Hi Bitbucket team,
Could a Bitbucket engineer please run garbage collection on sbc-admin/koitomo-backend? We can't open a support ticket - the form needs a workspace admin to grant dataaccess, and our workspace owner is in another timezone.
We rewrote history with BFG Repo-Cleaner to drop superseded large binaries, thenforce-pushed every branch. The rewrite worked, but the reported size went UPrather than down, because the now-unreachable objects are still stored:
before : 4,261,598,005 bytes
after : 4,315,626,152 bytes
limit : 4,294,967,296 bytes (over by ~20.7 MB)
Every push is now rejected, so the whole team is blocked. Running`git gc --prune=now` on the rewritten mirror locally takes it from 4.0 GB to1.5 GB, so a server-side collection should land this well under the limit.
Automatic GC has not run. We triggered the documented condition twice - aref-adding push, then a ref-removing push. Both were accepted and both returnedthe "Garbage collection will run automatically ... can take up to an hour"message, but more than four hours later the reported size and updated_on areunchanged. (The size did move by 192 bytes on the ref deletion, so it is beingrecomputed - just nothing is being collected.)
We also checked that we are not hitting the problem from this thread, where a GCfreed almost nothing because the old objects were still anchored by retained
refs/pull-requests/*:
https://community.atlassian.com/forums/Bitbucket-questions/Repository-size-not-reduced-after-history-rewrite-request-server/qaq-p/3266582
Here the pre-rewrite branch tips are not fetchable by SHA - while a current mid-history commit is, so the check is meaningful - and refs/pull-requests/* advertises no refs. The old objects are genuinely unreferenced, so a collectionshould reclaim them in full.
Thanks very much.
