database/postgres
1
0
Fork 0
mirror of https://github.com/postgres/postgres.git synced 2025-11-10 17:42:29 +03:00
Code Activity

Fix recently-understood problems with handling of XID freezing, particularly

Browse Source 
in PITR scenarios.  We now WAL-log the replacement of old XIDs with
FrozenTransactionId, so that such replacement is guaranteed to propagate to
PITR slave databases.  Also, rather than relying on hint-bit updates to be
preserved, pg_clog is not truncated until all instances of an XID are known to
have been replaced by FrozenTransactionId.  Add new GUC variables and
pg_autovacuum columns to allow management of the freezing policy, so that
users can trade off the size of pg_clog against the amount of freezing work
done.  Revise the already-existing code that forces autovacuum of tables
approaching the wraparound point to make it more bulletproof; also, revise the
autovacuum logic so that anti-wraparound vacuuming is done per-table rather
than per-database.  initdb forced because of changes in pg_class, pg_database,
and pg_autovacuum catalogs.  Heikki Linnakangas, Simon Riggs, and Tom Lane.
This commit is contained in:
Tom Lane
2006-11-05 22:42:10 +00:00
parent 10c70b8602
commit 48188e1621
43 changed files with 1284 additions and 1062 deletions
Download Patch File Download Diff File Expand all files Collapse all files

