On 08.05.2012 04:15, Ryan Kelly wrote:
> On Mon, May 07, 2012 at 05:31:40PM +0100, tom Tom wrote:
>> Nope, this was just a benchmark script that caused this, any sane person
>> would use an intbig index instead I guess. A better error message would be
>> nice though, I was pretty confused when this happened.
>>
>> This can also bring down postgresql - it happens occasionally and causes
>> the server to terminate. Someone in #postgresql said this happens when the
>> failure to acquire the lock occurs in a "critical section"? That might be
>> cause for concern.
> Occasionally, it causes a PANIC instead of an ERROR. I have the logs
> from the IRC session if anyone is in need of them.
Good point.
I wonder if we should reserve a few of the lwlock "slots" for critical
sections, to make this less likely to happen. Not only in this case, but
in general. We haven't seen this problem often, but it would be quite
trivial to reserve a few slots.
--
Heikki Linnakangas
EnterpriseDB http://www.enterprisedb.com