Re: Race condition between repack and loose-objects maintenance task

On Wed, Jun 29, 2022 at 01:03:54PM -0400, Taylor Blau wrote:

I haven't looked at the maintenance code too closely as of yet, but I have a plausible explanation for why maintenance is removing loose *unreachable* objects.

Ordinarily the loose-objects task alone will not remove any loose objects which don't already appear in a pack. That's because we first try to run prune-packed, and then the equivalent of `git repack -d` to consolidate loose objects together into a single pack.

But, we only do that for the first 50,000 loose objects. Anything else gets left around and then is likely removed by the `git gc` task which is run separately, and after the loose-objects task.

Using cruft packs won't change the race fundamentally, but storing the unreachable objects in a cruft pack will mean that you can keep more objects around stored together in a single pack. In other words, making it safe to increase the grace period beyond what you would ordinarily be able to do with the classic pruning mechanism.

Thanks, Taylor