297
src/backend/access/heap/heapam.c
View File

@@ -8,7 +8,7 @@
*
*
* IDENTIFICATION
* $PostgreSQL: pgsql/src/backend/access/heap/heapam.c,v 1.220 2006/10/04 00:29:48 momjian Exp $
* $PostgreSQL: pgsql/src/backend/access/heap/heapam.c,v 1.221 2006/11/05 22:42:07 tgl Exp $
*
*
* INTERFACE ROUTINES
@@ -2809,6 +2809,166 @@ heap_inplace_update(Relation relation, HeapTuple tuple)
}
/*
* heap_freeze_tuple
*
* Check to see whether any of the XID fields of a tuple (xmin, xmax, xvac)
* are older than the specified cutoff XID. If so, replace them with
* FrozenTransactionId or InvalidTransactionId as appropriate, and return
* TRUE. Return FALSE if nothing was changed.
*
* It is assumed that the caller has checked the tuple with
* HeapTupleSatisfiesVacuum() and determined that it is not HEAPTUPLE_DEAD
* (else we should be removing the tuple, not freezing it).
*
* NB: cutoff_xid *must* be <= the current global xmin, to ensure that any
* XID older than it could neither be running nor seen as running by any
* open transaction. This ensures that the replacement will not change
* anyone's idea of the tuple state. Also, since we assume the tuple is
* not HEAPTUPLE_DEAD, the fact that an XID is not still running allows us
* to assume that it is either committed good or aborted, as appropriate;
* so we need no external state checks to decide what to do. (This is good
* because this function is applied during WAL recovery, when we don't have
* access to any such state, and can't depend on the hint bits to be set.)
*
* In lazy VACUUM, we call this while initially holding only a shared lock
* on the tuple's buffer. If any change is needed, we trade that in for an
* exclusive lock before making the change. Caller should pass the buffer ID
* if shared lock is held, InvalidBuffer if exclusive lock is already held.
*
* Note: it might seem we could make the changes without exclusive lock, since
* TransactionId read/write is assumed atomic anyway. However there is a race
* condition: someone who just fetched an old XID that we overwrite here could
* conceivably not finish checking the XID against pg_clog before we finish
* the VACUUM and perhaps truncate off the part of pg_clog he needs. Getting
* exclusive lock ensures no other backend is in process of checking the
* tuple status. Also, getting exclusive lock makes it safe to adjust the
* infomask bits.
*/
bool
heap_freeze_tuple(HeapTupleHeader tuple, TransactionId cutoff_xid,
Buffer buf)
{
bool changed = false;
TransactionId xid;
xid = HeapTupleHeaderGetXmin(tuple);
if (TransactionIdIsNormal(xid) &&
TransactionIdPrecedes(xid, cutoff_xid))
{
if (buf != InvalidBuffer)
{
/* trade in share lock for exclusive lock */
LockBuffer(buf, BUFFER_LOCK_UNLOCK);
LockBuffer(buf, BUFFER_LOCK_EXCLUSIVE);
buf = InvalidBuffer;
}
HeapTupleHeaderSetXmin(tuple, FrozenTransactionId);
/*
* Might as well fix the hint bits too; usually XMIN_COMMITTED will
* already be set here, but there's a small chance not.
*/
Assert(!(tuple->t_infomask & HEAP_XMIN_INVALID));
tuple->t_infomask |= HEAP_XMIN_COMMITTED;
changed = true;
}
/*
* When we release shared lock, it's possible for someone else to change
* xmax before we get the lock back, so repeat the check after acquiring
* exclusive lock. (We don't need this pushup for xmin, because only
* VACUUM could be interested in changing an existing tuple's xmin,
* and there's only one VACUUM allowed on a table at a time.)
*/
recheck_xmax:
if (!(tuple->t_infomask & HEAP_XMAX_IS_MULTI))
