提交 62377619 编写于 作者: X Xiaoyu Wang

enh: optimizing stt file reading in last query

上级 25778754
......@@ -1136,6 +1136,7 @@ typedef struct {
int64_t numOfInsertSuccessReqs;
int64_t numOfBatchInsertReqs;
int64_t numOfBatchInsertSuccessReqs;
int32_t numOfCacheTables;
} SVnodeLoad;
typedef struct {
......
......@@ -55,6 +55,8 @@ void *taosLRUCacheValue(SLRUCache *cache, LRUHandle *handle);
size_t taosLRUCacheGetUsage(SLRUCache *cache);
size_t taosLRUCacheGetPinnedUsage(SLRUCache *cache);
int32_t taosLRUCacheGetElems(SLRUCache *cache);
void taosLRUCacheSetCapacity(SLRUCache *cache, size_t capacity);
size_t taosLRUCacheGetCapacity(SLRUCache *cache);
......
......@@ -198,9 +198,10 @@ int32_t tsdbRetrieveCacheRows(void *pReader, SSDataBlock *pResBlock, const int32
void *tsdbCacherowsReaderClose(void *pReader);
int32_t tsdbGetTableSchema(SVnode *pVnode, int64_t uid, STSchema **pSchema, int64_t *suid);
void tsdbCacheSetCapacity(SVnode *pVnode, size_t capacity);
size_t tsdbCacheGetCapacity(SVnode *pVnode);
size_t tsdbCacheGetUsage(SVnode *pVnode);
void tsdbCacheSetCapacity(SVnode *pVnode, size_t capacity);
size_t tsdbCacheGetCapacity(SVnode *pVnode);
size_t tsdbCacheGetUsage(SVnode *pVnode);
int32_t tsdbCacheGetElems(SVnode *pVnode);
// tq
typedef struct SMetaTableInfo {
......@@ -264,7 +265,7 @@ int32_t tqReaderSetTbUidList(STqReader *pReader, const SArray *tbUidList);
int32_t tqReaderAddTbUidList(STqReader *pReader, const SArray *tbUidList);
int32_t tqReaderRemoveTbUidList(STqReader *pReader, const SArray *tbUidList);
int32_t tqSeekVer(STqReader *pReader, int64_t ver, const char* id);
int32_t tqSeekVer(STqReader *pReader, int64_t ver, const char *id);
int32_t tqNextBlock(STqReader *pReader, SFetchRet *ret);
int32_t tqReaderSetSubmitReq2(STqReader *pReader, void *msgStr, int32_t msgLen, int64_t ver);
......
......@@ -748,7 +748,7 @@ struct SDiskDataBuilder {
int32_t tMergeTreeOpen(SMergeTree *pMTree, int8_t backward, SDataFReader *pFReader, uint64_t suid, uint64_t uid,
STimeWindow *pTimeWindow, SVersionRange *pVerRange, SSttBlockLoadInfo *pBlockLoadInfo,
bool destroyLoadInfo, const char *idStr);
bool destroyLoadInfo, const char *idStr, bool strictTimeRange);
void tMergeTreeAddIter(SMergeTree *pMTree, SLDataIter *pIter);
bool tMergeTreeNext(SMergeTree *pMTree);
TSDBROW tMergeTreeGetRow(SMergeTree *pMTree);
......
......@@ -632,8 +632,8 @@ static int32_t getNextRowFromFSLast(void *iter, TSDBROW **ppRow, bool *pIgnoreEa
}
tMergeTreeOpen(&state->mergeTree, 1, *state->pDataFReader, state->suid, state->uid,
&(STimeWindow){.skey = TSKEY_MIN, .ekey = TSKEY_MAX},
&(SVersionRange){.minVer = 0, .maxVer = UINT64_MAX}, state->pLoadInfo, false, NULL);
&(STimeWindow){.skey = state->lastTs, .ekey = TSKEY_MAX},
&(SVersionRange){.minVer = 0, .maxVer = UINT64_MAX}, state->pLoadInfo, false, NULL, true);
state->pMergeTree = &state->mergeTree;
bool hasVal = tMergeTreeNext(&state->mergeTree);
if (!hasVal) {
......@@ -1727,6 +1727,15 @@ size_t tsdbCacheGetUsage(SVnode *pVnode) {
return usage;
}
int32_t tsdbCacheGetElems(SVnode *pVnode) {
int32_t elems = 0;
if (pVnode->pTsdb != NULL) {
elems = taosLRUCacheGetElems(pVnode->pTsdb->lruCache);
}
return elems;
}
static void getBICacheKey(int32_t fid, int64_t commitID, char *key, int *len) {
struct {
int32_t fid;
......
