Ensure ANALYZE phase is not skipped because of canceled truncate.

Patch b19e4250b4 attempted to
preserve existing behavior regarding statistics generation in the
case that a truncation attempt was canceled due to lock conflicts.
It failed to do this accurately in two regards: (1) autovacuum had
previously generated statistics if the truncate attempt failed to
initially get the lock rather than having started the attempt, and
(2) the VACUUM ANALYZE command had always generated statistics.

Both of these changes were unintended, and are reverted by this
patch.  On review, there seems to be consensus that the previous
failure to generate statistics when the truncate was terminated
was more an unfortunate consequence of how that effort was
previously terminated than a feature we want to keep; so this
patch generates statistics even when an autovacuum truncation
attempt terminates early.  Another unintended change which is kept
on the basis that it is an improvement is that when a VACUUM
command is truncating, it will the new heuristic for avoiding
blocking other processes, rather than keeping an
AccessExclusiveLock on the table for however long the truncation
takes.

Per multiple reports, with some renaming per patch by Jeff Janes.

Backpatch to 9.0, where problem was created.
This commit is contained in:
Kevin Grittner 2013-04-29 13:05:26 -05:00
parent 91fa8532f4
commit 5fc893760f
1 changed files with 22 additions and 36 deletions

View File

@ -78,9 +78,9 @@
* that the potential for improvement was great enough to merit the cost of * that the potential for improvement was great enough to merit the cost of
* supporting them. * supporting them.
*/ */
#define AUTOVACUUM_TRUNCATE_LOCK_CHECK_INTERVAL 20 /* ms */ #define VACUUM_TRUNCATE_LOCK_CHECK_INTERVAL 20 /* ms */
#define AUTOVACUUM_TRUNCATE_LOCK_WAIT_INTERVAL 50 /* ms */ #define VACUUM_TRUNCATE_LOCK_WAIT_INTERVAL 50 /* ms */
#define AUTOVACUUM_TRUNCATE_LOCK_TIMEOUT 5000 /* ms */ #define VACUUM_TRUNCATE_LOCK_TIMEOUT 5000 /* ms */
/* /*
* Guesstimation of number of dead tuples per page. This is used to * Guesstimation of number of dead tuples per page. This is used to
@ -285,17 +285,10 @@ lazy_vacuum_rel(Relation onerel, VacuumStmt *vacstmt,
new_frozen_xid, new_frozen_xid,
new_min_multi); new_min_multi);
/* /* report results to the stats collector, too */
* Report results to the stats collector, too. An early terminated pgstat_report_vacuum(RelationGetRelid(onerel),
* lazy_truncate_heap attempt suppresses the message and also cancels the onerel->rd_rel->relisshared,
* execution of ANALYZE, if that was ordered. new_rel_tuples);
*/
if (!vacrelstats->lock_waiter_detected)
pgstat_report_vacuum(RelationGetRelid(onerel),
onerel->rd_rel->relisshared,
new_rel_tuples);
else
vacstmt->options &= ~VACOPT_ANALYZE;
/* and log the action if appropriate */ /* and log the action if appropriate */
if (IsAutoVacuumWorkerProcess() && Log_autovacuum_min_duration >= 0) if (IsAutoVacuumWorkerProcess() && Log_autovacuum_min_duration >= 0)
@ -1347,28 +1340,21 @@ lazy_truncate_heap(Relation onerel, LVRelStats *vacrelstats)
*/ */
CHECK_FOR_INTERRUPTS(); CHECK_FOR_INTERRUPTS();
if (++lock_retry > (AUTOVACUUM_TRUNCATE_LOCK_TIMEOUT / if (++lock_retry > (VACUUM_TRUNCATE_LOCK_TIMEOUT /
AUTOVACUUM_TRUNCATE_LOCK_WAIT_INTERVAL)) VACUUM_TRUNCATE_LOCK_WAIT_INTERVAL))
{ {
/* /*
* We failed to establish the lock in the specified number of * We failed to establish the lock in the specified number of
* retries. This means we give up truncating. Suppress the * retries. This means we give up truncating.
* ANALYZE step. Doing an ANALYZE at this point will reset the
* dead_tuple_count in the stats collector, so we will not get
* called by the autovacuum launcher again to do the truncate.
*/ */
vacrelstats->lock_waiter_detected = true; vacrelstats->lock_waiter_detected = true;
ereport(LOG, ereport(elevel,
(errmsg("automatic vacuum of table \"%s.%s.%s\": " (errmsg("\"%s\": stopping truncate due to conflicting lock request",
"could not (re)acquire exclusive "
"lock for truncate scan",
get_database_name(MyDatabaseId),
get_namespace_name(RelationGetNamespace(onerel)),
RelationGetRelationName(onerel)))); RelationGetRelationName(onerel))));
return; return;
} }
pg_usleep(AUTOVACUUM_TRUNCATE_LOCK_WAIT_INTERVAL); pg_usleep(VACUUM_TRUNCATE_LOCK_WAIT_INTERVAL);
} }
/* /*
@ -1448,8 +1434,6 @@ count_nondeletable_pages(Relation onerel, LVRelStats *vacrelstats)
{ {
BlockNumber blkno; BlockNumber blkno;
instr_time starttime; instr_time starttime;
instr_time currenttime;
instr_time elapsed;
/* Initialize the starttime if we check for conflicting lock requests */ /* Initialize the starttime if we check for conflicting lock requests */
INSTR_TIME_SET_CURRENT(starttime); INSTR_TIME_SET_CURRENT(starttime);
@ -1467,24 +1451,26 @@ count_nondeletable_pages(Relation onerel, LVRelStats *vacrelstats)
/* /*
* Check if another process requests a lock on our relation. We are * Check if another process requests a lock on our relation. We are
* holding an AccessExclusiveLock here, so they will be waiting. We * holding an AccessExclusiveLock here, so they will be waiting. We
* only do this in autovacuum_truncate_lock_check millisecond * only do this once per VACUUM_TRUNCATE_LOCK_CHECK_INTERVAL, and we
* intervals, and we only check if that interval has elapsed once * only check if that interval has elapsed once every 32 blocks to
* every 32 blocks to keep the number of system calls and actual * keep the number of system calls and actual shared lock table
* shared lock table lookups to a minimum. * lookups to a minimum.
*/ */
if ((blkno % 32) == 0) if ((blkno % 32) == 0)
{ {
instr_time currenttime;
instr_time elapsed;
INSTR_TIME_SET_CURRENT(currenttime); INSTR_TIME_SET_CURRENT(currenttime);
elapsed = currenttime; elapsed = currenttime;
INSTR_TIME_SUBTRACT(elapsed, starttime); INSTR_TIME_SUBTRACT(elapsed, starttime);
if ((INSTR_TIME_GET_MICROSEC(elapsed) / 1000) if ((INSTR_TIME_GET_MICROSEC(elapsed) / 1000)
>= AUTOVACUUM_TRUNCATE_LOCK_CHECK_INTERVAL) >= VACUUM_TRUNCATE_LOCK_CHECK_INTERVAL)
{ {
if (LockHasWaitersRelation(onerel, AccessExclusiveLock)) if (LockHasWaitersRelation(onerel, AccessExclusiveLock))
{ {
ereport(elevel, ereport(elevel,
(errmsg("\"%s\": suspending truncate " (errmsg("\"%s\": suspending truncate due to conflicting lock request",
"due to conflicting lock request",
RelationGetRelationName(onerel)))); RelationGetRelationName(onerel))));
vacrelstats->lock_waiter_detected = true; vacrelstats->lock_waiter_detected = true;