{
xid = HeapTupleHeaderGetXmax(tuple);
if (TransactionIdIsNormal(xid) &&
TransactionIdPrecedes(xid, cutoff_xid))
{
if (buf != InvalidBuffer)
{
/* trade in share lock for exclusive lock */
LockBuffer(buf, BUFFER_LOCK_UNLOCK);
LockBuffer(buf, BUFFER_LOCK_EXCLUSIVE);
buf = InvalidBuffer;
goto recheck_xmax; /* see comment above */
}
HeapTupleHeaderSetXmax(tuple, InvalidTransactionId);
/*
* The tuple might be marked either XMAX_INVALID or
* XMAX_COMMITTED + LOCKED. Normalize to INVALID just to be
* sure no one gets confused.
*/
tuple->t_infomask &= ~HEAP_XMAX_COMMITTED;
tuple->t_infomask |= HEAP_XMAX_INVALID;
changed = true;
}
}
else
{
/*----------
* XXX perhaps someday we should zero out very old MultiXactIds here?
*
* The only way a stale MultiXactId could pose a problem is if a
* tuple, having once been multiply-share-locked, is not touched by
* any vacuum or attempted lock or deletion for just over 4G MultiXact
* creations, and then in the probably-narrow window where its xmax
* is again a live MultiXactId, someone tries to lock or delete it.
* Even then, another share-lock attempt would work fine. An
* exclusive-lock or delete attempt would face unexpected delay, or
* in the very worst case get a deadlock error. This seems an
* extremely low-probability scenario with minimal downside even if
* it does happen, so for now we don't do the extra bookkeeping that
* would be needed to clean out MultiXactIds.
*----------
*/
}
/*
* Although xvac per se could only be set by VACUUM, it shares physical
* storage space with cmax, and so could be wiped out by someone setting
* xmax. Hence recheck after changing lock, same as for xmax itself.
*/
recheck_xvac:
if (tuple->t_infomask & HEAP_MOVED)
{
xid = HeapTupleHeaderGetXvac(tuple);
if (TransactionIdIsNormal(xid) &&
TransactionIdPrecedes(xid, cutoff_xid))
{
if (buf != InvalidBuffer)
{
/* trade in share lock for exclusive lock */
LockBuffer(buf, BUFFER_LOCK_UNLOCK);
LockBuffer(buf, BUFFER_LOCK_EXCLUSIVE);
buf = InvalidBuffer;
goto recheck_xvac; /* see comment above */
}
/*
* If a MOVED_OFF tuple is not dead, the xvac transaction must
* have failed; whereas a non-dead MOVED_IN tuple must mean the
* xvac transaction succeeded.
*/
if (tuple->t_infomask & HEAP_MOVED_OFF)
HeapTupleHeaderSetXvac(tuple, InvalidTransactionId);
else
HeapTupleHeaderSetXvac(tuple, FrozenTransactionId);
/*
* Might as well fix the hint bits too; usually XMIN_COMMITTED will
* already be set here, but there's a small chance not.
*/
Assert(!(tuple->t_infomask & HEAP_XMIN_INVALID));
tuple->t_infomask |= HEAP_XMIN_COMMITTED;
changed = true;
}
}
return changed;
}
/* ----------------
* heap_markpos - mark scan position
* ----------------
@@ -2877,6 +3037,9 @@ heap_restrpos(HeapScanDesc scan)
/*
* Perform XLogInsert for a heap-clean operation. Caller must already
* have modified the buffer and marked it dirty.
*
* Note: for historical reasons, the entries in the unused[] array should
* be zero-based tuple indexes, not one-based.
*/
XLogRecPtr
log_heap_clean(Relation reln, Buffer buffer, OffsetNumber *unused, int uncnt)
@@ -2920,6 +3083,57 @@ log_heap_clean(Relation reln, Buffer buffer, OffsetNumber *unused, int uncnt)
return recptr;
}
/*
* Perform XLogInsert for a heap-freeze operation. Caller must already
* have modified the buffer and marked it dirty.
*
* Unlike log_heap_clean(), the offsets[] entries are one-based.
*/
XLogRecPtr
log_heap_freeze(Relation reln, Buffer buffer,
