--- On Mon, 12/7/09, Tom Lane <wrote:
> Have you looked into pg_locks to see if it's blocked
> waiting for a lock?
> The TRUNCATE in particular would require exclusive lock on
> the table, so it could be waiting for some other process
> that's touched the table.
Thanks Tom - while pg_locks did not reveal any locks, you gave me the clue that solved the mystery: the truncate
command.
pg_locks show no locks at all on this table, but just in case, I tried removing an ANLAYZE on the table being
truncated,that was being done right at the beginning of the sproc, and should have finished long before the truncate
wasbeing run, and hey presto no deadlocks.
A couple of weird things about this: the fact the pg_locks showed no locks, and secondly the fact that it used to work
finebefore we started partitioning an unrelated table.