Обсуждение: Advancing the archiver position safely
I've encountered a rather unusual situation (PostgreSQL 9.6). On a particular server, for reasons I've not fully diagnosed,the archiver thinks that the current WAL segment to be archived is 0000000200003B6800000062. This is unfortunate,because the oldest WAL segment that actually exists on disk is 0000000200003F1100000004, so the archive scriptis failing repeatedly because of the missing segment. The system is not actually missing important (for recovery) WAL segments, at least: Latest checkpoint's REDO WAL file: 000000020000417600000029 I'd like to "catch up" the archiver such that it is operating on files that actually exist; besides setting archive_commandto '/bin/true' and letting it chew through the old ones, is there a way of safely advancing the position ofthe archiver? -- -- Christophe Pettus xof@thebuild.com
Christophe Pettus <xof@thebuild.com> writes: > I've encountered a rather unusual situation (PostgreSQL 9.6). On a > particular server, for reasons I've not fully diagnosed, the archiver > thinks that the current WAL segment to be archived is > 0000000200003B6800000062. This is unfortunate, because the oldest WAL > segment that actually exists on disk is 0000000200003F1100000004, so > the archive script is failing repeatedly because of the missing > segment. > > The system is not actually missing important (for recovery) WAL segments, at least: > > Latest checkpoint's REDO WAL file: 000000020000417600000029 > > I'd like to "catch up" the archiver such that it is operating on files > that actually exist; besides setting archive_command to '/bin/true' > and letting it chew through the old ones, is there a way of safely > advancing the position of the archiver? Take a look at the contents of your pg_xlog/archive_status directory where you will likely find a .ready file corresponding to the missing segment and perhaps others. Deleting the .ready file should allow the archiver to get past the missing file. Make certain you're *not* mucking with the WAL files themselves. > -- > -- Christophe Pettus > xof@thebuild.com > > > > -- Jerry Sievers Postgres DBA/Development Consulting e: postgres.consulting@comcast.net
> On Aug 6, 2020, at 18:45, Jerry Sievers <gsievers19@comcast.net> wrote: > Deleting the .ready file should allow the archiver to get past the > missing file. Ah, excellent, yes. -- -- Christophe Pettus xof@thebuild.com