Re: some postgres 2s SELECT queries are deadlocking forever in __sendto (libsystem_kernel.dylib)

john gale <john(at)smadness(dot)com> writes:
> We are running a Postgres 9.3.2 server on a Mac Pro running OS X 10.9.0. As an automation results server it's heavily inserted into (30-80 queries per sec), periodically selected from (once every few minutes).
> We created this instance with Postgres 9.3.0 and after a time ran into an issue where some SELECTs that usually completed in 2s were deadlocking, still around hours or days later. I read something that seemed related to this issue under the release notes for 9.3.2 although can't find it again now. After upgrading to 9.3.2, the issue went away for a month.

I don't see any reason to think that's a deadlock. The stack trace shows
the backend is waiting to send data to its client. Since the client also
appears to be waiting for data to come in, it seems like this represents
some kind of network-stack bug. You might try to get Apple interested.
Although the first thing they're gonna ask you is if 10.9.1 fixes it.

regards, tom lane