On Fri, 3 Nov 2006, Tom Lane wrote:
> Jeff Frost <jeff@frostconsultingllc.com> writes:
>> Well, I spoke to soon on the it all works front. So, it's been
>> reindexed and appears to be working properly now. I guess I'll keep
>> an eye on it for a while. I didn't get your query suggestion in time,
>> so hopefully I grabbed the right binary file..though it did seem to
>> disappear after the reindex, so I think it's likely the correct one.
>> Definitely got the correct second one.
>
> Well, I can't find anything wrong :-(. There are some differences in
> the list of contained keys, but they're all up near the end of the
> range, which is consistent with the assumption that the table is live
> and had some changes between your two dumps of the index. In
> particular, there's no difference in the entries for the troublesome
> key value:
>
> 38635629 24080 25
> 38635629 24080 26
> 38635629 24080 27
>
> So I dunno what to make of it. If it happens again, we need to look
> more closely.
Tom, I know we shouldn't have to REINDEX in the 8.1.x days. Do you have any
idea what might have allowed this to happen? A while back this particular
server was unable to send e-mail and so we weren't getting the vacuum verbose
output. As a consequence the FSM settings were too low. That has been
remedied, but I'm wondering if it's possible that the FSM settings being too
low would allow the INDEX to somehow get damaged?
--
Jeff Frost, Owner <jeff@frostconsultingllc.com>
Frost Consulting, LLC http://www.frostconsultingllc.com/
Phone: 650-780-7908 FAX: 650-649-1954