Hello,

I have an Odoo 11 Community instance deployed in a Docker container stored in a cloud service, with a Point of Sale, 9K products and 14k customers, and the POS crashes when trying to load the res.partner model.

At first I got a "psycopg2.InterfaceError: connection already closed" error, which I fixed by activating workers in odoo.conf (This may be useful for someone. Remember setting up and opening ports for longpolling too!). That's when things get interesting: That error doesn't show anymore, and the POS works fine, but only if I access Odoo directly through the IP/port. That is, if I access through http://200.111.222.333:9000 in my browser (obviously the numbers aren't the real ones).  When I try to access through http://myodooinstance.mydomain.com, the POS crashes, but instead I get a 502 error.

I've seen this same error in other situations, particularly when dealing with slow, large calls - trying to backup a database for example, or importing a large CSV. Another interesting thing I've found when facing this issue, is that the call is still being processed after returning error. For example, imagine I import 5000 products with a CSV. An error shows up after a while, but if I check minutes later I find out products are still being added in the background, without giving any notification or log registry. In some more extreme cases, this happens even when accessing through IP:port.

From all of this I'm inclined to think it's a timeout issue. My question is, does anyone know what's triggering this timeout? Is it a misconfiguration on the DNS, the container, the host server?

Thanks a lot.