[LU-11490] run_rr_alloc() defect: FSNAME should be used instead of lustre

• Type: Bug
• Resolution: Fixed
• Priority: Minor
• Fix Version/s: Lustre 2.14.0
• Affects Version/s: Lustre 2.12.0
• Labels: esay
• Severity: 3

Epic

test

• Assignee: WC Triage
• Reporter: Elena Gryaznova

Description

Patch fixes rr_alloc() test to use FSNAME instead of lustre in the parameter path.

Comments

[Peter Jones]

Landed for 2.14

[Gerrit Updater]

Oleg Drokin ([e-mail]) merged in patch https://review.whamcloud.com/36694/
Subject: LU-11490 tests: use variable for file system name
Project: fs/lustre-release
Branch: master
Current Patch Set:
Commit: 134f3eced1535bb6e867d552075d507eadb7cc48

[Gerrit Updater]

James Nunez ([e-mail]) uploaded a new patch: https://review.whamcloud.com/36694
Subject: LU-11490 tests: use variable for file system name
Project: fs/lustre-release
Branch: master
Current Patch Set: 1
Commit: ab64ee277197a093787b914421dbebaf70442c1a

[Peter Jones]

Elena's patch has landed for 2.12 but we should leave the ticket open until the other places have been fixed too

[Gerrit Updater]

Oleg Drokin ([e-mail]) merged in patch https://review.whamcloud.com/33333/
Subject: LU-11490 tests: fix rr_alloc() test to use FSNAME
Project: fs/lustre-release
Branch: master
Current Patch Set:
Commit: c0a1ef281a45fb0baf1f12a4282ea717efe9051d

[James Nunez (Inactive)]

I think there's a few more instances of assuming 'lustre' is the fs name in the Lustre test suites:

sanity test 805

19427         local old
19428         local quota
19429         local pref="osd-zfs.lustre-MDT0000."
19430 
19431         # limit available space on MDS dataset to meet nospace issue

and sanityn test 93

4237         local old_rr=$(do_facet $SINGLEMDS lctl get_param -n \
4238                 'lod.lustre-MDT*/qos_threshold_rr' | sed -e 's/%//')
4239         do_facet $SINGLEMDS lctl set_param -n \
4240                 'lod.lustre-MDT*/qos_threshold_rr' 100
4241         #define OBD_FAIL_MDS_LOV_CREATE_RACE     0x163
4242         do_facet $SINGLEMDS "lctl set_param fail_loc=0x00000163"
4243 
4244         $SETSTRIPE -c -1 $DIR1/$tfile-1/file1 &
4245         local PID1=$!
4246         sleep 1
4247         $SETSTRIPE -c -1 $DIR2/$tfile-2/file2 &
4248         local PID2=$!
4249         wait $PID2
4250         wait $PID1
4251         do_facet $SINGLEMDS "lctl set_param fail_loc=0x0"
4252         do_facet $SINGLEMDS "lctl set_param -n \
4253                 'lod.lustre-MDT*/qos_threshold_rr' $old_rr"

and sanity-flr test 33

1040         # read file again with ost2 failed
1041         $LCTL set_param ldlm.namespaces.lustre-*-osc-ffff*.lru_size=clear

[Andreas Dilger]

Thanks for this patch. Are there other places in the test code that assume "lustre" is the fsname?

I've been thinking that it would be useful to have a random FSNAME setting per test session so that we avoid tests being added that assume the wrong FSNAME, as well as avoid potential conflicts between running or stale filesystem components. In the past we've seen some issues during testing because of an old OSS or MDS VM with the same FSNAME is trying to join a newly-formatted filesystem, which breaks the currently-running testing in strange and hard-to-debug ways.

[Gerrit Updater]

Elena Gryaznova ([e-mail]) uploaded a new patch: https://review.whamcloud.com/33333
Subject: LU-11490 tests: fix rr_alloc() test to use FSNAME
Project: fs/lustre-release
Branch: master
Current Patch Set: 1
Commit: 274f00cfe171a2215dd71a2dbab38fee8bd4fe4d