[LU-15728] 'relatime' is not working properly

• Type: Bug
• Resolution: Fixed
• Priority: Minor
• Fix Version/s: Lustre 2.16.0
• Affects Version/s: Lustre 2.12.7, Lustre 2.15.0
• Severity: 3
• Assignee: Aurelien Degremont (Inactive)
• Reporter: Aurelien Degremont (Inactive)

Description

relatime behavior is properly managed by VFS, however Lustre also stores acmtime on OST objects and atime updates for OST objects should honor relatime behavior.

ci_noatime feature was introduced by LU-3832 to properly honor noatime option for OST objects. However, the relatime management was not added at the same time, even if it uses the same function.

relatime could be fixed adding the missing parts from touch_atime() to also support relatime.

Issue Links

• is related to: LU-1783 sanity test_39l failed: atime is not updated

Comments

[Peter Jones]

Landed for 2.16

[Gerrit Updater]

"Oleg Drokin <[e-mail]>" merged in patch https://review.whamcloud.com/c/fs/lustre-release/+/50009/
Subject: LU-15728 mdd: fix sanity-56oc failure
Project: fs/lustre-release
Branch: master
Current Patch Set:
Commit: 8beeec77c3b426a01e1f10ca51149c7ca7e01b7e

[Aurelien Degremont (Inactive)]

Patch to fix the regression: https://review.whamcloud.com/c/fs/lustre-release/+/50009/

[Andreas Dilger]

Reopen to track sanity test_56oc failure triggered by this patch.

[Andreas Dilger]

Any tips to find all failing occurences in Maloo with some query? I'm failing to do it.

Two easy ways to find similar failures. For any subtest failure report in Maloo, you can click on the failed subtest number and it will generate a search for failures of that subtest in the past week. You can then modify the query for specific date ranges, or limit it to specific branches, etc. There is a "link to this page" at the bottom to save the URL for future reference, like the following which searches from 2023-01-15 to 2023-02-15 and shows that the problem was hit once by the patch before landing on 2023-02-01, and then by other patches starting on 2023-02-07 after the patch landed:

https://testing.whamcloud.com/search?client_branch_type_id=24a6947e-04a9-11e1-bb5f-52540025f9af&server_branch_type_id=24a6947e-04a9-11e1-bb5f-52540025f9af&status%5B%5D=FAIL&test_set_script_id=f9516376-32bc-11e0-aaee-52540025f9ae&sub_test_script_id=cb9003f2-f26b-11e9-b62b-52540065bddc&start_date=2023-01-15&end_date=2023-02-15&source=sub_tests#redirect

Alternately, you can get to this search page from the "Results->Search" link at the top of any Maloo page, and then select "Sub tests" to limit the query to specific subtest failures, as opposed to whole "Test sets" failures.

[Aurelien Degremont (Inactive)]

I think i found the reason.

 

 It also forces atime to disk on MDD if ondisk atime is older than ondisk mtime/ctime to match relatime (even if relatime is not enabled)

This code change shows the culprit: https://review.whamcloud.com/c/fs/lustre-release/+/47017/1/lustre/tests/sanity.sh#5123

That's a change that Andreas asked for in https://review.whamcloud.com/c/fs/lustre-release/+/47017/1/lustre/tests/sanity.sh#5123

 

I reproduced the failure and did not without that change in that Gerrit patch (https://review.whamcloud.com/c/fs/lustre-release/+/50009/2)

 

I need to think about it

 

[Alex Zhuravlev]

subtest search in Maloo for sanity/test_56oc ?