I had bookmarked ucacher as a really interesting caching approach and I'm trying it out on a workflow with quite a lot of jobs. While I'm seeing some really promising results for a few of our time-consuming steps, the number one blocker so far is that I'm getting this error quite frequently:

failed to upload DB to backend: failed to put database file into backend: failed to reserve GHA cache: failed to delete existing cache entry: HTTP 404: 404 Not Found

This can happen both after running a command (because of a cache miss) as well as after successfully restoring a cache and not running the command.

My best guess is that this is a concurrency problem: our workflow starts a number of jobs in parallel, and they might all start by installing packages or building an app. In other words, they run the same long-running command at the same time.

My impression is that the above error occurs under the same circumstances as when the "official" actions/cache says Failed to save: Unable to reserve cache with key some-key-here, another job may be creating this cache. Having said that, this doesn't explain that I've also seen the error occur after restoring a cache. With ucacher being closed-source, I have no idea what it's doing, of course.