Re: 8.3.0 Core with concurrent vacuum fulls

"Gavin M. Roy" <gmr(at)myyearbook(dot)com> writes:
> On Wed, Mar 5, 2008 at 10:13 AM, Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> wrote:
>> Actually, maybe it *has* been seen before. Gavin, are you in the habit
>> of running concurrent VACUUM FULLs on system catalogs, and if so have
>> you noted that they occasionally get deadlock failures?

> Generally no, I've never noticed deadlocks before, but I'll go back and
> look at some of the other the machines.

After digging a bit deeper, it seems that pre-8.2 releases wouldn't have
been at risk for a deadlock here anyway, because
CatalogCacheInitializeCache didn't lock the system catalog it was
initializing a cache for. (That had *other* risks, but not this one.)
So possibly the lack of prior reports is just because not too many
people are in the habit of using concurrent VACUUM FULLs with late-model
Postgres. I can reproduce the deadlock (though not the ensuing PANIC)
in 8.2, so it's definitely not heap_page_prune's fault.

regards, tom lane