Config fix for postgres docker defaults

[Fmstrat]

Default Docker configurations in Alpine have a different override path, which causes problems if people switch to the Ansible deployment.

Reproduction of the issue:

• Fire up a Lemmy DB in Docker without a volume mount for /etc/postgresql.conf
• Use SHOW config_file to see the location of the config file (it will be in /var/lib/data)
• Take down the DB, and add a volume mount for /etc/postgresql.conf
• Use SHOW config_file to see the location of the config file (it will still be in /var/lib/data)

Also, official docker instructions for Postgres is to use /etc/postgresql/postgresql.conf, though this is irrelevant given this PR.

To force the use of /etc/postgresql.conf for if users switch from bare mode to Ansible, the compose file was edited to include a forced command.

In addition, the default for PG15 alpine listen_address is localhost, not *, which is required for the lemmy container to connect. This default is overwritten by the docker container normally, but is not included in the tuning samples provided.

NOTE: It is highly possible there are a number of smaller self-hosted instances who have no idea their poor performance is because PG is not properly tuned, as the defaults are pretty meager.

[codyro]

Taking a quick look, I think we may run into issues with the /etc/postgresql.conf being mounted with permissions the postgres user cannot read:

/ # ls -lh /etc/postgresql.conf
-rw-------    1 1000     1000         733 Oct  2 18:12 /etc/postgresql.conf

/ # grep postgres /etc/passwd
postgres:x:70:70:Linux User,,,:/var/lib/postgresql:/bin/sh

2023-10-07 20:01:21.001 GMT [1] LOG:  could not open configuration file "/etc/postgresql.conf": Permission denied
2023-10-07 20:01:21.001 GMT [1] FATAL:  configuration file "/etc/postgresql.conf" contains errors

The playbooks are distributing the postgres configuration with 1000:1000 ownership and 0600 permissions (https://github.com/LemmyNet/lemmy-ansible/blob/main/lemmy.yml#L222-L224 & https://github.com/LemmyNet/lemmy-ansible/blob/main/lemmy-almalinux.yml#L189-L191). I see no reason why this is needed, so we could either change the ownership to 70:70 and keep the 0600 permissions or just set the user and owner to root with 0644 permissions (IMO the most logical).

Could you fix this in your PR?

EDIT

I realized you might have been coming from the primary project, which references this docker-compose.yml. If you're uncomfortable adjusting the playbooks (or don't have time), let me know, and I'll submit a separate PR to fix it :).

Good luck with the move!

[Fmstrat]

Happy to do it, while I don't deploy lemmy this way (I just happened to notice it) I'm prretty familiar with ansible. I'll knock it out after the move. And thanks!

[ticoombs]

Thanks @Fmstrat ! I've been meaning to fix this up. As i've helped solve this on the matrix rooms multiple times.

Once #186 is merged please rebase and we'll merge