TransactionId cutoff_xid,
OffsetNumber *offsets, int offcnt)
{
xl_heap_freeze xlrec;
XLogRecPtr recptr;
XLogRecData rdata[2];
/* Caller should not call me on a temp relation */
Assert(!reln->rd_istemp);
xlrec.node = reln->rd_node;
xlrec.block = BufferGetBlockNumber(buffer);
xlrec.cutoff_xid = cutoff_xid;
rdata[0].data = (char *) &xlrec;
rdata[0].len = SizeOfHeapFreeze;
rdata[0].buffer = InvalidBuffer;
rdata[0].next = &(rdata[1]);
/*
* The tuple-offsets array is not actually in the buffer, but pretend
* that it is. When XLogInsert stores the whole buffer, the offsets array
* need not be stored too.
*/
if (offcnt > 0)
{
rdata[1].data = (char *) offsets;
rdata[1].len = offcnt * sizeof(OffsetNumber);
}
else
{
rdata[1].data = NULL;
rdata[1].len = 0;
}
rdata[1].buffer = buffer;
rdata[1].buffer_std = true;
rdata[1].next = NULL;
recptr = XLogInsert(RM_HEAP2_ID, XLOG_HEAP2_FREEZE, rdata);
return recptr;
}
/*
* Perform XLogInsert for a heap-update operation. Caller must already
* have modified the buffer(s) and marked them dirty.
@@ -3057,6 +3271,7 @@ heap_xlog_clean(XLogRecPtr lsn, XLogRecord *record)
while (unused < unend)
{
/* unused[] entries are zero-based */
lp = PageGetItemId(page, *unused + 1);
lp->lp_flags &= ~LP_USED;
unused++;
@@ -3071,6 +3286,55 @@ heap_xlog_clean(XLogRecPtr lsn, XLogRecord *record)
UnlockReleaseBuffer(buffer);
}
static void
heap_xlog_freeze(XLogRecPtr lsn, XLogRecord *record)
{
xl_heap_freeze *xlrec = (xl_heap_freeze *) XLogRecGetData(record);
TransactionId cutoff_xid = xlrec->cutoff_xid;
Relation reln;
Buffer buffer;
Page page;
if (record->xl_info & XLR_BKP_BLOCK_1)
return;
reln = XLogOpenRelation(xlrec->node);
buffer = XLogReadBuffer(reln, xlrec->block, false);
if (!BufferIsValid(buffer))
return;
page = (Page) BufferGetPage(buffer);
if (XLByteLE(lsn, PageGetLSN(page)))
{
UnlockReleaseBuffer(buffer);
return;
}
if (record->xl_len > SizeOfHeapFreeze)
{
OffsetNumber *offsets;
OffsetNumber *offsets_end;
offsets = (OffsetNumber *) ((char *) xlrec + SizeOfHeapFreeze);
offsets_end = (OffsetNumber *) ((char *) xlrec + record->xl_len);
while (offsets < offsets_end)
{
/* offsets[] entries are one-based */
ItemId lp = PageGetItemId(page, *offsets);
HeapTupleHeader tuple = (HeapTupleHeader) PageGetItem(page, lp);
(void) heap_freeze_tuple(tuple, cutoff_xid, InvalidBuffer);
offsets++;
}
}
PageSetLSN(page, lsn);
PageSetTLI(page, ThisTimeLineID);
MarkBufferDirty(buffer);
UnlockReleaseBuffer(buffer);
}
static void
heap_xlog_newpage(XLogRecPtr lsn, XLogRecord *record)
{
@@ -3546,6 +3810,18 @@ heap_redo(XLogRecPtr lsn, XLogRecord *record)
elog(PANIC, "heap_redo: unknown op code %u", info);
}
void
heap2_redo(XLogRecPtr lsn, XLogRecord *record)
{
uint8 info = record->xl_info & ~XLR_INFO_MASK;
info &= XLOG_HEAP_OPMASK;
if (info == XLOG_HEAP2_FREEZE)
heap_xlog_freeze(lsn, record);
else
elog(PANIC, "heap2_redo: unknown op code %u", info);
}
static void
out_target(StringInfo buf, xl_heaptid *target)
{
@@ -3645,3 +3921,22 @@ heap_desc(StringInfo buf, uint8 xl_info, char *rec)
else
appendStringInfo(buf, "UNKNOWN");
}
void
heap2_desc(StringInfo buf, uint8 xl_info, char *rec)
{
uint8 info = xl_info & ~XLR_INFO_MASK;
info &= XLOG_HEAP_OPMASK;
if (info == XLOG_HEAP2_FREEZE)
{
xl_heap_freeze *xlrec = (xl_heap_freeze *) rec;
appendStringInfo(buf, "freeze: rel %u/%u/%u; blk %u; cutoff %u",
xlrec->node.spcNode, xlrec->node.dbNode,
xlrec->node.relNode, xlrec->block,
xlrec->cutoff_xid);
}
else
appendStringInfo(buf, "UNKNOWN");
}

