We did not recover the subtransaction IDs of prepared transactions
when starting a hot standby from a shutdown checkpoint. As a result,
such subtransactions were considered as aborted, rather than
in-progress. That would lead to hint bits being set incorrectly, and
the subtransactions suddenly becoming visible to old snapshots when
the prepared transaction was committed.
To fix, update pg_subtrans with prepared transactions's subxids when
starting hot standby from a shutdown checkpoint. The snapshots taken
from that state need to be marked as "suboverflowed", so that we also
check the pg_subtrans.
Backport to all supported versions.
Discussion: https://www.postgresql.org/message-id/
6b852e98-2d49-4ca1-9e95-
db419a2696e0@iki.fi
* This is never called at the end of recovery - we use
* RecoverPreparedTransactions() at that point.
*
- * The lack of calls to SubTransSetParent() calls here is by design;
- * those calls are made by RecoverPreparedTransactions() at the end of recovery
- * for those xacts that need this.
+ * This updates pg_subtrans, so that any subtransactions will be correctly
+ * seen as in-progress in snapshots taken during recovery.
*/
void
StandbyRecoverPreparedTransactions(void)
buf = ProcessTwoPhaseBuffer(xid,
gxact->prepare_start_lsn,
- gxact->ondisk, false, false);
+ gxact->ondisk, true, false);
if (buf != NULL)
pfree(buf);
}
RunningTransactionsData running;
TransactionId latestCompletedXid;
+ /* Update pg_subtrans entries for any prepared transactions */
+ StandbyRecoverPreparedTransactions();
+
/*
* Construct a RunningTransactions snapshot representing a
* shut down server, with only prepared transactions still
*/
running.xcnt = nxids;
running.subxcnt = 0;
- running.subxid_overflow = false;
+ running.subxid_status = SUBXIDS_IN_SUBTRANS;
running.nextXid = XidFromFullTransactionId(checkPoint.nextFullXid);
running.oldestRunningXid = oldestActiveXID;
latestCompletedXid = XidFromFullTransactionId(checkPoint.nextFullXid);
running.xids = xids;
ProcArrayApplyRecoveryInfo(&running);
-
- StandbyRecoverPreparedTransactions();
}
}
oldestActiveXID = PrescanPreparedTransactions(&xids, &nxids);
+ /* Update pg_subtrans entries for any prepared transactions */
+ StandbyRecoverPreparedTransactions();
+
/*
* Construct a RunningTransactions snapshot representing a shut
* down server, with only prepared transactions still alive. We're
*/
running.xcnt = nxids;
running.subxcnt = 0;
- running.subxid_overflow = false;
+ running.subxid_status = SUBXIDS_IN_SUBTRANS;
running.nextXid = XidFromFullTransactionId(checkPoint.nextFullXid);
running.oldestRunningXid = oldestActiveXID;
latestCompletedXid = XidFromFullTransactionId(checkPoint.nextFullXid);
running.xids = xids;
ProcArrayApplyRecoveryInfo(&running);
-
- StandbyRecoverPreparedTransactions();
}
/* ControlFile->checkPointCopy always tracks the latest ckpt XID */
* If the snapshot isn't overflowed or if its empty we can reset our
* pending state and use this snapshot instead.
*/
- if (!running->subxid_overflow || running->xcnt == 0)
+ if (running->subxid_status != SUBXIDS_MISSING || running->xcnt == 0)
{
/*
* If we have already collected known assigned xids, we need to
* missing, so conservatively assume the last one is latestObservedXid.
* ----------
*/
- if (running->subxid_overflow)
+ if (running->subxid_status == SUBXIDS_MISSING)
{
standbyState = STANDBY_SNAPSHOT_PENDING;
standbyState = STANDBY_SNAPSHOT_READY;
standbySnapshotPendingXmin = InvalidTransactionId;
+
+ /*
+ * If the 'xids' array didn't include all subtransactions, we have to
+ * mark any snapshots taken as overflowed.
