-
Bug
-
Resolution: Fixed
-
Medium
-
None
-
None
-
3
-
9223372036854775807
sanity-quota test_63 (test_dom) fails on ZFS with:
FAIL: granted 20480 > 9216 after cleanup
raised by wait_quota_released() via wait_quota_release_zfs() (sanity-quota.sh test_dom), the quota-release wait added by LU-17077 (commit 4a930dda75). After the DoM test files are deleted and the ZFS destroys complete, the summed MDT+OST bhardlimit grant for quota_usr stays at 20480 and never drops to the 9216 maximum within the wait period.
Seen twice, with identical numbers, in the last 30 days:
- lustre-master full-dne-zfs-part-2 (el10.1), 2026-10-29, build 4767: https://testing.whamcloud.com/test_sets/0d9c1b70-70fa-4ebc-a931-52d39d91bf41
- lustre-reviews review-dne-zfs-part-4 (el10.1), 2026-10-30, change 67603 ps11 build 134758: https://testing.whamcloud.com/test_sets/f639bb46-0fa0-433d-bd2c-2f13f7d75963
The master hit shows it is not caused by change 67603 (an mdd/connect-flag patch with no quota code). Rate is low: about 0.2% of lustre-master suite runs, and test_63 fails 1.3% in this review group overall, almost all with other errors.
Not LU-17077 ("write failed, expect succeed") and not LU-21011 (sanity-quota timeout). Either quota grant is not being released on the slave after DoM deletes on ZFS, or the 9216 limit in the wait is too tight.
Raised by Patch Watcher run pw-engineer-67603-ps11-43c58bed316a on https://review.whamcloud.com/67603
VVVVVVV DO NOT REMOVE LINES BELOW, Added by Maloo for auto-association VVVVVVV
sanity-quota test_63 - granted 20480 > 9216 after cleanup