74
src/backend/access/transam/clog.c
View File

@@ -24,7 +24,7 @@
* Portions Copyright (c) 1996-2006, PostgreSQL Global Development Group
* Portions Copyright (c) 1994, Regents of the University of California
*
* $PostgreSQL: pgsql/src/backend/access/transam/clog.c,v 1.40 2006/10/04 00:29:49 momjian Exp $
* $PostgreSQL: pgsql/src/backend/access/transam/clog.c,v 1.41 2006/11/05 22:42:07 tgl Exp $
*
*-------------------------------------------------------------------------
*/
@@ -69,6 +69,7 @@ static SlruCtlData ClogCtlData;
static int ZeroCLOGPage(int pageno, bool writeXlog);
static bool CLOGPagePrecedes(int page1, int page2);
static void WriteZeroPageXlogRec(int pageno);
static void WriteTruncateXlogRec(int pageno);
/*
@@ -309,16 +310,17 @@ ExtendCLOG(TransactionId newestXact)
/*
* Remove all CLOG segments before the one holding the passed transaction ID
*
* When this is called, we know that the database logically contains no
* reference to transaction IDs older than oldestXact. However, we must
* not truncate the CLOG until we have performed a checkpoint, to ensure
* that no such references remain on disk either; else a crash just after
* the truncation might leave us with a problem. Since CLOG segments hold
* a large number of transactions, the opportunity to actually remove a
* segment is fairly rare, and so it seems best not to do the checkpoint
* unless we have confirmed that there is a removable segment. Therefore
* we issue the checkpoint command here, not in higher-level code as might
* seem cleaner.
* Before removing any CLOG data, we must flush XLOG to disk, to ensure
* that any recently-emitted HEAP_FREEZE records have reached disk; otherwise
* a crash and restart might leave us with some unfrozen tuples referencing
* removed CLOG data. We choose to emit a special TRUNCATE XLOG record too.
* Replaying the deletion from XLOG is not critical, since the files could
* just as well be removed later, but doing so prevents a long-running hot
* standby server from acquiring an unreasonably bloated CLOG directory.
*
* Since CLOG segments hold a large number of transactions, the opportunity to
* actually remove a segment is fairly rare, and so it seems best not to do
* the XLOG flush unless we have confirmed that there is a removable segment.
*/
void
TruncateCLOG(TransactionId oldestXact)
@@ -335,8 +337,8 @@ TruncateCLOG(TransactionId oldestXact)
if (!SlruScanDirectory(ClogCtl, cutoffPage, false))
return; /* nothing to remove */
/* Perform a CHECKPOINT */
RequestCheckpoint(true, false);
/* Write XLOG record and flush XLOG to disk */
WriteTruncateXlogRec(cutoffPage);
/* Now we can remove the old CLOG segment(s) */
SimpleLruTruncate(ClogCtl, cutoffPage);
@@ -386,6 +388,29 @@ WriteZeroPageXlogRec(int pageno)
(void) XLogInsert(RM_CLOG_ID, CLOG_ZEROPAGE | XLOG_NO_TRAN, &rdata);
}
/*
* Write a TRUNCATE xlog record
*
* We must flush the xlog record to disk before returning --- see notes
* in TruncateCLOG().
*
* Note: xlog record is marked as outside transaction control, since we
* want it to be redone whether the invoking transaction commits or not.
*/
static void
WriteTruncateXlogRec(int pageno)
{
XLogRecData rdata;
XLogRecPtr recptr;
rdata.data = (char *) (&pageno);
rdata.len = sizeof(int);
rdata.buffer = InvalidBuffer;
rdata.next = NULL;
recptr = XLogInsert(RM_CLOG_ID, CLOG_TRUNCATE | XLOG_NO_TRAN, &rdata);
XLogFlush(recptr);
}
/*
* CLOG resource manager's routines
*/
@@ -409,6 +434,22 @@ clog_redo(XLogRecPtr lsn, XLogRecord *record)
LWLockRelease(CLogControlLock);
}
else if (info == CLOG_TRUNCATE)
{
int pageno;
memcpy(&pageno, XLogRecGetData(record), sizeof(int));
/*
* During XLOG replay, latest_page_number isn't set up yet; insert
* a suitable value to bypass the sanity test in SimpleLruTruncate.
*/
ClogCtl->shared->latest_page_number = pageno;
SimpleLruTruncate(ClogCtl, pageno);
}
else
elog(PANIC, "clog_redo: unknown op code %u", info);
}
void
@@ -423,6 +464,13 @@ clog_desc(StringInfo buf, uint8 xl_info, char *rec)
memcpy(&pageno, rec, sizeof(int));
appendStringInfo(buf, "zeropage: %d", pageno);
}
else if (info == CLOG_TRUNCATE)
{
int pageno;
memcpy(&pageno, rec, sizeof(int));
appendStringInfo(buf, "truncate before: %d", pageno);
}
else
appendStringInfo(buf, "UNKNOWN");
}

