/*
* smgrclosenode() -- Close SMgrRelation object for given RelFileNode,
* if one exists.
*
* This has the same effects as smgrclose(smgropen(rnode)), but it avoids
* uselessly creating a hashtable entry only to drop it again when no
* such entry exists already.
*/
void
smgrclosenode(RelFileNodeBackend rnode)
{
SMgrRelation reln;
/* Nothing to do if hashtable not set up */
if (SMgrRelationHash == NULL)
return;
reln = (SMgrRelation) hash_search(SMgrRelationHash,
(void *) &rnode,
HASH_FIND, NULL);
if (reln != NULL)
smgrclose(reln);
}
/*
* smgrcloseall() -- Close all existing SMgrRelation objects.
*/
void
smgrcloseall(void)
{
HASH_SEQ_STATUS status;
SMgrRelation reln;
/* Nothing to do if hashtable not set up */
if (SMgrRelationHash == NULL)
return;
hash_seq_init(&status, SMgrRelationHash);
while ((reln = (SMgrRelation) hash_seq_search(&status)) != NULL)
smgrclose(reln);
}
/*
* smgrDoPendingDeletes() -- Take care of relation deletes at end of xact.
*
* This also runs when aborting a subxact; we want to clean up a failed
* subxact immediately.
*/
void
smgrDoPendingDeletes(bool isCommit)
{
int nestLevel = GetCurrentTransactionNestLevel();
PendingRelDelete *pending;
PendingRelDelete *prev;
PendingRelDelete *next;
prev = NULL;
for (pending = pendingDeletes; pending != NULL; pending = next)
{
next = pending->next;
if (pending->nestLevel < nestLevel)
{
/* outer-level entries should not be processed yet */
prev = pending;
}
else
{
/* unlink list entry first, so we don't retry on failure */
if (prev)
prev->next = next;
else
pendingDeletes = next;
/* do deletion if called for */
if (pending->atCommit == isCommit)
{
SMgrRelation srel;
int i;
srel = smgropen(pending->relnode);
for (i = 0; i <= MAX_FORKNUM; i++)
{
if (smgrexists(srel, i))
smgrdounlink(srel,
i,
pending->isTemp,
false);
}
smgrclose(srel);
}
/* must explicitly free the list entry */
pfree(pending);
/* prev does not change */
}
}
}
static void
_xl_remove_hash_entry(XLogRelDesc *rdesc)
{
Form_pg_class tpgc = rdesc->reldata.rd_rel;
XLogRelCacheEntry *hentry;
rdesc->lessRecently->moreRecently = rdesc->moreRecently;
rdesc->moreRecently->lessRecently = rdesc->lessRecently;
hentry = (XLogRelCacheEntry *) hash_search(_xlrelcache,
(void *) &(rdesc->reldata.rd_node), HASH_REMOVE, NULL);
if (hentry == NULL)
elog(PANIC, "_xl_remove_hash_entry: file was not found in cache");
if (rdesc->reldata.rd_fd >= 0)
smgrclose(DEFAULT_SMGR, &(rdesc->reldata));
memset(rdesc, 0, sizeof(XLogRelDesc));
memset(tpgc, 0, sizeof(FormData_pg_class));
rdesc->reldata.rd_rel = tpgc;
return;
}
static int
FileRepPrimary_ResyncBufferPoolIncrementalWrite(ChangeTrackingRequest *request)
{
int status = STATUS_OK;
Page page;
Buffer buf;
BlockNumber numBlocks = 0;
SMgrRelation smgr_relation = NULL;
char relidstr[OIDCHARS + 1 + OIDCHARS + 1 + OIDCHARS + 1];
int ii;
XLogRecPtr loc;
XLogRecPtr loc1;
int count = 0;
int thresholdCount = 0;
bool mirrorDataLossOccurred = FALSE;
int NumberOfRelations = request->count;
FileRepResyncHashEntry_s entry;
ChangeTrackingResult *result = NULL;
while (1)
{
/* allow flushing buffers from buffer pool during scan */
FileRepResync_SetReadBufferRequest();
if ((result = ChangeTracking_GetChanges(request)) != NULL)
{
FileRepResync_ResetReadBufferRequest();
for (ii = 0; ii < result->count; ii++)
{
if (smgr_relation == NULL)
{
NumberOfRelations--;
smgr_relation = smgropen(result->entries[ii].relFileNode);
snprintf(relidstr, sizeof(relidstr), "%u/%u/%u",
smgr_relation->smgr_rnode.spcNode,