+ */
+ if (running->subxid_status == SUBXIDS_IN_SUBTRANS)
+ procArray->lastOverflowedXid = latestObservedXid;
+ else
+ {
+ Assert(running->subxid_status == SUBXIDS_IN_ARRAY);
+ procArray->lastOverflowedXid = InvalidTransactionId;
+ }
}
/*
CurrentRunningXacts->xcnt = count - subcount;
CurrentRunningXacts->subxcnt = subcount;
- CurrentRunningXacts->subxid_overflow = suboverflowed;
+ CurrentRunningXacts->subxid_status = suboverflowed ? SUBXIDS_IN_SUBTRANS : SUBXIDS_IN_ARRAY;
CurrentRunningXacts->nextXid = XidFromFullTransactionId(ShmemVariableCache->nextFullXid);
CurrentRunningXacts->oldestRunningXid = oldestRunningXid;
CurrentRunningXacts->latestCompletedXid = latestCompletedXid;
running.xcnt = xlrec->xcnt;
running.subxcnt = xlrec->subxcnt;
- running.subxid_overflow = xlrec->subxid_overflow;
+ running.subxid_status = xlrec->subxid_overflow ? SUBXIDS_MISSING : SUBXIDS_IN_ARRAY;
running.nextXid = xlrec->nextXid;
running.latestCompletedXid = xlrec->latestCompletedXid;
running.oldestRunningXid = xlrec->oldestRunningXid;
xlrec.xcnt = CurrRunningXacts->xcnt;
xlrec.subxcnt = CurrRunningXacts->subxcnt;
- xlrec.subxid_overflow = CurrRunningXacts->subxid_overflow;
+ xlrec.subxid_overflow = (CurrRunningXacts->subxid_status != SUBXIDS_IN_ARRAY);
xlrec.nextXid = CurrRunningXacts->nextXid;
xlrec.oldestRunningXid = CurrRunningXacts->oldestRunningXid;
xlrec.latestCompletedXid = CurrRunningXacts->latestCompletedXid;
recptr = XLogInsert(RM_STANDBY_ID, XLOG_RUNNING_XACTS);
- if (CurrRunningXacts->subxid_overflow)
+ if (xlrec.subxid_overflow)
elog(trace_recovery(DEBUG2),
"snapshot of %u running transactions overflowed (lsn %X/%X oldest xid %u latest complete %u next xid %u)",
CurrRunningXacts->xcnt,
* almost immediately see the data we need to begin executing queries.
*/
+typedef enum
+{
+ SUBXIDS_IN_ARRAY, /* xids array includes all running subxids */
+ SUBXIDS_MISSING, /* snapshot overflowed, subxids are missing */
+ SUBXIDS_IN_SUBTRANS, /* subxids are not included in 'xids', but
+ * pg_subtrans is fully up-to-date */
+} subxids_array_status;
+
typedef struct RunningTransactionsData
{
int xcnt; /* # of xact ids in xids[] */
int subxcnt; /* # of subxact ids in xids[] */
- bool subxid_overflow; /* snapshot overflowed, subxids missing */
+ subxids_array_status subxid_status;
TransactionId nextXid; /* xid from ShmemVariableCache->nextFullXid */
TransactionId oldestRunningXid; /* *not* oldestXmin */
TransactionId latestCompletedXid; /* so we can set xmax */
use PostgresNode;
use TestLib;
-use Test::More tests => 24;
+use Test::More tests => 27;
my $psql_out = '';
my $psql_rc = '';
$cur_master->psql('postgres', "COMMIT PREPARED 'xact_009_12'");
+###############################################################################
+# Check visibility of prepared transactions in standby after a restart while
+# primary is down.
+###############################################################################
+
+$cur_master->psql(
+ 'postgres', "
+ CREATE TABLE t_009_tbl_standby_mvcc (id int, msg text);
+ BEGIN;
+ INSERT INTO t_009_tbl_standby_mvcc VALUES (1, 'issued to ${cur_master_name}');
+ SAVEPOINT s1;
+ INSERT INTO t_009_tbl_standby_mvcc VALUES (2, 'issued to ${cur_master_name}');
+ PREPARE TRANSACTION 'xact_009_standby_mvcc';
+ ");
+$cur_master->stop;
+$cur_standby->restart;
+
+# Acquire a snapshot in standby, before we commit the prepared transaction
+my $standby_session = $cur_standby->background_psql('postgres', on_error_die => 1);
+$standby_session->query_safe("BEGIN ISOLATION LEVEL REPEATABLE READ");
+$psql_out = $standby_session->query_safe(
+ "SELECT count(*) FROM t_009_tbl_standby_mvcc");
+is($psql_out, '0',
+ "Prepared transaction not visible in standby before commit");
+
+# Commit the transaction in primary
+$cur_master->start;
+$cur_master->psql('postgres', "
+SET synchronous_commit='remote_apply'; -- To ensure the standby is caught up
+COMMIT PREPARED 'xact_009_standby_mvcc'
+");
+
+# Still not visible to the old snapshot
+$psql_out = $standby_session->query_safe(
+ "SELECT count(*) FROM t_009_tbl_standby_mvcc");
+is($psql_out, '0',
+ "Committed prepared transaction not visible to old snapshot in standby");
+
+# Is visible to a new snapshot
+$standby_session->query_safe("COMMIT");
+$psql_out = $standby_session->query_safe(
+ "SELECT count(*) FROM t_009_tbl_standby_mvcc");
+is($psql_out, '2',
+ "Committed prepared transaction is visible to new snapshot in standby");
+
###############################################################################
# Check for a lock conflict between prepared transaction with DDL inside and
# replay of XLOG_STANDBY_LOCK wal record.
substitute_actual_parameters_context
substitute_actual_srf_parameters_context
substitute_phv_relids_context
+subxids_array_status
svtype
symbol
tablespaceinfo
projects / postgresql.git / commitdiff