[LU-2003] conf-sanity 21d @@@@@@ FAIL: import is not in FULL state

• Type: Bug
• Resolution: Cannot Reproduce
• Priority: Minor
• Affects Version/s: Lustre 2.3.0, Lustre 2.4.0
• Labels: tests
• Environment: MDS and MGS are the same server but two different disk are used for the MDT and MGT.
• Severity: 3
• Assignee: Jian Yu
• Reporter: James A Simmons

Description

Test conf-sanity 21d failed with

conf-sanity test_21d: @@@@@@ FAIL: import is not in FULL state
Trace dump:
= ./../tests/test-framework.sh:3709:error_noexit()
= ./../tests/test-framework.sh:3731:error()
= ./../tests/test-framework.sh:4764:wait_osc_import_state()
= ./conf-sanity.sh:721:test_21d()
= ./../tests/test-framework.sh:3967:run_one()
= ./../tests/test-framework.sh:3997:run_one_logged()
= ./../tests/test-framework.sh:3819:run_test()
= ./conf-sanity.sh:731:main()

Comments

[James Nunez (Inactive)]

James (Simmons),

Have you seen this error recently in your testing? I've tried over the past two days to hit this error with the latest master running conf-sanity and just conf-sanity test 21d alone and can't trigger it. As you can see, Yu, Jian can't trigger the problem either.

If you haven't seen this error, please let me know if you are comfortable closing this ticket.

Thanks,
James (Nunez)

[Jian Yu]

Still cannot reproduce the failure on master branch (build #2052) by running conf-sanity from test 0 to 22 for more than 10 times. And also failed to reproduce the failure on Lustre 2.4.3.

[Jian Yu]

I'm sorry for the late reply. I've manually running conf-sanity test 21d alone with separate MGS and MDT on the latest master branch for more than 10 times but did not reproduce the failure. I'll do more experiments.

[James Nunez (Inactive)]

After several runs of conf-sanity and running conf-sanity test 21d alone, I'm able to reproduce this error. I have a separate but co-located MGS and MDS as in the original setup.

Results for this failure are at https://maloo.whamcloud.com/test_sessions/cc509aae-6806-11e3-a01f-52540035b04c

[James Nunez (Inactive)]

From dmesg on the OSS:

[ 1207.629766] Lustre: DEBUG MARKER: == conf-sanity test 21d: start mgs then ost and then mds == 10:53:04 (1348239184)
[ 1467.507155] Lustre: DEBUG MARKER: rpc : @@@@@@ FAIL: can't put import for osc.lustre-OST0001-osc-MDT0000.ost_server_uuid into FULL state after 140 sec, have DISCONN
[ 1469.733721] Lustre: DEBUG MARKER: conf-sanity test_21d: @@@@@@ FAIL: import is not in FULL state

[Peter Jones]

Yujian

Could you please look into this one?

Thanks

Peter