Greg Stark <gsstark@mit.edu> writes:
> Tom Lane <tgl@sss.pgh.pa.us> writes:
>> And what PG version is this? Alvaro fixed the
>> foreign-keys-take-exclusive-locks problem in 8.1 ...
> Except I don't think this is taking an exclusive lock at all. The original
> post had the deadlock detection fire on a SharedLock.
Yeah, but it was a ShareLock on a transaction ID, which is the trace
of something doing XactLockTableWait, which is only done if we're
blocking on a locked or updated-but-uncommitted row.
Since Brendan says he's using 8.1, the FK theory is out, and I think
what this probably is is a garden-variety deadlock on tuple updates, ie,
two concurrent transactions tried to update the same tuples in different
orders.
regards, tom lane