[LU-17046] sanity-quota test_1g: user write success, but expect EDQUOT

• Status: Resolved
• Project: Lustre
• Fix Version/s: Lustre 2.16.0
• Type: Bug
• Priority: Minor
• Reporter: Maloo
• Assignee: James A Simmons
• Resolution: Fixed
• Severity: 3

Description

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

This issue relates to the following test suite run: https://testing.whamcloud.com/test_sets/199b4874-fb26-4062-b506-70ca71268b02

test_1g failed with the following error:

user write success, but expect EDQUOT

Test session details:
clients: https://build.whamcloud.com/job/lustre-reviews/97057 - 4.18.0-477.15.1.el8_8.x86_64
servers: https://build.whamcloud.com/job/lustre-reviews/97057 - 4.18.0-477.15.1.el8_lustre.x86_64

<<Please provide additional information about the failure here>>

VVVVVVV DO NOT REMOVE LINES BELOW, Added by Maloo for auto-association VVVVVVV
sanity-quota test_1g - user write success, but expect EDQUOT

Issue Links

• is related to: LU-13917 sanity-quota test_1c: user write succ...
• is related to: LU-15257 sanity-quota: test_1g wait_delete_com...
• is related to: LU-13810 Check OST pool quota hard limit at fi...
• is related to: LU-14279 sanity-quota test_3b: write success, ...
• is related to: LU-17191 sanity-quota test_1b, 1d, 1f, 1i: FAI...

Comments

[James A Simmons]

Note this was opened Aug 23 and the Xarray patch landed Sept 6. So this means the problem existed before the Xarray patch landed.

[Gerrit Updater]

"James Simmons <[e-mail]>" uploaded a new patch: https://review.whamcloud.com/c/fs/lustre-release/+/52995
Subject: LU-17046 tests: increase sleep for sanity-quota 1g
Project: fs/lustre-release
Branch: master
Current Patch Set: 1
Commit: f3e49e5c530451426a0a2505b9cd16ba6662e4f4

[James A Simmons]

Tracked down the commit that is causing this failure.

commit 5176c0494338de34be9b2bc35e55f91daaab67a6
("LU-13810 tests: increase limit for 1g")

Further test suggest this bug existed before this patch but this patch really makes it come out.

[Sergey Cheremencev]

From https://testing.whamcloud.com/test_sets/199b4874-fb26-4062-b506-70ca71268b02

 [dd] [if=/dev/zero] [bs=1M] [of=/mnt/lustre/d1g.sanity-quota/f1g.sanity-quota-0] [count=10] [seek=10]
dd: error writing '/mnt/lustre/d1g.sanity-quota/f1g.sanity-quota-0': Disk quota exceeded
8+0 records in
7+0 records out

Due to preacquire requests it gets EDQUOT earlier and only 7MB instead of 10M. Despite there is no logs about changing edquot from 1 to 0(somehow there is nothing about that and I believe this part of logs missed), I think as was written only 17MB, not 20, OSTs could send release requests causing to change edquot flag from 1 to 0. Thus when it writes the rest 2MB(there are 2 OSTs) it doesn't fail as there is no EDQUOT flag anymore and finally it writes just 19(17+2) MB instead of 22(20+2).

[Gerrit Updater]

"Sergey Cheremencev <[e-mail]>" uploaded a new patch: https://review.whamcloud.com/c/fs/lustre-release/+/53038
Subject: LU-17046 tests: fix write success in 1g
Project: fs/lustre-release
Branch: master
Current Patch Set: 1
Commit: 072e51154c13e7abf035a75018e52dc94c04806c

[Sergey Cheremencev]

There is a lot of failures of sanity-quota_1g last time. However, the most of them have different reasons. https://review.whamcloud.com/c/fs/lustre-release/+/53038 is only to fix the issue from the description. Other failures could be fixed by https://review.whamcloud.com/c/fs/lustre-release/+/52713 (LU-17191). If after landing these both patches sanity-quota_1g will be still failing, new failures should be worked out in a separate ticket.

[Gerrit Updater]

"Oleg Drokin <[e-mail]>" merged in patch https://review.whamcloud.com/c/fs/lustre-release/+/53038/
Subject: LU-17046 tests: fix write success in 1g
Project: fs/lustre-release
Branch: master
Current Patch Set:
Commit: de352465eb6a02aeb20357208c54e903585e12e3

[James A Simmons]

Thank you Oleg.