smgr_relation->smgr_rnode.dbNode,
smgr_relation->smgr_rnode.relNode);
numBlocks = smgrnblocks(smgr_relation);
if (Debug_filerep_print)
elog(LOG, "resynchronize buffer pool relation '%u/%u/%u' "
"number of blocks:'%u' ",
smgr_relation->smgr_rnode.spcNode,
smgr_relation->smgr_rnode.dbNode,
smgr_relation->smgr_rnode.relNode,
numBlocks);
thresholdCount = Min(numBlocks, 1024);
}
loc1 = result->entries[ii].lsn_end;
/*
* if relation was truncated then block_num from change tracking can be beyond numBlocks
*/
if (result->entries[ii].block_num >= numBlocks)
{
ereport(LOG,
(errmsg("could not resynchonize buffer pool relation '%s' block '%d' (maybe due to truncate), "
"lsn change tracking '%s(%u/%u)' "
"number of blocks '%d' ",
relidstr,
result->entries[ii].block_num,
XLogLocationToString(&loc1),
loc1.xlogid,
loc1.xrecoff,
numBlocks),
FileRep_errcontext()));
goto flush_check;
}
/* allow flushing buffers from buffer pool during scan */
FileRepResync_SetReadBufferRequest();
buf = ReadBuffer_Resync(smgr_relation,
result->entries[ii].block_num,
relidstr);
FileRepResync_ResetReadBufferRequest();
Assert(result->entries[ii].block_num < numBlocks);
LockBuffer(buf, BUFFER_LOCK_EXCLUSIVE);
page = BufferGetPage(buf);
loc = PageGetLSN(page);
if(Debug_filerep_print)
{
elog(LOG,
"incremental resync buffer pool identifier '%s' num blocks '%d' blkno '%d' lsn page '%s(%u/%u)' "
"lsn end change tracking '%s(%u/%u)' ",
relidstr,
numBlocks,
result->entries[ii].block_num,
XLogLocationToString(&loc),
loc.xlogid,
loc.xrecoff,
XLogLocationToString(&loc1),
result->entries[ii].lsn_end.xlogid,
result->entries[ii].lsn_end.xrecoff);
}
else
{
char tmpBuf[FILEREP_MAX_LOG_DESCRIPTION_LEN];
snprintf(tmpBuf, sizeof(tmpBuf),
"incremental resync buffer pool identifier '%s' num blocks '%d' blkno '%d' lsn page '%s(%u/%u)' ",
relidstr,
numBlocks,
result->entries[ii].block_num,
XLogLocationToString(&loc),
loc.xlogid,
loc.xrecoff);
FileRep_InsertConfigLogEntry(tmpBuf);
snprintf(tmpBuf, sizeof(tmpBuf),
"incremental resync buffer pool identifier '%s' lsn end change tracking '%s(%u/%u)' ",
relidstr,
XLogLocationToString(&loc1),
result->entries[ii].lsn_end.xlogid,
result->entries[ii].lsn_end.xrecoff);
FileRep_InsertConfigLogEntry(tmpBuf);
}
if (XLByteLE(result->entries[ii].lsn_end, PageGetLSN(page)))
{
if (! XLByteEQ(PageGetLSN(page), result->entries[ii].lsn_end))
{
ereport(LOG,
(errmsg("Resynchonize buffer pool relation '%s' block '%d' has page lsn less than CT lsn, "
"lsn end change tracking '%s(%u/%u)' lsn page '%s(%u/%u)' "
"number of blocks '%d'",
relidstr,
result->entries[ii].block_num,
XLogLocationToString(&loc),
loc.xlogid,
loc.xrecoff,
XLogLocationToString(&loc1),
loc1.xlogid,
loc1.xrecoff,
numBlocks),
FileRep_errcontext()));
}
/*
* It's safe and better to perform write of the page to mirror,
* for this case, as primary and mirror data pages should always
* be same. So, we might do some extra work but definitely won't
* loose out blocks, or error out and need to perform full recovery.
* Need to cover for this case as there are some known scenarios where
* CT file can have extra records which should have been discarded,
* but as we loose out information of xlog LSN cannot be discarded.
* One such case is when CT_TRANSIENT being compacted to CT_COMPACT
* with specific xlog LSN (to discard extra records) in CT mode gets
* interrupted by resync. Compaction during Resync collects all the
* CT records and doesn't have xlog LSN information to discard any
* extra records from CT_TRANSIENT.
