You can click on the Google or Yahoo buttons to sign-in with these identity providers,
or you just type your identity uri and click on the little login button.
explain in the documentation that pgbouncer should be used and remove the connection pooler
I have send patches for the approach "make the pooler more intelligent"
We might want to inspect the pooling capability of SQLAlchemy before investigating this one once this ticket is done https://www.cubicweb.org/ticket/17258007
Reference documentation https://docs.sqlalchemy.org/en/13/core/pooling.html
See also https://stackoverflow.com/questions/25767484/how-to-disable-connection-pooling-with-postgresql-in-flask-sqlalchemy and https://dba.stackexchange.com/questions/36828/how-to-best-use-connection-pooling-in-sqlalchemy-for-pgbouncer-transaction-level
I'd not make pgbouncer mandatory as this is not trivial to install and configure.
But rather improve our connection pooler, maybe using SQLAlchemy one like Bram suggest. The principal issue with our connection pooler is that it's statically allocated, per process. It mean we open "connections-pool-size" connections per started process. I think we should have a min/max and ttl for such connections, otherwise use of the pool is inefficient.
Suite à discussion hébdo, le ticket pourrait se transformer en "rendre le connexion pooler de cubicweb plus intéligent", pour corriger deux problèmes:
Le code dans cubicweb est ici: https://hg.logilab.org/review/cubicweb/file/f137b90efdae/cubicweb/server/repository.py#l146
Ticket #17241878 - latest update on 2020/01/07, created on 2019/10/14 by Nicolas Chauvat