Lonni J Friedman wrote:
> I'm running postgresql-9.1.3 on a Linux-x86_64 (Fedora16, if it
> matters) system. I noticed the existence of pg_basebackup starting in
> 9.1, and figured I'd try it out and see if it would simplify our
> backup & management processes.
>
> $ pg_basebackup -P -v -D /tmp/backup -x -Ft -z -U postgres
> xlog start point: C6/64000020
> 135733616/135733616 kB (100%), 1/1 tablespace
> xlog end point: C6/640000A0
> pg_basebackup: base backup completed
>
> So after running through this, I tried to use (restore) the backup
> that was generated. While everything appears to be working ok from a
> functional perspective, in the server log I saw the following:
> ######
> LOG: creating missing WAL directory "pg_xlog/archive_status"
> LOG: database system was not properly shut down; automatic recovery in progress
> LOG: redo starts at C6/66000078
> LOG: could not open file "pg_xlog/00000001000000C600000067" (log file
> 198, segment 103): No such file or directory
> LOG: redo done at C6/660000A0
> FATAL: the database system is starting up
> LOG: autovacuum launcher started
> LOG: database system is ready to accept connections
> #####
>
> Just to be clear, here's what I did after pg_basebackup had completed
> successfully:
> 0) shutdown postgresql gracefully, and verified that it was fully shutdown
> 1) moved $PGDATA to $PGDATA.old
> 2) created $PGDATA as postgres user
> 3) extracted the basebackup tarball as postgres user
> cd $PGDATA && tar xzvpf /tmp/backup/base.tar.gz
> 4) started postgresql up
>
> I would have expected that I wouldn't have gotten the 'not properly
> shutdown' warning, or the 'could not open file' warning by following
> this process. Am I doing something wrong?
From what you quote it looks like everything went well and as expected.
An online backup is not a consistent state of the database, so WAL
files will have to be applied to recover the database. See
http://www.postgresql.org/docs/current/static/continuous-archiving.html#BACKUP-PITR-RECOVERY
Your method worked because you used the -x flag to copy WAL files
along with the backup. You were lucky because all the necessary WAL
files were still there (see the note in pg_basebackup's
documentation).
If there is activity on the database while you take your backup and
wal_keep_segments is not high enough, you method will not work.
In general, you should also copy the archived WAL files and follow
the procedure in the above link to restore the database.
Yours,
Laurenz Albe