[LU-8985] conf-sanity test_50i: FAIL: wrong != 3 for striped_dir Created: 04/Jan/17  Updated: 16/Apr/17  Resolved: 16/Apr/17

Status: Resolved
Project: Lustre
Component/s: None
Affects Version/s: Lustre 2.10.0
Fix Version/s: Lustre 2.10.0

Type: Bug Priority: Blocker
Reporter: Jian Yu Assignee: WC Triage
Resolution: Cannot Reproduce Votes: 0
Labels: None

Issue Links:
Related
is related to LU-8922 deactive MDT might cause 2 dir stripe... Resolved
Severity: 3
Rank (Obsolete): 9223372036854775807

 Description   

While testing patch https://review.whamcloud.com/24033 on master branch, conf-sanity test 50i failed in review-dne-part-1 test session as follows:

CMD: trevis-8vm1.trevis.hpdd.intel.com /usr/sbin/lctl get_param -n mdc.lustre-MDT0001-mdc-[!M]*.active
Updated after 6s: wanted '0' got '0'
error on LL_IOC_LMV_SETSTRIPE '/mnt/lustre/d50i.conf-sanity/2' (3): No such device
error: mkdir: create stripe dir '/mnt/lustre/d50i.conf-sanity/2' failed
error opening /mnt/lustre/d50i.conf-sanity/striped_dir: Cannot send after transport endpoint shutdown (108)
llapi_semantic_traverse: Failed to open '/mnt/lustre/d50i.conf-sanity/striped_dir': Cannot send after transport endpoint shutdown (108)
error: getdirstripe failed for /mnt/lustre/d50i.conf-sanity/striped_dir.
/usr/lib64/lustre/tests/conf-sanity.sh: line 3751: [: -eq: unary operator expected
 conf-sanity test_50i: @@@@@@ FAIL: wrong  != 3 for striped_dir 

Maloo report: https://testing.hpdd.intel.com/test_sets/072949d2-d0d2-11e6-9b20-5254006e85c2



 Comments   
Comment by Jian Yu [ 04/Jan/17 ]

More failure instances on master branch:
https://testing.hpdd.intel.com/test_sets/1da5b118-d0b5-11e6-bd53-5254006e85c2
https://testing.hpdd.intel.com/test_sets/a5c88f42-d09d-11e6-a836-5254006e85c2

Comment by Bruno Faccini (Inactive) [ 05/Jan/17 ]

+1 for a master/lustre-review at https://testing.hpdd.intel.com/test_sets/e32b4830-d2e5-11e6-95bb-5254006e85c2

Comment by Bruno Faccini (Inactive) [ 05/Jan/17 ]

+1 for a master/lustre-review at https://testing.hpdd.intel.com/test_sets/16e3b688-d2de-11e6-9f62-5254006e85c2

Comment by nasf (Inactive) [ 06/Jan/17 ]

+1 on master:
https://testing.hpdd.intel.com/test_sets/49c31a48-d3ac-11e6-aa78-5254006e85c2

Comment by Jian Yu [ 06/Jan/17 ]

This is blocking patch review testing on master branch.

Comment by Bruno Faccini (Inactive) [ 06/Jan/17 ]

Well, all my master reviews are failing due this patch since I have recently rebased...

Having a look to conf-sanity/test_50i it appears that the precisely failing part has been added by commit 1be03a1088d2aa583c2651435634ba336b48f939 (LU-8922 lod: check master stripes properly) which has landed 3 days ago.

My guess is that the new "$LFS mkdir -i0 -c$MDSCOUNT $DIR/$tdir/striped_dir" in conf-sanity/test_50i fails with "LustreError: 16258:0:(llite_lib.c:2300:ll_prep_inode()) new_inode -fatal: rc -108" on Client side because MDT/DNE-master has prepared/provided striped infos/FIDs for the whole set of MDTs, but then this fails on Client because the mdc connection to MDT0001 has been deactivated for the purpose of the test.

May be DI can help on this ??

Comment by Steve Guminski (Inactive) [ 09/Jan/17 ]

Another on master:
https://testing.hpdd.intel.com/test_sets/d90889a4-d4b0-11e6-85b1-5254006e85c2

Comment by Bruno Faccini (Inactive) [ 17/Jan/17 ]

Looks like the master patch/commit/test for LU-8922, I was pointing to in my previous update, has been recently (> Fri Jan 6 18:29:55 2017) reverted by commit b78fb445555916e380b1661546c821df14098596 from James. So a rebase is required for local master branches to get rid of these auto-tests recurring errors.

Comment by Peter Jones [ 16/Apr/17 ]

AFAICT this issue no longer occurs because the offending patch causing the regression was reverted

Generated at Sat Feb 10 02:22:15 UTC 2024 using Jira 9.4.14#940014-sha1:734e6822bbf0d45eff9af51f82432957f73aa32c.