*/
smgrwrite(smgr_relation,
result->entries[ii].block_num,
(char *)BufferGetBlock(buf),
FALSE);
}
#ifdef FAULT_INJECTOR
FaultInjector_InjectFaultIfSet(
FileRepResyncWorker,
DDLNotSpecified,
"", // databaseName
""); // tableName
#endif
UnlockReleaseBuffer(buf);
#ifdef FAULT_INJECTOR
FaultInjector_InjectFaultIfSet(
FileRepResyncWorker,
DDLNotSpecified,
"", // databaseName
""); // tableName
#endif
flush_check:
if (((ii + 1) == result->count) ||
! (result->entries[ii].relFileNode.spcNode == result->entries[ii+1].relFileNode.spcNode &&
result->entries[ii].relFileNode.dbNode == result->entries[ii+1].relFileNode.dbNode &&
result->entries[ii].relFileNode.relNode == result->entries[ii+1].relFileNode.relNode))
{
if (result->ask_for_more == false)
{
smgrimmedsync(smgr_relation);
smgrclose(smgr_relation);
smgr_relation = NULL;
FileRep_GetRelationPath(
entry.fileName,
result->entries[ii].relFileNode,
0 /* segment file number is always 0 for Buffer Pool */);
status = FileRepResync_UpdateEntry(&entry);
if (status != STATUS_OK)
{
break;
}
}
}
if (count > thresholdCount)
{
count = 0;
FileRepSubProcess_ProcessSignals();
if (! (FileRepSubProcess_GetState() == FileRepStateReady &&
dataState == DataStateInResync))
{
mirrorDataLossOccurred = TRUE;
break;
}
}
else
count++;
} // for (ii = 0; ii < result->count; ii++)
} // if ((result = ChangeTracking_GetChanges(request)) != NULL)
FileRepResync_ResetReadBufferRequest();
if (result != NULL && result->ask_for_more == true)
{
Assert(request->count == 1);
request->entries[0].lsn_start = result->next_start_lsn;
}
else
{
break;
}
} // while(1)
ChangeTracking_FreeRequest(request);
ChangeTracking_FreeResult(result);
Insist(NumberOfRelations == 0);
if (mirrorDataLossOccurred)
status = STATUS_ERROR;
return status;
}
static int
FileRepPrimary_ResyncWrite(FileRepResyncHashEntry_s *entry)
{
int status = STATUS_OK;
Page page;
Buffer buf;
BlockNumber numBlocks;
BlockNumber blkno;
SMgrRelation smgr_relation;
char relidstr[OIDCHARS + 1 + OIDCHARS + 1 + OIDCHARS + 1];
XLogRecPtr loc;
int count = 0;
int thresholdCount = 0;
bool mirrorDataLossOccurred = FALSE;
switch (entry->relStorageMgr)
{
case PersistentFileSysRelStorageMgr_BufferPool:
switch (entry->mirrorDataSynchronizationState)
{
case MirroredRelDataSynchronizationState_BufferPoolScanIncremental:
case MirroredRelDataSynchronizationState_FullCopy:
smgr_relation = smgropen(entry->relFileNode);
numBlocks = smgrnblocks(smgr_relation);
snprintf(relidstr, sizeof(relidstr), "%u/%u/%u",
smgr_relation->smgr_rnode.spcNode,
smgr_relation->smgr_rnode.dbNode,
smgr_relation->smgr_rnode.relNode);
if (Debug_filerep_print)
elog(LOG, "resync buffer pool relation '%s' number of blocks '%d' ",
relidstr, numBlocks);
thresholdCount = Min(numBlocks, 1024);
/*
* required in order to report how many blocks were synchronized
* if gp_persistent_relation_node does not return that information
*/
if (entry->mirrorBufpoolResyncChangedPageCount == 0)
{
entry->mirrorBufpoolResyncChangedPageCount = numBlocks - entry->mirrorBufpoolResyncCkptBlockNum;
}
for (blkno = entry->mirrorBufpoolResyncCkptBlockNum; blkno < numBlocks; blkno++)
{
XLogRecPtr endResyncLSN = (isFullResync() ?
