Closed matyasselmeci closed 3 weeks ago
I just remembered that the containers use tini not supervisord. I'd like to avoid switching back to supervisord just to run a logrotate cron job. Since it's our deployment that configures pelican to log to a file, maybe our deployment should handle the log rotation via a sidecar container.
I created a Jira issue to track it: https://opensciencegrid.atlassian.net/browse/SOFTWARE-5887 It's not really Pelican's job, it should be fixed in the deployment. It is, however, something to keep in mind if we start shipping things like Helm charts or other pre-packaged Kubernetes config.
Pelican Service:
Logs, especially xrootd logs, grow fairly quickly, filling up /var/log/pelican if it's on a PVC. The RPM installs have logrotate configs; we should add log rotation to the containers as well.