......@@ -408,6 +408,7 @@ int32_t tsdbRetrieveCacheRows(void* pReader, SSDataBlock* pResBlock, const int32
if (hasNotNullRow) {
pr->lastTs = minTs;
tsdbInfo("%p have cache read %d tables", pr, i + 1);
}
}
......@@ -418,6 +419,7 @@ int32_t tsdbRetrieveCacheRows(void* pReader, SSDataBlock* pResBlock, const int32
saveOneRow(pLastCols, pResBlock, pr, slotIds, pRes, pr->idstr);
}
tsdbInfo("have cached %d tables", taosLRUCacheGetElems(lruCache));
} else if (HASTYPE(pr->type, CACHESCAN_RETRIEVE_TYPE_ALL)) {
for (int32_t i = pr->tableIndex; i < pr->numOfTables; ++i) {
STableKeyInfo* pKeyInfo = &pr->pTableList[i];
......
......@@ -31,7 +31,8 @@ struct SLDataIter {
SSttBlockLoadInfo *pBlockLoadInfo;
};
SSttBlockLoadInfo *tCreateLastBlockLoadInfo(STSchema *pSchema, int16_t *colList, int32_t numOfCols, int32_t numOfSttTrigger) {
SSttBlockLoadInfo *tCreateLastBlockLoadInfo(STSchema *pSchema, int16_t *colList, int32_t numOfCols,
int32_t numOfSttTrigger) {
SSttBlockLoadInfo *pLoadInfo = taosMemoryCalloc(numOfSttTrigger, sizeof(SSttBlockLoadInfo));
if (pLoadInfo == NULL) {
terrno = TSDB_CODE_OUT_OF_MEMORY;
......@@ -162,7 +163,8 @@ static SBlockData *loadLastBlock(SLDataIter *pIter, const char *idStr) {
pInfo->blockIndex[pInfo->currentLoadBlockIndex] = pIter->iSttBlk;
pIter->iRow = (pIter->backward) ? pInfo->blockData[pInfo->currentLoadBlockIndex].nRow : -1;
tsdbDebug("last block index list:%d, %d, rowIndex:%d %s", pInfo->blockIndex[0], pInfo->blockIndex[1], pIter->iRow, idStr);
tsdbDebug("last block index list:%d, %d, rowIndex:%d %s", pInfo->blockIndex[0], pInfo->blockIndex[1], pIter->iRow,
idStr);
return &pInfo->blockData[pInfo->currentLoadBlockIndex];
_exit:
......@@ -263,7 +265,7 @@ static int32_t binarySearchForStartRowIndex(uint64_t *uidList, int32_t num, uint
int32_t tLDataIterOpen(struct SLDataIter **pIter, SDataFReader *pReader, int32_t iStt, int8_t backward, uint64_t suid,
uint64_t uid, STimeWindow *pTimeWindow, SVersionRange *pRange, SSttBlockLoadInfo *pBlockLoadInfo,
const char *idStr) {
const char *idStr, bool strictTimeRange) {
int32_t code = TSDB_CODE_SUCCESS;
*pIter = taosMemoryCalloc(1, sizeof(SLDataIter));
......@@ -340,6 +342,16 @@ int32_t tLDataIterOpen(struct SLDataIter **pIter, SDataFReader *pReader, int32_t
