Comment 2 for bug 943794

Gunicorn support (with more than 1 worker process) is unfortunately limited to completely 'installed' databases. It is not possible to use it to create a new database, or to update any data for which the server has a local cache (e.g. record rules or translations).

We have a patch already available to coordinate the multiple processes so that their state (model pool and caches) is kept in sync at https://code.launchpad.net/~openerp-dev/openobject-server/trunk-gunicorn-signaling-vmt

This also requires a little change change client-side to use a blocking operation when creating a new database instead of a polling one (which would hit different (not in sync) workers.