pFileInfo->firstVer >= 0);
int64_t idxOffset = (entry.ver - pFileInfo->firstVer) * sizeof(SWalIdxEntry);
wDebug("vgId:%d, write index, index:%" PRId64 ", offset:%" PRId64 ", at %" PRId64, pWal->cfg.vgId, ver, offset, wDebug("vgId:%d, write index, index:%" PRId64 ", offset:%" PRId64 ", at %" PRId64, pWal->cfg.vgId, ver, offset,
idxOffset); idxOffset);
int64_t size = taosWriteFile(pWal->pIdxFile, &entry, sizeof(SWalIdxEntry)); int64_t size = taosWriteFile(pWal->pIdxFile, &entry, sizeof(SWalIdxEntry));
if (size != sizeof(SWalIdxEntry)) { if (size != sizeof(SWalIdxEntry)) {
wError("vgId:%d, failed to write idx entry due to %s. ver:%lld", pWal->cfg.vgId, strerror(errno), ver);
terrno = TAOS_SYSTEM_ERROR(errno); terrno = TAOS_SYSTEM_ERROR(errno);
// TODO truncate
return -1; return -1;
} }
return 0; return 0;
} }
// TODO gurantee atomicity by truncate failed writing
static FORCE_INLINE int32_t walWriteImpl(SWal *pWal, int64_t index, tmsg_t msgType, SWalSyncInfo syncMeta, static FORCE_INLINE int32_t walWriteImpl(SWal *pWal, int64_t index, tmsg_t msgType, SWalSyncInfo syncMeta,
const void *body, int32_t bodyLen) { const void *body, int32_t bodyLen) {
int64_t code = 0; int64_t code = 0;
int64_t offset = walGetCurFileOffset(pWal); int64_t offset = walGetCurFileOffset(pWal);
SWalFileInfo *pFileInfo = walGetCurFileInfo(pWal);
ASSERT(pFileInfo != NULL);
if (pFileInfo->firstVer == -1) {
pFileInfo->firstVer = index;
}
pWal->writeHead.head.version = index; pWal->writeHead.head.version = index;
pWal->writeHead.head.bodyLen = bodyLen; pWal->writeHead.head.bodyLen = bodyLen;
pWal->writeHead.head.msgType = msgType; pWal->writeHead.head.msgType = msgType;
@ -437,11 +445,9 @@ static FORCE_INLINE int32_t walWriteImpl(SWal *pWal, int64_t index, tmsg_t msgTy
pWal->writeHead.cksumHead = walCalcHeadCksum(&pWal->writeHead); pWal->writeHead.cksumHead = walCalcHeadCksum(&pWal->writeHead);
pWal->writeHead.cksumBody = walCalcBodyCksum(body, bodyLen); pWal->writeHead.cksumBody = walCalcBodyCksum(body, bodyLen);
wDebug("vgId:%d, wal write log %ld, msgType: %s", pWal->cfg.vgId, index, TMSG_INFO(msgType)); wDebug("vgId:%d, wal write log %ld, msgType: %s", pWal->cfg.vgId, index, TMSG_INFO(msgType));
if (taosWriteFile(pWal->pLogFile, &pWal->writeHead, sizeof(SWalCkHead)) != sizeof(SWalCkHead)) { if (taosWriteFile(pWal->pLogFile, &pWal->writeHead, sizeof(SWalCkHead)) != sizeof(SWalCkHead)) {
// TODO ftruncate
terrno = TAOS_SYSTEM_ERROR(errno); terrno = TAOS_SYSTEM_ERROR(errno);
wError("vgId:%d, file:%" PRId64 ".log, failed to write since %s", pWal->cfg.vgId, walGetLastFileFirstVer(pWal), wError("vgId:%d, file:%" PRId64 ".log, failed to write since %s", pWal->cfg.vgId, walGetLastFileFirstVer(pWal),
strerror(errno)); strerror(errno));
@ -450,7 +456,6 @@ static FORCE_INLINE int32_t walWriteImpl(SWal *pWal, int64_t index, tmsg_t msgTy
} }
if (taosWriteFile(pWal->pLogFile, (char *)body, bodyLen) != bodyLen) { if (taosWriteFile(pWal->pLogFile, (char *)body, bodyLen) != bodyLen) {
// TODO ftruncate
terrno = TAOS_SYSTEM_ERROR(errno); terrno = TAOS_SYSTEM_ERROR(errno);
wError("vgId:%d, file:%" PRId64 ".log, failed to write since %s", pWal->cfg.vgId, walGetLastFileFirstVer(pWal), wError("vgId:%d, file:%" PRId64 ".log, failed to write since %s", pWal->cfg.vgId, walGetLastFileFirstVer(pWal),
strerror(errno)); strerror(errno));
@ -460,7 +465,6 @@ static FORCE_INLINE int32_t walWriteImpl(SWal *pWal, int64_t index, tmsg_t msgTy
code = walWriteIndex(pWal, index, offset); code = walWriteIndex(pWal, index, offset);
if (code < 0) { if (code < 0) {
// TODO ftruncate
goto END; goto END;
} }
@ -468,14 +472,27 @@ static FORCE_INLINE int32_t walWriteImpl(SWal *pWal, int64_t index, tmsg_t msgTy
if (pWal->vers.firstVer == -1) pWal->vers.firstVer = index; if (pWal->vers.firstVer == -1) pWal->vers.firstVer = index;
pWal->vers.lastVer = index; pWal->vers.lastVer = index;
pWal->totSize += sizeof(SWalCkHead) + bodyLen; pWal->totSize += sizeof(SWalCkHead) + bodyLen;
if (walGetCurFileInfo(pWal)->firstVer == -1) { pFileInfo->lastVer = index;
walGetCurFileInfo(pWal)->firstVer = index; pFileInfo->fileSize += sizeof(SWalCkHead) + bodyLen;
}
walGetCurFileInfo(pWal)->lastVer = index;
walGetCurFileInfo(pWal)->fileSize += sizeof(SWalCkHead) + bodyLen;
return 0; return 0;
END: END:
// recover in a reverse order
if (taosFtruncateFile(pWal->pLogFile, offset) < 0) {
wFatal("vgId:%d, failed to ftruncate logfile to offset:%lld during recovery due to %s", pWal->cfg.vgId, offset,
strerror(errno));
terrno = TAOS_SYSTEM_ERROR(errno);
ASSERT(0 && "failed to recover from error");
}
int64_t idxOffset = (index - pFileInfo->firstVer) * sizeof(SWalIdxEntry);
if (taosFtruncateFile(pWal->pIdxFile, idxOffset) < 0) {
wFatal("vgId:%d, failed to ftruncate idxfile to offset:%lld during recovery due to %s", pWal->cfg.vgId, idxOffset,
strerror(errno));
terrno = TAOS_SYSTEM_ERROR(errno);
ASSERT(0 && "failed to recover from error");
}
return -1; return -1;
} }