Re: PITR Recovery and out-of-sync indexes

Simon Riggs <simon(at)2ndquadrant(dot)com> writes:
> On Thu, 2007-10-04 at 09:21 -0600, Brian Wipf wrote:
>> The Apple Xserve is easy to maintain and rock solid reliable. If it
>> had better performance to its Fibre Channel RAID array, it would be a
>> better main server too. The Linux box is a better performer, but in
>> our experience at least, more difficult to maintain when things go
>> wrong.

> It is a small gain for much risk. If you have weird problems, remember
> that you are doing something we (or at least I) said not to... even if
> it works for you it may not do for someone else that tries.

Well, the other side of the coin is that using a nonidentical backup
server might protect him against some types of common-mode failures.
Consider for example a worm that can penetrate only one of the two OSes.

He needs to get the locale sort orders in sync, but beyond that I
can't think of any hard reason why he can't use this combination.

BTW, the reasons I don't like the "we'll reindex after we bring up
the backup server" theory are:
1. I'm not convinced that the locale mismatch couldn't cause a failure
while following the WAL log.
2. It's the sort of step that you are certain to forget when it's
four AM and you're really in urgent need of that backup server.
Fix the locales, instead.

regards, tom lane