FileRepResync_GetEndFullResyncLSN() :
FileRepResync_GetEndIncrResyncLSN());
#ifdef FAULT_INJECTOR
FaultInjector_InjectFaultIfSet(
FileRepResyncWorkerRead,
DDLNotSpecified,
"", //databaseName
""); // tableName
#endif
FileRepResync_SetReadBufferRequest();
buf = ReadBuffer_Resync(smgr_relation, blkno, relidstr);
FileRepResync_ResetReadBufferRequest();
LockBuffer(buf, BUFFER_LOCK_EXCLUSIVE);
page = BufferGetPage(buf);
loc = PageGetLSN(page);
if (Debug_filerep_print)
{
elog(LOG,
"full resync buffer pool identifier '%s' num blocks '%d' blkno '%d' lsn begin change tracking '%s(%u/%u)' "
"lsn page '%s(%u/%u)' lsn end change tracking '%s(%u/%u)' ",
relidstr,
numBlocks,
blkno,
XLogLocationToString(&entry->mirrorBufpoolResyncCkptLoc),
entry->mirrorBufpoolResyncCkptLoc.xlogid,
entry->mirrorBufpoolResyncCkptLoc.xrecoff,
XLogLocationToString(&loc),
loc.xlogid,
loc.xrecoff,
XLogLocationToString(&endResyncLSN),
endResyncLSN.xlogid,
endResyncLSN.xrecoff);
}
else
{
char tmpBuf[FILEREP_MAX_LOG_DESCRIPTION_LEN];
snprintf(tmpBuf, sizeof(tmpBuf),
"full resync buffer pool identifier '%s' num blocks '%d' blkno '%d' lsn begin change tracking '%s(%u/%u)' ",
relidstr,
numBlocks,
blkno,
XLogLocationToString(&entry->mirrorBufpoolResyncCkptLoc),
entry->mirrorBufpoolResyncCkptLoc.xlogid,
entry->mirrorBufpoolResyncCkptLoc.xrecoff);
FileRep_InsertConfigLogEntry(tmpBuf);
snprintf(tmpBuf, sizeof(tmpBuf),
"full resync buffer pool identifier '%s' lsn page '%s(%u/%u)' lsn end change tracking '%s(%u/%u)' ",
relidstr,
XLogLocationToString(&loc),
loc.xlogid,
loc.xrecoff,
XLogLocationToString(&endResyncLSN),
endResyncLSN.xlogid,
endResyncLSN.xrecoff);
FileRep_InsertConfigLogEntry(tmpBuf);
}
if (XLByteLE(PageGetLSN(page), endResyncLSN) &&
XLByteLE(entry->mirrorBufpoolResyncCkptLoc, PageGetLSN(page)))
{
smgrwrite(smgr_relation,
blkno,
(char *)BufferGetBlock(buf),
FALSE);
}
#ifdef FAULT_INJECTOR
FaultInjector_InjectFaultIfSet(
FileRepResyncWorker,
DDLNotSpecified,
"", // databaseName
""); // tableName
#endif
UnlockReleaseBuffer(buf);
if (count > thresholdCount)
{
count = 0;
FileRepSubProcess_ProcessSignals();
if (! (FileRepSubProcess_GetState() == FileRepStateReady &&
dataState == DataStateInResync))
{
mirrorDataLossOccurred = TRUE;
break;
}
}
else
count++;
}
if (mirrorDataLossOccurred)
break;
if (entry->mirrorDataSynchronizationState != MirroredRelDataSynchronizationState_FullCopy)
{
LockRelationForResyncExtension(&smgr_relation->smgr_rnode, ExclusiveLock);
numBlocks = smgrnblocks(smgr_relation);
smgrtruncate(smgr_relation,
numBlocks,
TRUE /* isTemp, TRUE means to not record in XLOG */,
FALSE /* isLocalBuf */,
&entry->persistentTid,
entry->persistentSerialNum);
UnlockRelationForResyncExtension(&smgr_relation->smgr_rnode, ExclusiveLock);
}
smgrimmedsync(smgr_relation);
smgrclose(smgr_relation);
smgr_relation = NULL;
break;
case MirroredRelDataSynchronizationState_None:
case MirroredRelDataSynchronizationState_DataSynchronized:
break;
default:
ereport(LOG,
(errmsg("could not resynchronize relation '%u/%u/%u' "
"mirror synchronization state:'%s(%d)' ",
entry->relFileNode.relNode,
entry->relFileNode.spcNode,
entry->relFileNode.dbNode,
MirroredRelDataSynchronizationState_Name(entry->mirrorDataSynchronizationState),
entry->mirrorDataSynchronizationState)));
break;
}
break;
case PersistentFileSysRelStorageMgr_AppendOnly:
{
MirroredAppendOnlyOpen mirroredOpen;
int primaryError;
bool mirrorDataLossOccurred;
char *buffer = NULL;
int64 endOffset = entry->mirrorAppendOnlyNewEof;
int64 startOffset = entry->mirrorAppendOnlyLossEof;
int32 bufferLen = 0;
int retval = 0;
switch (entry->mirrorDataSynchronizationState)
{
case MirroredRelDataSynchronizationState_AppendOnlyCatchup:
case MirroredRelDataSynchronizationState_FullCopy:
/*
* required in order to report how many blocks were synchronized
* if gp_persistent_relation_node does not return that information
*/
if (entry->mirrorBufpoolResyncChangedPageCount == 0)
{
entry->mirrorBufpoolResyncChangedPageCount = (endOffset - startOffset) / BLCKSZ;
}
/*
* The MirroredAppendOnly_OpenResynchonize routine knows we are a resynch worker and
* will open BOTH, but write only the MIRROR!!!
