Mladen Gogala <mladen(dot)gogala(at)vmsinfo(dot)com> writes:
> Tom Lane wrote:
>> My guess would be overstressed disk subsystem. A COMMIT doesn't require
>> much except fsync'ing the commit WAL record down to disk ...
> Doesn't the "commit" statement also release all the locks held by the
> transaction?
Yeah, and there's a nontrivial amount of other cleanup too, but it all
amounts to just changes in in-memory data structures. I don't see that
taking five seconds, especially not if commits of similar transactions
usually take much less than that.
regards, tom lane