Purge the artifacts in the cache directory when they are not being used

[adityasood]

Issue Type

• Feature enhancement

Summary

Currently, GoCD does not clean the artifacts stored in the artifacts/cache directory. With setups with larger artifacts being generated in each run of the job, this can eat up quite some space on the GoCD server and cause the server itself to starve. This further causes the server to be preemptive and delete the artifacts causing some jobs to fail.

[ketan]

I did some quick analysis of the current code and here's a summary of the existing implementation:

1. The stages table in the DB keeps track of stages where artifacts have been purged. We run a a query to find the oldest stages that can be purged.

2. For each stage that can be purged, GoCD will attempt to clear the artifacts dir for that stage, and the cache dir. If a delete fails (open file handles, among other reasons,) the cleanup will silently continue. This stage will then be marked as "purged" so we don't clean it up again in the future.

   This implementation another side-effect in that if an artifact cache is created after a stage has been marked as "purged", the cache will not be cleaned up.

---

A possible solution:

ZipArtifactCache (via its superclass ArtifactCache), maintains a list of artifacts currently being zipped. We can possibly use this information to make the object a bit more smarter by:

• adding a LRU cache, to know when a zip file was accessed last
• adding a reference count of which files are being borrowed (for purpose for sending to clients requesting the file)

When a file in the cache expires after a set (configurable) TTL, it can be safely purged (as long as the reference count for that file is 0)

[adityasood]

@ketan's fix solves the problem of open file handles.

[stale]

This issue has been automatically marked as stale because it has not had activity in the last 90 days.
If you can still reproduce this error on the master branch using local development environment or on the latest GoCD Release, please reply with all of the information you have about it in order to keep the issue open.
Thank you for all your contributions.