[LU-12805] replay-single/36 is broken

• Type: Bug
• Resolution: Unresolved
• Priority: Minor
• Labels: always_except
• Severity: 3
• Assignee: WC Triage
• Reporter: Alex Zhuravlev

Description

replay-single/36 seem to be very old, written at times of CFS when all tests were local:
if $LCTL dk | grep "stale lock .*cookie"; then
error "cancel after replay failed"
i.e. it checks for server-side messages on the client.
I noticed that having it failing constantly locally. then tried to change so that server side's lock is grepped: https://review.whamcloud.com/#/c/36127/ – all testing sessions failed in autotest.
AFAIU, ELC is the root cause - it packs cancels, then client doesn't replay cancelled locks but the original pack is sent anyway.
I think we should disable 36 at least.

Comments

[Gerrit Updater]

"Oleg Drokin <[e-mail]>" merged in patch https://review.whamcloud.com/c/fs/lustre-release/+/36291/
Subject: LU-12805 tests: disable replay-single/36
Project: fs/lustre-release
Branch: master
Current Patch Set:
Commit: bdbc7f9f42b9b06aa4d93aabd4c4d559f2bbced1

[Alex Zhuravlev]

for example: https://testing.whamcloud.com/test_sets/31eccc00-dfbd-11e9-a0ba-52540065bddc
CMD: onyx-32vm12 /usr/sbin/lctl dk
00010000:00010000:1.0:1569433691.407285:0:31650:0:(ldlm_lockd.c:1708:ldlm_request_cancel()) ### server-side cancel handler stale lock (cookie 18295147502878251446)
00010000:00010000:1.0:1569433691.407286:0:31650:0:(ldlm_lockd.c:1708:ldlm_request_cancel()) ### server-side cancel handler stale lock (cookie 18295147502878251264)
00010000:00010000:1.0:1569433691.407287:0:31650:0:(ldlm_lockd.c:1708:ldlm_request_cancel()) ### server-side cancel handler stale lock (cookie 18295147502878251418)
replay-single test_36: @@@@@@ FAIL: cancel after replay failed

[Alex Zhuravlev]

hm, the test verifies that there should be no cancels after recovery, right? but they do come due to ELC.

[Andreas Dilger]

I have a patch to fix this I thought? https://review.whamcloud.com/35311

[Gerrit Updater]

Alex Zhuravlev ([e-mail]) uploaded a new patch: https://review.whamcloud.com/36291
Subject: LU-12805 tests: disable replay-single/36
Project: fs/lustre-release
Branch: master
Current Patch Set: 1
Commit: 4017b4b99bd83303db1a1b8eb5305139c20c1b34