if ((*pIter)->iSttBlk != -1) {
(*pIter)->pSttBlk = taosArrayGet(pBlockLoadInfo->aSttBlk, (*pIter)->iSttBlk);
(*pIter)->iRow = ((*pIter)->backward) ? (*pIter)->pSttBlk->nRow : -1;
if ((!backward) && ((strictTimeRange && (*pIter)->pSttBlk->minKey >= (*pIter)->timeWindow.ekey) ||
(!strictTimeRange && (*pIter)->pSttBlk->minKey > (*pIter)->timeWindow.ekey))) {
(*pIter)->pSttBlk = NULL;
}
if (backward && ((strictTimeRange && (*pIter)->pSttBlk->maxKey <= (*pIter)->timeWindow.skey) ||
(!strictTimeRange && (*pIter)->pSttBlk->maxKey < (*pIter)->timeWindow.skey))) {
(*pIter)->pSttBlk = NULL;
}
}
return code;
......@@ -421,7 +433,7 @@ static void findNextValidRow(SLDataIter *pIter, const char *idStr) {
pBlockData->aUid != NULL) {
i = binarySearchForStartRowIndex((uint64_t *)pBlockData->aUid, pBlockData->nRow, pIter->uid, pIter->backward);
if (i == -1) {
tsdbDebug("failed to find the data in pBlockData, uid:%"PRIu64" , %s", pIter->uid, idStr);
tsdbDebug("failed to find the data in pBlockData, uid:%" PRIu64 " , %s", pIter->uid, idStr);
pIter->iRow = -1;
return;
}
......@@ -508,7 +520,7 @@ bool tLDataIterNextRow(SLDataIter *pIter, const char *idStr) {
}
// set start row index
pIter->iRow = pIter->backward? pBlockData->nRow-1:0;
pIter->iRow = pIter->backward ? pBlockData->nRow - 1 : 0;
}
}
......@@ -551,7 +563,7 @@ static FORCE_INLINE int32_t tLDataIterDescCmprFn(const SRBTreeNode *p1, const SR
int32_t tMergeTreeOpen(SMergeTree *pMTree, int8_t backward, SDataFReader *pFReader, uint64_t suid, uint64_t uid,
STimeWindow *pTimeWindow, SVersionRange *pVerRange, SSttBlockLoadInfo *pBlockLoadInfo,
bool destroyLoadInfo, const char *idStr) {
bool destroyLoadInfo, const char *idStr, bool strictTimeRange) {
pMTree->backward = backward;
pMTree->pIter = NULL;
pMTree->pIterList = taosArrayInit(4, POINTER_BYTES);
......@@ -573,7 +585,7 @@ int32_t tMergeTreeOpen(SMergeTree *pMTree, int8_t backward, SDataFReader *pFRead
for (int32_t i = 0; i < pFReader->pSet->nSttF; ++i) { // open all last file
struct SLDataIter *pIter = NULL;
code = tLDataIterOpen(&pIter, pFReader, i, pMTree->backward, suid, uid, pTimeWindow, pVerRange,
&pMTree->pLoadInfo[i], pMTree->idStr);
&pMTree->pLoadInfo[i], pMTree->idStr, strictTimeRange);
if (code != TSDB_CODE_SUCCESS) {
goto _end;
}
......
