"Joshua D. Drake" <jd(at)commandprompt(dot)com> writes:
> We ran into a problem with a customer this weekend. They had >128,000
> tables and we were trying to run a pg_dump. When we reached
> max_locks_per_transaction, the dump just hung waiting to lock the next
> table.
> Would it make sense to have some sort of timeout for that?
I don't think you have diagnosed this correctly. Running out of lock
table slots generates an "out of shared memory" error, with a HINT that
you might want to increase max_locks_per_transaction. If you can prove
otherwise, please supply a test case.
regards, tom lane