4
src/backend/access/transam/rmgr.c
View File

@@ -3,7 +3,7 @@
*
* Resource managers definition
*
* $PostgreSQL: pgsql/src/backend/access/transam/rmgr.c,v 1.24 2006/08/07 16:57:56 tgl Exp $
* $PostgreSQL: pgsql/src/backend/access/transam/rmgr.c,v 1.25 2006/11/05 22:42:07 tgl Exp $
*/
#include "postgres.h"
@@ -32,7 +32,7 @@ const RmgrData RmgrTable[RM_MAX_ID + 1] = {
{"MultiXact", multixact_redo, multixact_desc, NULL, NULL, NULL},
{"Reserved 7", NULL, NULL, NULL, NULL, NULL},
{"Reserved 8", NULL, NULL, NULL, NULL, NULL},
{"Reserved 9", NULL, NULL, NULL, NULL, NULL},
{"Heap2", heap2_redo, heap2_desc, NULL, NULL, NULL},
{"Heap", heap_redo, heap_desc, NULL, NULL, NULL},
{"Btree", btree_redo, btree_desc, btree_xlog_startup, btree_xlog_cleanup, btree_safe_restartpoint},
{"Hash", hash_redo, hash_desc, NULL, NULL, NULL},

82
src/backend/access/transam/varsup.c
View File

@@ -6,7 +6,7 @@
* Copyright (c) 2000-2006, PostgreSQL Global Development Group
*
* IDENTIFICATION
* $PostgreSQL: pgsql/src/backend/access/transam/varsup.c,v 1.75 2006/10/04 00:29:49 momjian Exp $
* $PostgreSQL: pgsql/src/backend/access/transam/varsup.c,v 1.76 2006/11/05 22:42:07 tgl Exp $
*
*-------------------------------------------------------------------------
*/
@@ -17,6 +17,8 @@
#include "access/subtrans.h"
#include "access/transam.h"
#include "miscadmin.h"
#include "postmaster/autovacuum.h"
#include "storage/pmsignal.h"
#include "storage/proc.h"
#include "utils/builtins.h"
@@ -47,20 +49,31 @@ GetNewTransactionId(bool isSubXact)
xid = ShmemVariableCache->nextXid;
/*
/*----------
* Check to see if it's safe to assign another XID. This protects against
* catastrophic data loss due to XID wraparound. The basic rules are:
* warn if we're past xidWarnLimit, and refuse to execute transactions if
* we're past xidStopLimit, unless we are running in a standalone backend
* (which gives an escape hatch to the DBA who ignored all those
* warnings).
*
* If we're past xidVacLimit, start trying to force autovacuum cycles.
* If we're past xidWarnLimit, start issuing warnings.
* If we're past xidStopLimit, refuse to execute transactions, unless
* we are running in a standalone backend (which gives an escape hatch
* to the DBA who somehow got past the earlier defenses).
*
* Test is coded to fall out as fast as possible during normal operation,
* ie, when the warn limit is set and we haven't violated it.
* ie, when the vac limit is set and we haven't violated it.
*----------
*/
if (TransactionIdFollowsOrEquals(xid, ShmemVariableCache->xidWarnLimit) &&
TransactionIdIsValid(ShmemVariableCache->xidWarnLimit))
if (TransactionIdFollowsOrEquals(xid, ShmemVariableCache->xidVacLimit) &&
TransactionIdIsValid(ShmemVariableCache->xidVacLimit))
{
/*
* To avoid swamping the postmaster with signals, we issue the
* autovac request only once per 64K transaction starts. This
* still gives plenty of chances before we get into real trouble.
*/
if (IsUnderPostmaster && (xid % 65536) == 0)
SendPostmasterSignal(PMSIGNAL_START_AUTOVAC);
if (IsUnderPostmaster &&
TransactionIdFollowsOrEquals(xid, ShmemVariableCache->xidStopLimit))
ereport(ERROR,
@@ -69,7 +82,7 @@ GetNewTransactionId(bool isSubXact)
NameStr(ShmemVariableCache->limit_datname)),
errhint("Stop the postmaster and use a standalone backend to vacuum database \"%s\".",
NameStr(ShmemVariableCache->limit_datname))));
else
else if (TransactionIdFollowsOrEquals(xid, ShmemVariableCache->xidWarnLimit))
ereport(WARNING,
(errmsg("database \"%s\" must be vacuumed within %u transactions",
NameStr(ShmemVariableCache->limit_datname),
@@ -178,28 +191,29 @@ ReadNewTransactionId(void)
/*
* Determine the last safe XID to allocate given the currently oldest
* datminxid (ie, the oldest XID that might exist in any database
* datfrozenxid (ie, the oldest XID that might exist in any database
* of our cluster).
*/
void
SetTransactionIdLimit(TransactionId oldest_datminxid,
SetTransactionIdLimit(TransactionId oldest_datfrozenxid,
Name oldest_datname)
{
TransactionId xidVacLimit;
TransactionId xidWarnLimit;
TransactionId xidStopLimit;
TransactionId xidWrapLimit;
TransactionId curXid;
Assert(TransactionIdIsValid(oldest_datminxid));
Assert(TransactionIdIsNormal(oldest_datfrozenxid));
/*
* The place where we actually get into deep trouble is halfway around
* from the oldest existing XID. (This calculation is probably off by one
* or two counts, because the special XIDs reduce the size of the loop a
* little bit. But we throw in plenty of slop below, so it doesn't
* matter.)
* from the oldest potentially-existing XID. (This calculation is
* probably off by one or two counts, because the special XIDs reduce the
* size of the loop a little bit. But we throw in plenty of slop below,
* so it doesn't matter.)
*/
xidWrapLimit = oldest_datminxid + (MaxTransactionId >> 1);
xidWrapLimit = oldest_datfrozenxid + (MaxTransactionId >> 1);
if (xidWrapLimit < FirstNormalTransactionId)
xidWrapLimit += FirstNormalTransactionId;
@@ -229,8 +243,28 @@ SetTransactionIdLimit(TransactionId oldest_datminxid,
if (xidWarnLimit < FirstNormalTransactionId)
xidWarnLimit -= FirstNormalTransactionId;
/*
* We'll start trying to force autovacuums when oldest_datfrozenxid
* gets to be more than autovacuum_freeze_max_age transactions old.
*
* Note: guc.c ensures that autovacuum_freeze_max_age is in a sane
* range, so that xidVacLimit will be well before xidWarnLimit.
*
* Note: autovacuum_freeze_max_age is a PGC_POSTMASTER parameter so that
* we don't have to worry about dealing with on-the-fly changes in its
* value. It doesn't look practical to update shared state from a GUC
* assign hook (too many processes would try to execute the hook,
* resulting in race conditions as well as crashes of those not
* connected to shared memory). Perhaps this can be improved someday.
*/
xidVacLimit = oldest_datfrozenxid + autovacuum_freeze_max_age;
if (xidVacLimit < FirstNormalTransactionId)
xidVacLimit += FirstNormalTransactionId;
/* Grab lock for just long enough to set the new limit values */
LWLockAcquire(XidGenLock, LW_EXCLUSIVE);
ShmemVariableCache->oldestXid = oldest_datfrozenxid;
ShmemVariableCache->xidVacLimit = xidVacLimit;
ShmemVariableCache->xidWarnLimit = xidWarnLimit;
ShmemVariableCache->xidStopLimit = xidStopLimit;
ShmemVariableCache->xidWrapLimit = xidWrapLimit;
@@ -242,6 +276,18 @@ SetTransactionIdLimit(TransactionId oldest_datminxid,
ereport(DEBUG1,
(errmsg("transaction ID wrap limit is %u, limited by database \"%s\"",
xidWrapLimit, NameStr(*oldest_datname))));
/*
* If past the autovacuum force point, immediately signal an autovac
* request. The reason for this is that autovac only processes one
* database per invocation. Once it's finished cleaning up the oldest
* database, it'll call here, and we'll signal the postmaster to start
* another iteration immediately if there are still any old databases.
*/
if (TransactionIdFollowsOrEquals(curXid, xidVacLimit) &&
IsUnderPostmaster)
SendPostmasterSignal(PMSIGNAL_START_AUTOVAC);
/* Give an immediate warning if past the wrap warn point */
if (TransactionIdFollowsOrEquals(curXid, xidWarnLimit))
ereport(WARNING,

