[LU-2540] test: sanity FAIL: test_74c Touch successful

• Type: Bug
• Resolution: Cannot Reproduce
• Priority: Minor
• Affects Version/s: Lustre 2.3.0
• Labels: tests
• Environment: server: lustre-2.1.3-1nasS centos 6.3; client: lustre-2.3.0-1nasC sles11sp2 3.0.42-0.7.3
• Severity: 3
• Assignee: Bob Glossman (Inactive)
• Reporter: Jay Lan (Inactive)

Description

sanity.sh test 74c failed: FAIL: test_74c Touch successful.

/var/log/messages at the client showed:
Dec 27 15:54:48 service331 kernel: [ 1397.672701] Lustre: DEBUG MARKER: == sanity test 74c: ldlm_lock_create error path, (shouldn't LBUG) ==================================== 15:54:48 (1356652488)
Dec 27 15:54:48 service331 kernel: [ 1397.719360] Lustre: *** cfs_fail_loc=319, val=0***
Dec 27 15:54:48 service331 kernel: [ 1397.719385] LustreError: 9373:0:(mdc_locks.c:773:mdc_enqueue()) ldlm_cli_enqueue: -12
Dec 27 15:54:48 service331 kernel: [ 1397.742866] LustreError: 9373:0:(file.c:2333:ll_inode_revalidate_fini()) failure -12 inode 29
Dec 27 15:54:48 service331 kernel: [ 1397.898789] Lustre: DEBUG MARKER: sanity test_74c: @@@@@@ FAIL: Touch successful
Dec 27 15:54:50 service331 kernel: [ 1399.583195] Lustre: DEBUG MARKER: == sanity sanity.sh test complete, duration 7 sec ==================================================== 15:54:50 (1356652490)

Comments

[Peter Jones]

Thanks Jay!

[Jay Lan (Inactive)]

This can be closed. I have not seen failure on this test for a long while... maybe since 2.4?

[Dmitry Eremin (Inactive)]

Made this test more robust in patch http://review.whamcloud.com/9766/

[nasf (Inactive)]

Another failure instance:

https://maloo.whamcloud.com/test_sets/27eed06a-efc1-11e2-b957-52540035b04c

[Nathaniel Clark]

https://maloo.whamcloud.com/test_sets/348ae5ee-c1b2-11e2-9e40-52540035b04c
master (circa 2.4.50)

[Bob Glossman (Inactive)]

When I put up a pure b2_3 sles11 sp2 client I still wasn't able to reproduce the problem.

The msg output reported suggests the client is seeing -12 (-ENOMEM), which is the expected error generated by the fail_loc value set in the test. strace output of the touch command suggests that error is being reported up to the touch command. I'm not sure how the test could fail that being the case.

[Bob Glossman (Inactive)]

Haven't been able to reproduce this problem with current (2.3.57) Centos servers and sles11 sp2 client. Suggests this problem is probably version specific.

[Peter Jones]

Bob

Could you please look into this one?

Thanks

Peter