fix(celery-worker): Should mount the /sshkeys just like the server

[abellotti]

• Celery worker should be mounting the /sshkeys just like the server as tasks would require that if configured to authenticate as such.

[mirekdlugosz]

+1

Obviously not blocking, but I wonder if these units should be generated from single source? Would also make it easier to deploy multiple Celery workers. The only difference between API server and Celery worker is, and should be, the entrypoint.

systemd has the concept of template units (Fedora Magazine). I don't know if you can merge them with quadlets, but then you could have discovery@api, discovery@worker-1, discovery@worker-2 etc.

[abellotti]

The original prototype with Quadlet had templates implemented for celery-worker for scalability, with script to scale the workers, discovery-celery-worker@1, discovery-celery-worker@2, etc. However, this was overkill as we're running on a single system (vs. OCP pods and such). we decided to use the Celery process auto-scaler and a single pod.

See commit 0642de4 for the switch.