On 03/30/11 4:08 PM, Mike Orr wrote:
> I looked at PITR and WAL, but it looks like I can't have the backup
> database running and answering queries while it's WAL'ing. I'd have to
> log in and switch it to normal mode and start the webapp, and that's
> what I may not be able to do if the backup server were needed.
>
look at the 'warm standby' stuff. just realize that this backup HAS to
be readonly.
> The backup server exists in case there's something like an earthquake
> at the main data center. Something that could cause a regional
> communication blackout and prevent the sysadmins from accessing the
> backup server remotely. The site is a data-sharing tool for emergency
> responders across the country. They may be in another region working
> on an unrelated incident, and need the website. So the backup site has
> to be already running and loaded with data -- we can't depend on a
> sysadmin being able to log in remotely to turn it on
distributed failover is tricky stuff. the standby *could* be
monitoring the active server and initiate automatic failover if it goes
down, but you don't want spontaneous failover if communications between
the servers is disrupted or you end up with two masters.
