[LU-13967] Tests fail/hang with “can't notify - lge_glbl_data is not set” Created: 15/Sep/20 Updated: 25/Sep/20 Resolved: 25/Sep/20 |
|
| Status: | Resolved |
| Project: | Lustre |
| Component/s: | None |
| Affects Version/s: | Lustre 2.14.0 |
| Fix Version/s: | Lustre 2.14.0 |
| Type: | Bug | Priority: | Major |
| Reporter: | James Nunez (Inactive) | Assignee: | Sergey Cheremencev |
| Resolution: | Fixed | Votes: | 0 |
| Labels: | None | ||
| Severity: | 3 |
| Rank (Obsolete): | 9223372036854775807 |
| Description |
|
We are seeing the message (qmt_entry.c:805:qmt_adjust_edquot_qunit_notify()) lustre-QMT0000: can't notify - lge_glbl_data is not set in the MDS console log for many test suites starting on 19 AUG 2020. This message was introduced by " Some tests like sanity-pcc test 5 at https://testing.whamcloud.com/test_sets/b8d571f0-46fe-4651-a545-24c8142eebf8 fail or hang after this comment is printed, but it’s not clear if the statement is actually warning of a real problem or if the tests are failing/hanging for a different reason. For the test session https://testing.whamcloud.com/test_sessions/42070a37-3493-47fe-a89f-5b0d4c869bb8, we see messages like the following seven times in the MDS console log: [31796.726146] Lustre: DEBUG MARKER: /usr/sbin/lctl mark SKIP: sanity-sec test_33 need shared key feature for this test [31796.769617] Lustre: format at qmt_entry.c:805:qmt_adjust_edquot_qunit_notify doesn't end in newline [31796.769631] Lustre: format at qmt_entry.c:805:qmt_adjust_edquot_qunit_notify doesn't end in newline [31796.769634] Lustre: 4891:0:(qmt_entry.c:805:qmt_adjust_edquot_qunit_notify()) lustre-QMT0000: can't notify - lge_glbl_data is not set[31796.773585] Lustre: format at qmt_entry.c:805:qmt_adjust_edquot_qunit_notify doesn't end in newline [31796.773626] Lustre: format at qmt_entry.c:805:qmt_adjust_edquot_qunit_notify doesn't end in newline We need to determine if this message is a sign of a real problem of if it should be suppressed. |
| Comments |
| Comment by Sergey Cheremencev [ 15/Sep/20 ] |
|
This might happen if no slaves have enqueued global quota locks yet. I don't now any problems that may be caused in such case - we just can't reseed lqe_glbl_data with new edquot or qunit as it doesn't exist at this moment. Such case is valid only for quota report and release - quota acquire requires enqueued lock. I think this warning should be changed to CDEBUG(D_QUOTA, ...).
|
| Comment by Gerrit Updater [ 15/Sep/20 ] |
|
Sergey Cheremencev (sergey.cheremencev@hpe.com) uploaded a new patch: https://review.whamcloud.com/39921 |
| Comment by Gerrit Updater [ 25/Sep/20 ] |
|
Oleg Drokin (green@whamcloud.com) merged in patch https://review.whamcloud.com/39921/ |
| Comment by Peter Jones [ 25/Sep/20 ] |
|
Landed for 2.14 |