On Mon, Oct 19, 2009 at 11:06 AM, Brad Nicholson
<bnichols@ca.afilias.info> wrote:
> On Mon, 2009-10-19 at 10:53 -0600, Scott Marlowe wrote:
>> On Mon, Oct 19, 2009 at 10:44 AM, Tom Lane <tgl@sss.pgh.pa.us> wrote:
>> > Brad Nicholson <bnichols@ca.afilias.info> writes:
>> >> On Mon, 2009-10-19 at 12:07 -0400, Tom Lane wrote:
>> >>> That seems like a fundamentally stupid idea, unless you are unconcerned
>> >>> with the time and cost of getting the DB running again, which seemingly
>> >>> you are.
>> >
>> >> I disagree that this is fundamentally stupid. We are talking about a
>> >> situation where the server is about to die, HA solution kicks in and
>> >> moves it to standby.
>> >
>> > Moving it to standby immediately is a good idea, but it does not follow
>> > that you need to hit the DB over the head with a hammer. A fast-mode
>> > shutdown seems perfectly adequate. If it isn't, you're going to need
>> > nontrivial recovery effort anyhow.
>>
>> All of this is completely besides the point that a database that's
>> been shutdown immediately / had the power cord yanked comes back up
>> and doesn't start autovacuuming automatically, which seems a
>> non-optimal behaviour.
>
> It's also not going to endear us very much to the VLDB crowd - it will
> amounts to a multi-hour crash recovery for those folks while analyze
> regenerates statistics.
But this would be AOK behaviour for small transactional databases?
Again, besides the point, but important. The real point is a database
that doesn't run autovac after an emergency shutdown is broken by
design, and not just for one use case.
