From: | "Zeugswetter Andreas SB SD" <ZeugswetterA(at)spardat(dot)at> |
---|---|
To: | "Bruce Momjian" <pgman(at)candle(dot)pha(dot)pa(dot)us>, "Naomi Walker" <nwalker(at)eldocomp(dot)com> |
Cc: | <matthew(at)hairy(dot)beasts(dot)org>, <dmiller(at)espgroup(dot)net>, <pjw(at)rhyme(dot)com(dot)au>, <ncm(at)zembu(dot)com>, <pgsql-admin(at)postgresql(dot)org> |
Subject: | Re: Backup and Recovery (revisited) |
Date: | 2002-02-25 12:04:30 |
Message-ID: | 46C15C39FEB2C44BA555E356FBCD6FA41EB520@m0114.s-mxs.net |
Views: | Raw Message | Whole Thread | Download mbox | Resend email |
Thread: | |
Lists: | pgsql-admin |
> One use of replication can be point-in-time recovery. The log of
> replication messages can be played back against a nightly backup to
> restore the system to the current state. It would be nice to hit both
> of these at the same time.
WAL currently has all that is needed for full instance master slave replication.
Imho it would be a shame to add yet another txlog that is only used for
replication. Imho if the current WAL misses some info that would be needed
for partial and master-master replication, then the current WAL should be
extended to contain such info.
What instantly comes to mind is the old primary key value.
Since I guess we want partial replication we need a log sniffer anyway,
that extracts only what is needed for a specific target replicate.
Imho most efficiently replicating "delete from atab;" or other single statements
that affect many rows should not be a main issue in replication design.
Andreas
From | Date | Subject | |
---|---|---|---|
Next Message | Bruno Wolff III | 2002-02-25 15:28:48 | Re: Reset sequence number |
Previous Message | Jean-Christophe ARNU (JX) | 2002-02-25 08:21:25 | mirorring databases with small time gaps |