[LU-2674] sanity-quota test_1: @@@@@@ FAIL: user write success, but expect EDQUOT

• Status: Resolved
• Project: Lustre
• Affects Version/s: Lustre 2.4.0
• Fix Version/s: Lustre 2.4.0
• Type: Bug
• Priority: Blocker
• Reporter: Keith Mannthey (Inactive)
• Assignee: Niu Yawei (Inactive)
• Resolution: Fixed
• Labels: HB

Environment

A patch pushed to git.

• Severity: 3

Description

Auto test for a patch (that contains an unrelated test update) against Master encountered.

sanity-quota test_1: @@@@@@ FAIL: user write success, but expect EDQUOT

The error can be seen here:
https://maloo.whamcloud.com/test_sets/d16db8e0-660f-11e2-903d-52540035b04c

The error was reported as

Error: 'user write success, but expect EDQUOT'
Failure Rate: 10.00% of last 100 executions [all branches]

It seems there may be a quota issue.

Comments

[Keith Mannthey (Inactive)]

Also about of a 1/4 of the other sanity-quota tests failed in this test run some with EDQUOT errors some differently.

[Peter Jones]

Any ideas?

[Niu Yawei (Inactive)]

This is caused by recent commit: http://review.whamcloud.com/#change,4922, look at the maloo test result for this patch, the sanity-quota.sh was skipped because e2fsprogs on the test system isn't uptodate.

12:39:41:-----============= acceptance-small: sanity-quota ============----- Tue Jan 22 12:39:36 PST 2013
12:39:41:Loading modules from /usr/lib64/lustre
12:39:41:detected 1 online CPUs by sysfs
12:39:41:libcfs will create CPU partition based on online CPUs
12:39:52:debug=0x33f0404
12:39:52:subsystem_debug=0xffb7e3ff
12:39:52:gss/krb5 is not supported
12:39:52:Running: bash /usr/lib64/lustre/tests/sanity-quota.sh
12:39:52:CMD: wtm-5vm3 /usr/sbin/lctl get_param -n version
12:39:52:debugfs 1.41.12 (17-May-2010)
12:39:52:
12:39:52: SKIP: sanity-quota  e2fsprogs doesn't support quota
12:39:52:sanity-quota returned 0
12:39:52:running: sanity-sec 
12:39:52:run_suite sanity-sec /usr/lib64/lustre/tests/sanity-sec.sh

Hi, Chris, could you make sure that all our test systems have correct e2fsprogs now?

Actually, this problem can be fixed by: http://review.whamcloud.com/#change,4923 (see the lustre_lwp_connect(), we shouldn't pass OBD_CONNECT_IBITS connect flag for the lightweight osp connection, because quota enqueues plain locks over this connection).

[Niu Yawei (Inactive)]

Instead of waiting for http://review.whamcloud.com/#change,4923, I pushed a quick fix: http://review.whamcloud.com/#change,5167

[Keith Mannthey (Inactive)]

I had another patch hit this and I believe this is blocking a good deal of testing right now for others. Can the root patch (http://review.whamcloud.com/#change,4922) be dropped while this gets sorted out properly?

[Niu Yawei (Inactive)]

Given that 5167 is ready for landing now, I think we'd land this small fix rather than drop the big DNE patches.

BTW, the test shows Rosso Autotest is still skipping quota tests for incorrect e2fsprogs version. I think we'd fix it asap.

[Niu Yawei (Inactive)]

patch landed.

[Andreas Dilger]

Filed TT-1081 to track the issue about e2fsprogs not being installed.

[Andreas Dilger]

Niu, the tests that are actually running sanity-quota.sh are still failing with this error.

https://maloo.whamcloud.com/test_sets/bf8260bc-6d09-11e2-83d0-52540035b04c
https://maloo.whamcloud.com/test_sets/de653550-6bf0-11e2-b2a5-52540035b04c

[Niu Yawei (Inactive)]

The failed patch's parent is "abe64080697d28cbe8eab9ffae1318329e9ca997 LU-398 ptlrpc: Add the NRS CRR-N policy" -> "fadd88fe410127322876ada077dc6dc13050ecd3 LU-398 ptlrpc: Add the NRS framework and FIFO policy" -> "20040d7f9529b933fd9778da0d3f3be0df4d876b LU-2561 mdd: create decares attr set with parent attr", which didn't include the fix. Seems it needs be rebased from the first patch.