Re: Client deadlocks when connecting via ssl

On Sat, Apr 25, 2015 at 12:08 AM, Peter J. Farrell
<pfarrell(at)greatbiztools(dot)com> wrote:
> Amjith Ramanujam wrote on 10/08/2014 03:47 PM:
>> I'm accessing Postgresql from Python using Psycopg2. I'm making
>> multiple calls from threads and once every few times it causes the
>> script to lock up and become unresponsive. This only happens if I'm
>> also making HTTPS calls in addition to Postgres TCP calls with SSL.
>
> Just wanted to add that we're seeing this libpq pq_lockingcallback bug
> in our production web apps on Heroku using Gunicorn, New Relic and
> Postgres on Heroku. We've relieved some of the symptoms by implementing
> database connection pooling with high timeouts and switching from
> Gunicorn to UWsgi so we are Harakiri rogue workers that are deadlocked.
> Hoping this is fixed eventually...
>
> Initial Background:
> http://blog.maestropublishing.com/2015/01/12/gunicorn-dyno-death-spiral-on-heroku/
>
> Discovery of Bug:
> http://blog.maestropublishing.com/2015/01/16/gunicorn-dyno-death-spiral-on-heroku-part-ii/

Hello,

there is a pull request open on psycopg that could be what needed to
fix this problem: https://github.com/psycopg/psycopg2/pull/290

Could you please test if it fixes the issue? I can merge the patch and
release 2.6.1 in a few days.

Thank you,

-- Daniele