On Mon, Jul 6, 2009 at 8:38 PM, Jennifer
Spencer<jenniferm411@hotmail.com> wrote:
> Hello All -
>
> I will soon be receiving a new production and warm-standby machine. Prior
> to now, we have only had one database machine and no warm standby.
>
> We had planned to feed the standby with one backup and then use WAL files
> rsync'd over until either catastrophe or maintenance required on the primary
> machine. I will be able to properly close out my users on the primary,
> force a closeout of the log file (pg_switch_xlog), and shut it down. Then I
> can catch up and ingest the last WAL on my warm standby, let the users in,
> and have the warm standby become the new primary. All okay so far.
>
> But. When the primary is vacuumed, re-indexed and all clean and shiny
> again, HOW do I catch up with data changes that happened in the meantime on
> my warm standby without corruption or >30 minutes of user down-time? I
> cannot re-ingest WALs from the warm standby into the cleaned up primary or I
> get a PANIC, and I don't have time to run a full backup on the warm standby
> and ingest it into the primary leaving everything down.
If you've moved on, so to speak, with the new primary, you restart the
old primary, now warm standby, the same way you initially created the
warm standby. issue the start hot backup command to the primary, copy
over all the data dir and start shipping WAL files to it before you
start continuous recovery.