I've been working on backup restore scripts for a PostgreSQL DB hosted
on an Amazon EC2 instance. EC2 instances don't have any persistent
storage, so I archive the WAL files to Amazon S3. I do periodic
backups, and restore and roll forward when an instance is restarted.
I've followed the excellent documentation at http://www.postgresql.org/docs/8.0/interactive/backup-online.html
, and everything is working well, but I have a few further questions.
1. Purely out of curiousity, what's the nnnnnnnn.history file which is
requested from the archive when you restore? I realise that the
restore process looking for files which don't exist is not a problem,
I'm just being inquisitive.
2. When I do a backup, Postgres requests archiving of n.m.backup and
the WAL file n immediately, so I only copy WAL files with numbers > n
from pg_xlog. I've seen suggestions that I should be copying WAL files
with numbers >= n -- is there any reason to do so?
3. What's the best thing to do when I deliberately shut down
PostgreSQL (i.e. pg_ctl stop)? When I start again I will be restoring
from the most recent backup and rolling forward over the archived WAL
files. I believe that shutdown leaves me with unarchived WAL files in
pg_xlog. Can I simply copy these to S3 as if they were WAL files which
were copied from pg_xlog at the time of the last backup, and then put
them back into pg_xlog as part of my restore process?
4. I'm using PostgreSQL 8.0 -- are there any significant improvements
in on-line backups in later versions?
Thanks,
Tom