database/postgres
1
0
Fork 0
mirror of https://github.com/postgres/postgres.git synced 2025-11-09 06:21:09 +03:00
Code Activity

Fix MVCC bug with prepared xact with subxacts on standby

Browse Source 
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 commit is contained in:
Heikki Linnakangas
2024-06-27 21:06:32 +03:00
parent ecbf6ac51d
commit cbfbda7841
7 changed files with 85 additions and 17 deletions
Download Patch File Download Diff File Expand all files Collapse all files

7
src/backend/access/transam/twophase.c
View File

@@ -2035,9 +2035,8 @@ PrescanPreparedTransactions(TransactionId **xids_p, int *nxids_p)
* 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)
@@ -2057,7 +2056,7 @@ StandbyRecoverPreparedTransactions(void)
buf = ProcessTwoPhaseBuffer(xid,
gxact->prepare_start_lsn,
gxact->ondisk, false, false);
gxact->ondisk, true, false);
if (buf != NULL)
pfree(buf);
}