[ZODB-Dev] RelStorage PostgreSQLLocker deadlock/wait problem

[Sean Upton]

On Thu, Jul 17, 2014 at 2:37 PM, Leonardo Rochael Almeida
<leorochael at gmail.com> wrote:
> I seem to recall, though I couldn't find references to it now, that this
> could happen if there is a firewall between the RelStorage using
> processes/threads and your database.
>
> If a long time passes without any activity, this could cause the firewall to
> forget about the long lived TCP connections that RelStorage keeps, and drop
> packets silently instead of rejecting them, making the RelStorage end seem
> hung up.

Oddly, in my case, all connectivity to postgres is over a local Unix
domain socket.  I am not sure if there are situations where maybe I'm
inadvertently doing something to cause one committing thread's lock to
get stuck or deadlock.

Sean