Site keeps stopping since upgrade

[M0VSE (Phil Taylor)]

I have a site that has been running really well for 18 months+

I noticed that it was running on Debian buster so upgraded to bullseye and Discourse to 3.1.0.beta3.

Since then, each morning, I noticed that the site is down and have been trying to diagnose why. I am getting lots of strange errors in the logs:

./launcher logs app shows lots of:

svlogd: warning: unable to lock directory: /var/log/postgres: temporary failure
svlogd: fatal: no functional log directories.
svlogd: warning: unable to lock directory: /var/log/redis: temporary failure
svlogd: fatal: no functional log directories.

In /var/log/postgres/current I see:

2023-04-10 10:48:23.733 GMT [6918] LOG:  skipping missing configuration file "/shared/postgres_data/postgresql.auto.conf"
2023-04-10 10:48:23.734 UTC [6918] FATAL:  data directory "/shared/postgres_data" has wrong ownership
2023-04-10 10:48:23.734 UTC [6918] HINT:  The server must be started by the user that owns the data directory.

And in /var/log/redis/current:

3787:M 10 Apr 2023 10:49:09.043 * 100 changes in 300 seconds. Saving...
3787:M 10 Apr 2023 10:49:09.044 * Background saving started by pid 7014
7014:C 10 Apr 2023 10:49:09.045 # Failed opening the RDB file dump.rdb (in server root dir /shared/redis_data) for saving: Permission denied
3787:M 10 Apr 2023 10:49:09.144 # Background saving error

The owner of /shared/postgres_data is postgres:postgres and redis_data is redis:redis

I have tried ./launcher rebuild app multiple times.

Please help!

Thanks

Phil

[M0VSE (Phil Taylor)]

OK I found the problem!

When I first installed discourse, I created an ‘import’ container to import messages from groups.io. I removed this container and deleted the yml files from the containers folder, and it has been running fine (with a number of reboots) since.

Something in the docker upgrade appears to have re-activated it, I ran docker container list and there was a container called import!

I deleted it from docker and restarted the app container and all seems to be well.

Thanks

Phil