[LU-13543] hsm.actions file is broken on RHEL 8.2

• Type: Bug
• Resolution: Fixed
• Priority: Critical
• Fix Version/s: Lustre 2.14.0
• Affects Version/s: Lustre 2.14.0
• Environment: RHEL 8.2
• Severity: 3
• Assignee: Mikhail Pershin
• Reporter: Jian Yu

Description

This issue was created by maloo for jianyu <[e-mail]>

This issue relates to the following test suite run: https://testing.whamcloud.com/test_sets/5690066f-e39c-4964-86b6-ceb2f961126c

test_90 failed with the following error:

CMD: trevis-66vm4 /usr/sbin/lctl get_param -n mdt.lustre-MDT0000.hsm.actions | egrep 'WAITING|STARTED'
pdsh@trevis-66vm1: trevis-66vm4: ssh exited with exit code 1
CMD: trevis-66vm4 /usr/sbin/lctl get_param -n mdt.lustre-MDT0000.hsm.actions | egrep 'WAITING|STARTED'
pdsh@trevis-66vm1: trevis-66vm4: ssh exited with exit code 1
Cannot send HSM request (use of /mnt/lustre/d90.sanity-hsm/f90.sanity-hsm.1): Operation not permitted
 sanity-hsm test_90: @@@@@@ FAIL: cannot release a file list 

https://testing.whamcloud.com/test_sets/6badc2b6-c8a6-40db-a740-205913d2f371

VVVVVVV DO NOT REMOVE LINES BELOW, Added by Maloo for auto-association VVVVVVV
sanity-hsm test_90 - cannot release a file list

Issue Links

• is duplicated by: LU-13985 seq_file next function must change *pos
• is duplicated by: LU-14086 sanity-hsm test 260c fails with 'request on 0x200001b71:0x2b1:0x0 is not SUCCEED on mds1'
• is duplicated by: LU-13544 sanity-hsm test_104: Data field in records is () and not ([434541])
• is duplicated by: LU-13545 sanity-hsm test_260c: FAIL: request on 0x200000404:0x2bb:0x0 is not SUCCEED on mds1

Comments

[Peter Jones]

Landed for 2.14

[Gerrit Updater]

Oleg Drokin ([e-mail]) merged in patch https://review.whamcloud.com/40757/
Subject: LU-13543 lustre: update *pos in seq_file .next functions
Project: fs/lustre-release
Branch: master
Current Patch Set:
Commit: d5d0ff24a84f64e5196341f5ce946952d7fff8b7

[Gerrit Updater]

Mike Pershin ([e-mail]) uploaded a new patch: https://review.whamcloud.com/40757
Subject: LU-13543 lustre: update *pos in seq_file .next functions
Project: fs/lustre-release
Branch: master
Current Patch Set: 1
Commit: 31b0a733dd6bc6eae6bd8b1a606da7e6d274c910

[John Hammond]

This is likely due to the inclusion of https://github.com/torvalds/linux/commit/1f4aace60b0edc2d885aaa263abf4df42c8c65a8 in el8.2. Reading from the debugfs file directly while using Lustre trace and strace shows that hsm_actions_show_cb() is called lots of times but read() returns 0. When there are lots of records in the log then none are returned from read().

[John Hammond]

These tests (90, 104, 260c) consistently:

1. Pass with el8.1 client and el8.1 server: https://review.whamcloud.com/#/c/38675/
2. Fail with el8.1 client and el8.2 server: https://review.whamcloud.com/#/c/38676/

Test 52 consistently passes with 8.1/8.1, but failed 1/4 times with 8.1/8.2.

[Ben Evans (Inactive)]

Yes, I'm working on that now.

[Jian Yu]

Hi Ben,
It's likely a timing issue because the whole test session was run with ssh by autotest and only those three sub-tests failed regularly. Could you please improve the test scripts to resolve the issues?

[Ben Evans (Inactive)]

I think this is either the ssh issue mentioned above, or a timing issue.  I was able to run the test by hand with no issues.