As it has been discovered, regular update of postgresql cause issues when
the database gets restarted, cause the containers can no longer correctly
resolve the hostname and reconnect, therefore use ‹PartOf› to truly¹
force a restart of the container once the database gets restarted.
¹ Tried to fix a similar issue once already in 737acfd, based on the
observations, I have failed…
Signed-off-by: Matej Focko <me@mfocko.xyz>
Storing data for Forgejo and Vaultwarden in the postgres database results
in a rather wonky behavior during restarts of the VPS and potential
restarts of the postgres itself.
Based on some observations Forgejo is capable of recovering from such
issues, but Vaultwarden falls into a retry loop trying to reconnect to
the database while failing on resolving the hostname / connecting, e.g.,
[2024-11-27 16:52:43.646][r2d2][ERROR] connection to server at "host.containers.internal" (XXX.XXX.XXX.XXX), port 5432 failed: Connection refused
Therefore adjust the quadlet definitions to require and boot containers
»after« the postgres is running to minimalize the possible issues.
Signed-off-by: Matej Focko <me@mfocko.xyz>
This variable was not required when running from the user-space, but
after migrating to quadlets, I had to define it.
Just to satisfy the requirements, I set it to 1. Based on some docs I
have managed to find, it appears that the default is 10. Given the
amount of users of my deployment and the fact it's running on the VPS,
setting it to 8.
Signed-off-by: Matej Focko <me@mfocko.xyz>