8
src/backend/access/transam/xact.c
View File

@@ -10,7 +10,7 @@
*
*
* IDENTIFICATION
* $PostgreSQL: pgsql/src/backend/access/transam/xact.c,v 1.227 2006/10/04 00:29:49 momjian Exp $
* $PostgreSQL: pgsql/src/backend/access/transam/xact.c,v 1.228 2006/11/05 22:42:07 tgl Exp $
*
*-------------------------------------------------------------------------
*/
@@ -468,8 +468,12 @@ TransactionIdIsCurrentTransactionId(TransactionId xid)
* is what we need during bootstrap. (Bootstrap mode only inserts tuples,
* it never updates or deletes them, so all tuples can be presumed good
* immediately.)
*
* Likewise, InvalidTransactionId and FrozenTransactionId are certainly
* not my transaction ID, so we can just return "false" immediately for
* any non-normal XID.
*/
if (xid == BootstrapTransactionId)
if (!TransactionIdIsNormal(xid))
return false;
/*

32
src/backend/access/transam/xlog.c
View File

@@ -7,7 +7,7 @@
* Portions Copyright (c) 1996-2006, PostgreSQL Global Development Group
* Portions Copyright (c) 1994, Regents of the University of California
*
* $PostgreSQL: pgsql/src/backend/access/transam/xlog.c,v 1.252 2006/10/18 22:44:11 tgl Exp $
* $PostgreSQL: pgsql/src/backend/access/transam/xlog.c,v 1.253 2006/11/05 22:42:08 tgl Exp $
*
*-------------------------------------------------------------------------
*/
@@ -5343,36 +5343,6 @@ GetLastSegSwitchTime(void)
return result;
}
/*
* GetRecentNextXid - get the nextXid value saved by the most recent checkpoint
*
* This is currently used only by the autovacuum daemon. To check for
* impending XID wraparound, autovac needs an approximate idea of the current
* XID counter, and it needs it before choosing which DB to attach to, hence
* before it sets up a PGPROC, hence before it can take any LWLocks. But it
* has attached to shared memory, and so we can let it reach into the shared
* ControlFile structure and pull out the last checkpoint nextXID.
*
* Since we don't take any sort of lock, we have to assume that reading a
* TransactionId is atomic ... but that assumption is made elsewhere, too,
* and in any case the worst possible consequence of a bogus result is that
* autovac issues an unnecessary database-wide VACUUM.
*
* Note: we could also choose to read ShmemVariableCache->nextXid in an
* unlocked fashion, thus getting a more up-to-date result; but since that
* changes far more frequently than the controlfile checkpoint copy, it would
* pose a far higher risk of bogus result if we did have a nonatomic-read
* problem.
*
* A (theoretically) completely safe answer is to read the actual pg_control
* file into local process memory, but that certainly seems like overkill.
*/
TransactionId
GetRecentNextXid(void)
{
return ControlFile->checkPointCopy.nextXid;
}
/*
* GetNextXidAndEpoch - get the current nextXid value and associated epoch
*