Jack Orenstein <jorenstein@archivas.com> writes:
> I'm not sure I did this right, (it's been a while since I worked with
> gdb). But here's what I found.
Doesn't look like either of those can be trusted very far :-(. But both
of them seem to be waiting for input.
What this looks like to me is some kind of networking problem. Perhaps
the connection to the remote host got dropped and the local kernel
hasn't realized it yet? From memory, we use TCP KEEPALIVE on the
server socket but not in client-side code, so there's probably not any
timeout that will trigger in anything like a sane amount of time.
On the other hand, if the remote backend didn't die immediately, it
*would* have recognized a network timeout and gone away in at most an
hour or so. Did you check for any "unexpected EOF on client connection"
or similar messages in the remote postmaster's log?
regards, tom lane