......@@ -315,11 +315,11 @@ static int32_t ensureBlockScanInfoBuf(SBlockInfoBuf* pBuf, int32_t numOfTables)
}
if (pBuf->numOfTables > 0) {
STableBlockScanInfo **p = (STableBlockScanInfo**)taosArrayPop(pBuf->pData);
STableBlockScanInfo** p = (STableBlockScanInfo**)taosArrayPop(pBuf->pData);
taosMemoryFree(*p);
pBuf->numOfTables /= pBuf->numPerBucket;
}
int32_t num = (numOfTables - pBuf->numOfTables) / pBuf->numPerBucket;
int32_t remainder = (numOfTables - pBuf->numOfTables) % pBuf->numPerBucket;
if (pBuf->pData == NULL) {
......@@ -919,7 +919,7 @@ static int32_t doLoadFileBlock(STsdbReader* pReader, SArray* pIndexList, SBlockN
pBlockNum->numOfBlocks += 1;
}
if ((pScanInfo->pBlockList != NULL )&& (taosArrayGetSize(pScanInfo->pBlockList) > 0)) {
if ((pScanInfo->pBlockList != NULL) && (taosArrayGetSize(pScanInfo->pBlockList) > 0)) {
numOfQTable += 1;
}
}
......@@ -1798,7 +1798,7 @@ static bool nextRowFromLastBlocks(SLastBlockReader* pLastBlockReader, STableBloc
while (1) {
bool hasVal = tMergeTreeNext(&pLastBlockReader->mergeTree);
if (!hasVal) { // the next value will be the accessed key in stt
if (!hasVal) { // the next value will be the accessed key in stt
pScanInfo->lastKeyInStt += step;
return false;
}
......@@ -2481,7 +2481,7 @@ static bool initLastBlockReader(SLastBlockReader* pLBlockReader, STableBlockScan
pScanInfo->uid, pReader->idStr);
int32_t code = tMergeTreeOpen(&pLBlockReader->mergeTree, (pLBlockReader->order == TSDB_ORDER_DESC),
pReader->pFileReader, pReader->suid, pScanInfo->uid, &w, &pLBlockReader->verRange,
pLBlockReader->pInfo, false, pReader->idStr);
pLBlockReader->pInfo, false, pReader->idStr, false);
if (code != TSDB_CODE_SUCCESS) {
return false;
}
......@@ -3512,7 +3512,7 @@ static int32_t checkForNeighborFileBlock(STsdbReader* pReader, STableBlockScanIn
CHECK_FILEBLOCK_STATE* state) {
SFileBlockDumpInfo* pDumpInfo = &pReader->status.fBlockDumpInfo;
SBlockData* pBlockData = &pReader->status.fileBlockData;
bool asc = ASCENDING_TRAVERSE(pReader->order);
bool asc = ASCENDING_TRAVERSE(pReader->order);
*state = CHECK_FILEBLOCK_QUIT;
int32_t step = ASCENDING_TRAVERSE(pReader->order) ? 1 : -1;
......@@ -3927,7 +3927,8 @@ int32_t tsdbSetTableList(STsdbReader* pReader, const void* pTableList, int32_t n
if (code) {
return code;
}
pReader->status.uidList.tableUidList = (uint64_t*)taosMemoryRealloc(pReader->status.uidList.tableUidList, sizeof(uint64_t) * num);
pReader->status.uidList.tableUidList =
(uint64_t*)taosMemoryRealloc(pReader->status.uidList.tableUidList, sizeof(uint64_t) * num);
}
taosHashClear(pReader->status.pTableMap);
......
......@@ -382,6 +382,7 @@ int32_t vnodeGetLoad(SVnode *pVnode, SVnodeLoad *pLoad) {
pLoad->syncRestore = state.restored;
pLoad->syncCanRead = state.canRead;
pLoad->cacheUsage = tsdbCacheGetUsage(pVnode);
pLoad->numOfCacheTables = tsdbCacheGetElems(pVnode);
pLoad->numOfTables = metaGetTbNum(pVnode->pMeta);
pLoad->numOfTimeSeries = metaGetTimeSeriesNum(pVnode->pMeta);
pLoad->totalStorage = (int64_t)3 * 1073741824;
......
......@@ -580,6 +580,16 @@ static size_t taosLRUCacheShardGetUsage(SLRUCacheShard *shard) {
return usage;
}
static int32_t taosLRUCacheShardGetElems(SLRUCacheShard *shard) {
int32_t elems = 0;
taosThreadMutexLock(&shard->mutex);
elems = shard->table.elems;
taosThreadMutexUnlock(&shard->mutex);
return elems;
}
static size_t taosLRUCacheShardGetPinnedUsage(SLRUCacheShard *shard) {
size_t usage = 0;
......@@ -755,6 +765,16 @@ size_t taosLRUCacheGetUsage(SLRUCache *cache) {
return usage;
}
int32_t taosLRUCacheGetElems(SLRUCache *cache) {
int32_t elems = 0;
for (int i = 0; i < cache->numShards; ++i) {
elems += taosLRUCacheShardGetElems(&cache->shards[i]);
}
return elems;
}
size_t taosLRUCacheGetPinnedUsage(SLRUCache *cache) {
size_t usage = 0;
......
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