*/
MirroredAppendOnly_OpenResynchonize(
&mirroredOpen,
&entry->relFileNode,
entry->segmentFileNum,
startOffset,
&primaryError,
&mirrorDataLossOccurred);
if (primaryError != 0)
{
ereport(ERROR,
(errcode_for_file_access(),
errmsg("could not open file %u/%u/%u.%u : %s",
entry->relFileNode.dbNode,
entry->relFileNode.spcNode,
entry->relFileNode.relNode,
entry->segmentFileNum,
strerror(primaryError))));
break;
}
if (mirrorDataLossOccurred)
break;
/* AO and CO Data Store writes 64k size by default */
bufferLen = (Size) Min(2*BLCKSZ, endOffset - startOffset);
buffer = (char*) palloc(bufferLen);
if (buffer == NULL)
ereport(ERROR,
(errcode(ERRCODE_OUT_OF_MEMORY),
(errmsg("not enough memory for resynchronization"))));
MemSet(buffer, 0, bufferLen);
while (startOffset < endOffset)
{
retval = MirroredAppendOnly_Read(
&mirroredOpen,
buffer,
bufferLen);
if (retval != bufferLen)
{
ereport(ERROR,
(errcode_for_file_access(),
errmsg("could not read from position:" INT64_FORMAT " in file %u/%u/%u.%u : %m",
startOffset,
entry->relFileNode.dbNode,
entry->relFileNode.spcNode,
entry->relFileNode.relNode,
entry->segmentFileNum)));
break;
}
MirroredAppendOnly_Append(
&mirroredOpen,
buffer,
bufferLen,
&primaryError,
&mirrorDataLossOccurred);
if (mirrorDataLossOccurred)
break;
Assert(primaryError == 0); // No primary writes as resync worker.
startOffset += bufferLen;
/* AO and CO Data Store writes 64k size by default */
bufferLen = (Size) Min(2*BLCKSZ, endOffset - startOffset);
}
if (buffer)
{
pfree(buffer);
buffer = NULL;
}
if (mirrorDataLossOccurred)
break;
/* Flush written data on Mirror */
MirroredAppendOnly_Flush(
&mirroredOpen,
&primaryError,
&mirrorDataLossOccurred);
if (mirrorDataLossOccurred)
break;
Assert(primaryError == 0); // Not flushed on primary as resync worker.
/* Close Primary and Mirror */
MirroredAppendOnly_Close(
&mirroredOpen,
&mirrorDataLossOccurred);
break;
case MirroredRelDataSynchronizationState_None:
case MirroredRelDataSynchronizationState_DataSynchronized:
break;
default:
ereport(LOG,
(errmsg("could not resynchronize relation '%u/%u/%u' "
"mirror synchronization state:'%s(%d)' ",
entry->relFileNode.relNode,
entry->relFileNode.spcNode,
entry->relFileNode.dbNode,
MirroredRelDataSynchronizationState_Name(entry->mirrorDataSynchronizationState),
entry->mirrorDataSynchronizationState)));
break;
}
break;
} //case
default:
Assert(0);
break;
} //switch
if (mirrorDataLossOccurred)
status = STATUS_ERROR;
return status;
}
/*
* FinishPreparedTransaction: execute COMMIT PREPARED or ROLLBACK PREPARED
*/
void
FinishPreparedTransaction(const char *gid, bool isCommit)
{
GlobalTransaction gxact;
TransactionId xid;
char *buf;
char *bufptr;
TwoPhaseFileHeader *hdr;
TransactionId latestXid;
TransactionId *children;
RelFileNode *commitrels;
RelFileNode *abortrels;
RelFileNode *delrels;
int ndelrels;
int i;
/*
* Validate the GID, and lock the GXACT to ensure that two backends do not
* try to commit the same GID at once.
