getsentry
How to handle disk space usage of debug files
Problem Statement
I build for 3 Unity platforms, and the debug files that gets uploaded per platform is over 1gb. So per build that is over 3gb of debug files that gets uploaded.
I typically do beta builds around once a day or so. Which adds up to over 100gb of debug files per month.
What are my options for handling this? I only really care about the most recent beta and production debug files, and not previous builds.
Solution Brainstorm
No response
You could have a scheduled cron job that cleans up those debug files. If anything is older than a few days old. Something like https://github.com/getsentry/self-hosted/blob/da06c0f230454b91fada3980ba2e1e012cedf1d2/docker-compose.yml#L481, but cleaning up the sentry-data volume instead, which is what I think houses debug files.
So there's no built in system for it, for example something like SENTRY_EVENT_RETENTION_DAYS, but that is for debug files instead?
And I don't think deleting anything older than a couple days on its own would work? As according to documentation, debug files are deleted when not used for more than 90 days, rather than being 90 days old. So I would want to delete any files not used for more than 2 weeks or so potentially, which should cover making sure they are always available if needed.
There is not a built in system for this, no. I have looked into this more, and I believe our sentry cleanup job that runs periodically that should clean up the models corresponding to the files and debug file models that live in postgres. However, I believe in order to actually free up disk space you would need to do a VACUUM FULL or something similar to ensure that the disk space is freed up.
https://www.rdiachenko.com/posts/databases/postgresql/postgresql-does-not-free-up-physical-space-after-delete/#:~:text=Q%3A%20How%20Do%20I%20Reclaim,frees%20space%20for%20future%20use.
Could I propose this being added as a feature then? Seems like it would be useful for anyone uploading debug files on a regular basis. I have no knowledge of the code base, but I would assume it would just be converting whatever is running the duration for cleaning up debug files to use an environment variable instead so users can configure it.
Sure, this can be a feature request. Please be aware this is probably not something we can get to in the near term though.