[LU-7723] few tests in sanity.sh fail because of STRIPES_PER_OBJ=-1

• Status: Open
• Project: Lustre
• Affects Version/s: Lustre 2.8.0
• Type: Bug
• Priority: Trivial
• Reporter: Alex Zhuravlev
• Assignee: WC Triage
• Resolution: Unresolved
• Labels: easy
• Severity: 3

Description

sanity.sh sets STRIPES_PER_OBJ=-11 unconditionally while few subtests need default striping count = 1. autotest seem to work this around because the filesystem is formatted by external means (not by sanity.sh).

Issue Links

• is related to: LU-9324 sanity-pfl test 10 needs to reset the...
• is related to: LU-11918 Allow setting default file layout on ...

Comments

[Gerrit Updater]

Alex Zhuravlev ([e-mail]) uploaded a new patch: http://review.whamcloud.com/18216
Subject: LU-7723 tests: sanity.sh should set STRIPES_PER_OBJ=1
Project: fs/lustre-release
Branch: master
Current Patch Set: 1
Commit: 4f78fb1d9016aa4088977f27c209a49ca60a6c67

[Andreas Dilger]

It would be good to know which tests are failing, since IMHO it would be better to fix those tests to work with different stripe counts, rather than only ever testing with default stripe count = 1.

[Alex Zhuravlev]

the tests I'm getting failed constantly with "sh sanity.sh":

sanity: FAIL: test_27D llapi_layout_test failed
sanity: FAIL: test_56a /work/head1/lustre/utils/lfs getstripe --obd wrong: found 6, expected 3
sanity: FAIL: test_56u "lfs find -stripe-index 0 -type f /mnt/lustre/d56u.sanityu" wrong: found 16, expected 12
sanity: FAIL: test_65a lverify failed

with STRIPES_PER_OBJ=1 all they pass. I've got a fix for llapi_layout_test ( http://review.whamcloud.com/#/c/18196/ ),
but at some point got to think it's just easier to mimic autotest instead as it's not very convenient to have unrelated failures doing local development.
in general, I agree that it makes sense to fix the tests.

[Andreas Dilger]

I suspect all of these issues will also be hit with a default PFL layout, which is being tested/fixed under LU-9324.

IMHO, it is better to fix the individual tests to work with different stripe counts and (less optimal) explicitly specify the required layout for the test, instead of enforcing a global stripe count to make them pass.