*/
gxact = LockGXact(gid, GetUserId());
xid = gxact->proc.xid;
/*
* Read and validate the state file
*/
buf = ReadTwoPhaseFile(xid);
if (buf == NULL)
ereport(ERROR,
(errcode(ERRCODE_DATA_CORRUPTED),
errmsg("two-phase state file for transaction %u is corrupt",
xid)));
/*
* Disassemble the header area
*/
hdr = (TwoPhaseFileHeader *) buf;
Assert(TransactionIdEquals(hdr->xid, xid));
bufptr = buf + MAXALIGN(sizeof(TwoPhaseFileHeader));
children = (TransactionId *) bufptr;
bufptr += MAXALIGN(hdr->nsubxacts * sizeof(TransactionId));
commitrels = (RelFileNode *) bufptr;
bufptr += MAXALIGN(hdr->ncommitrels * sizeof(RelFileNode));
abortrels = (RelFileNode *) bufptr;
bufptr += MAXALIGN(hdr->nabortrels * sizeof(RelFileNode));
/* compute latestXid among all children */
latestXid = TransactionIdLatest(xid, hdr->nsubxacts, children);
/*
* The order of operations here is critical: make the XLOG entry for
* commit or abort, then mark the transaction committed or aborted in
* pg_clog, then remove its PGPROC from the global ProcArray (which means
* TransactionIdIsInProgress will stop saying the prepared xact is in
* progress), then run the post-commit or post-abort callbacks. The
* callbacks will release the locks the transaction held.
*/
if (isCommit)
RecordTransactionCommitPrepared(xid,
hdr->nsubxacts, children,
hdr->ncommitrels, commitrels);
else
RecordTransactionAbortPrepared(xid,
hdr->nsubxacts, children,
hdr->nabortrels, abortrels);
ProcArrayRemove(&gxact->proc, latestXid);
/*
* In case we fail while running the callbacks, mark the gxact invalid so
* no one else will try to commit/rollback, and so it can be recycled
* properly later. It is still locked by our XID so it won't go away yet.
*
* (We assume it's safe to do this without taking TwoPhaseStateLock.)
*/
gxact->valid = false;
/*
* We have to remove any files that were supposed to be dropped. For
* consistency with the regular xact.c code paths, must do this before
* releasing locks, so do it before running the callbacks.
*
* NB: this code knows that we couldn't be dropping any temp rels ...
*/
if (isCommit)
{
delrels = commitrels;
ndelrels = hdr->ncommitrels;
}
else
{
delrels = abortrels;
ndelrels = hdr->nabortrels;
}
for (i = 0; i < ndelrels; i++)
{
SMgrRelation srel = smgropen(delrels[i]);
ForkNumber fork;
for (fork = 0; fork <= MAX_FORKNUM; fork++)
{
if (smgrexists(srel, fork))
smgrdounlink(srel, fork, false, false);
}
smgrclose(srel);
}
/* And now do the callbacks */
if (isCommit)
ProcessRecords(bufptr, xid, twophase_postcommit_callbacks);
else
ProcessRecords(bufptr, xid, twophase_postabort_callbacks);
/* Count the prepared xact as committed or aborted */
AtEOXact_PgStat(isCommit);
/*
* And now we can clean up our mess.
*/
RemoveTwoPhaseFile(xid, true);
RemoveGXact(gxact);
pfree(buf);
}
static int64
PersistentBuild_TruncateAllGpRelationNode(void)
{
Relation pg_database;
HeapScanDesc scan;
HeapTuple tuple;
int64 count;
pg_database = heap_open(
DatabaseRelationId,
AccessShareLock);
/*
* Truncate gp_relation_node and its index in each database.
