From: | Brian Wipf <brian(at)clickspace(dot)com> |
---|---|
To: | Simon Riggs <simon(at)2ndquadrant(dot)com> |
Cc: | Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us>, pgsql-general(at)postgresql(dot)org |
Subject: | Re: Base Backups from PITR Standby |
Date: | 2007-10-30 22:18:01 |
Message-ID: | 46FA8F10-BCE3-4DBB-9DAB-DC350785BBCF@clickspace.com |
Views: | Raw Message | Whole Thread | Download mbox | Resend email |
Thread: | |
Lists: | pgsql-general |
On 30-Oct-07, at 2:42 PM, Simon Riggs wrote:
> It's safest to shutdown the standby first, take a backup then crank it
> up again.
>
> It's possible to do it online in the way you suggest, but only when
> running with either full_page_writes = on or when making the backup
> with
> rsync, or another method that uses atomic 8Kb reads
I'm using rsync with the -a flag (equivalent to -rlptgoD)
I can give rsync a block size with the -B flag. According to the rsync
man page, "[the blocksize] is normally selected based on the size of
each file being updated". Maybe this is all I need to set. I'll give
that a try.
> . The backup is only
> valid if you rollforward past the next restartpoint after the backup
> completes
I can ensure this is done.
> , which is not technically the right place but definitely far
> enough. So its not exactly for the feint-hearted, at the moment.
Thanks for the input Simon, that helps a lot. If I can't get the
online backup of the warm standby to work properly, I'll simply keep
two warm standby's replaying WALs on the backup server. I'll have one
installation continually replaying WALs for failover and a second
which can be shut down periodically for taking base backups.
Brian Wipf
<brian(at)clickspace(dot)com>
From | Date | Subject | |
---|---|---|---|
Next Message | Thomas Kellerer | 2007-10-30 22:18:51 | Re: PostgreSQL and AutoCad |
Previous Message | Ilan Volow | 2007-10-30 22:01:09 | Re: PostgreSQL and AutoCad |