*/
scan = heap_beginscan(pg_database, SnapshotNow, 0, NULL);
count = 0;
while ((tuple = heap_getnext(scan, ForwardScanDirection)) != NULL)
{
Form_pg_database form_pg_database =
(Form_pg_database)GETSTRUCT(tuple);
Oid dbOid;
Oid dattablespace;
RelFileNode relFileNode;
SMgrRelation smgrRelation;
Page btree_metapage;
dbOid = HeapTupleGetOid(tuple);
dattablespace = form_pg_database->dattablespace;
if (dbOid == HcatalogDbOid)
continue;
if (Debug_persistent_print)
elog(Persistent_DebugPrintLevel(),
"PersistentBuild_TruncateAllGpRelationNode: dbOid %u, '%s'",
dbOid,
form_pg_database->datname.data);
if (Debug_persistent_print)
elog(Persistent_DebugPrintLevel(),
"Truncating gp_relation_node %u/%u/%u in database oid %u ('%s')",
relFileNode.spcNode,
relFileNode.dbNode,
relFileNode.relNode,
dbOid,
form_pg_database->datname.data);
relFileNode.spcNode = dattablespace;
relFileNode.dbNode = dbOid;
relFileNode.relNode = GpRelfileNodeRelationId;
/*
* Truncate WITHOUT generating an XLOG record (i.e. pretend it is a temp relation).
*/
PersistentBuild_NonTransactionTruncate(&relFileNode);
count++;
/*
* And, the index. Unfortunately, the relfilenode OID can change due to a
* REINDEX {TABLE|INDEX} command.
*/
PersistentBuild_FindGpRelationNodeIndex(
dbOid,
dattablespace,
&relFileNode);
if (Debug_persistent_print)
elog(Persistent_DebugPrintLevel(),
"Truncating gp_relation_node_index %u/%u/%u in database oid %u ('%s'). relfilenode different %s, tablespace different %s",
relFileNode.spcNode,
relFileNode.dbNode,
relFileNode.relNode,
dbOid,
form_pg_database->datname.data,
((relFileNode.relNode != GpRelfileNodeOidIndexId) ? "true" : "false"),
((relFileNode.spcNode != dattablespace) ? "true" : "false"));
PersistentBuild_NonTransactionTruncate(&relFileNode);
// The BTree needs an empty meta-data block.
smgrRelation = smgropen(relFileNode);
btree_metapage = (Page)palloc(BLCKSZ);
_bt_initmetapage(btree_metapage, P_NONE, 0);
smgrwrite(
smgrRelation,
/* blockNum */ 0,
(char*)btree_metapage,
/* isTemp */ false);
smgrimmedsync(smgrRelation);
pfree(btree_metapage);
smgrclose(smgrRelation);
count++;
}
heap_endscan(scan);
heap_close(pg_database, AccessShareLock);
return count;
}
/*
* smgrDoPendingDeletes() -- Take care of relation deletes at end of xact.
*
* This also runs when aborting a subxact; we want to clean up a failed
* subxact immediately.
*
* Note: It's possible that we're being asked to remove a relation that has
* no physical storage in any fork. In particular, it's possible that we're
* cleaning up an old temporary relation for which RemovePgTempFiles has
* already recovered the physical storage.
*/
void
smgrDoPendingDeletes(bool isCommit)
{
int nestLevel = GetCurrentTransactionNestLevel();
PendingRelDelete *pending;
PendingRelDelete *prev;
PendingRelDelete *next;
int nrels = 0,
i = 0,
maxrels = 0;
SMgrRelation *srels = NULL;
prev = NULL;
for (pending = pendingDeletes; pending != NULL; pending = next)
{
next = pending->next;
if (pending->nestLevel < nestLevel)
{
/* outer-level entries should not be processed yet */
prev = pending;
}
else
{
/* unlink list entry first, so we don't retry on failure */
if (prev)
prev->next = next;
else
pendingDeletes = next;
/* do deletion if called for */
if (pending->atCommit == isCommit)
{
SMgrRelation srel;
srel = smgropen(pending->relnode, pending->backend);
/* allocate the initial array, or extend it, if needed */
if (maxrels == 0)
{
maxrels = 8;
srels = palloc(sizeof(SMgrRelation) * maxrels);
}
else if (maxrels <= nrels)
{
maxrels *= 2;
srels = repalloc(srels, sizeof(SMgrRelation) * maxrels);
}
srels[nrels++] = srel;
}
/* must explicitly free the list entry */
pfree(pending);
/* prev does not change */
}
}
if (nrels > 0)
{
smgrdounlinkall(srels, nrels, false);
for (i = 0; i < nrels; i++)
smgrclose(srels[i]);
pfree(srels);
}
}
