[LU-1911] Test failure on test suite sanity, subtest test_130a Created: 12/Sep/12  Updated: 18/Dec/12  Resolved: 14/Sep/12

Status: Closed
Project: Lustre
Component/s: None
Affects Version/s: Lustre 2.3.0
Fix Version/s: None

Type: Bug Priority: Blocker
Reporter: Maloo Assignee: Johann Lombardi (Inactive)
Resolution: Fixed Votes: 0
Labels: None

Issue Links:
Duplicate
is duplicated by LU-2004 Several sanity 130 test failures Resolved
Related
Severity: 3
Rank (Obsolete): 4231

 Description   

This issue was created by maloo for Johann <johann@whamcloud.com>

This issue relates to the following test suite run: https://maloo.whamcloud.com/test_sets/373e0ae6-fce6-11e1-a961-52540035b04c.

The sub-test test_130a failed with the following error:

FIEMAP on 1-stripe file(/mnt/lustre/f.sanity.130a) failed;

Info required for matching: sanity 130a

hm, it seems that all fiemap tests are failing on master ...
[12/09/12 17:28:11] Johann Lombardi: it seems to be due to an unexpected output format of filefrag
[12/09/12 17:28:32] Johann Lombardi: adilger: did we upgrade e2fsprogs recently?
[12/09/12 17:29:23] Johann Lombardi: Ian is telling me that it is the case
[12/09/12 17:29:50] Johann Lombardi: chris: could we revert to the previous version of e2fsprogs until this is fixed?
[12/09/12 17:30:01] Ian Colle: http://git.whamcloud.com/?p=tools/e2fsprogs.git;a=summary
[12/09/12 17:31:47] Ian Colle: Strange - summary page says it was last changed 5 hours ago, but I don't see anything in the summary from last night
[12/09/12 17:32:43] Johann Lombardi: http://git.whamcloud.com/?p=tools/e2fsprogs.git;a=shortlog;h=refs/heads/master-lustre
[12/09/12 17:32:55] Ian Colle: ahhh
[12/09/12 17:37:24] Johann Lombardi: output format was changed in one of Andreas' patch it seems:
[12/09/12 17:37:35] Johann Lombardi: http://git.whamcloud.com/?p=tools/e2fsprogs.git;a=blobdiff;f=misc/filefrag.c;h=cddb422cd9ee81f330e85ecd810296e7db4b5ac8;hp=56985a1b0f2fe2cd83c808c3562c7197fed5c62d;hb=439cc86d17de637fff93f03580aff7c683f295ac;hpb=ba0318dfc17b037982639c561fe6038881c0266e
[12/09/12 17:38:40] Johann Lombardi: it used to be:
 ext:     device_logical:        physical_offset: length:  dev: flags:
   0:        0..      63:   33691548..  33691611:     64: 0005: network,eof

It is now:
 ext  logical   physical   expected length flags
   0:        0..      63:   33692596..  33692659:     64: 0002: network,eof

Somehow, it seems to confuse our fiemap tests ...
[12/09/12 17:40:35] Johann Lombardi: ah, the tests expect the ':' after column titles :(


 Comments   
Comment by Johann Lombardi (Inactive) [ 12/Sep/12 ]

Tentative patch:
http://review.whamcloud.com/3956

Comment by Andreas Dilger [ 12/Sep/12 ]

This is an error during merge on my part. The lack of change to the header line is a mistake, because there is no "expected" column in the output either. I'll fix this in e2fsprogs again.

However, it is good to be lenient in what is accepted by the test, so I'm not against the patch you submitted. That said, the breakage should also be fixed in e2fsprogs or it will cause problems on every other branch as well.

Comment by Peter Jones [ 12/Sep/12 ]

This is causing a lot of failures. Is is better to deal with it centrally to avoid having to make changes to every branch? Otherwise will we have all kinds of interop failures?

Comment by Oleg Drokin [ 12/Sep/12 ]

I vote for reverting to older e2fsprogs version until newer one is in place. This avoids needs to rebases and will have an immediate positive impact.

Comment by Ian Colle (Inactive) [ 12/Sep/12 ]

I thought Chris had done just that while we worked Johann's patch in parallel.

Comment by Peter Jones [ 12/Sep/12 ]

Well everything still seems to be failing with this error...

Comment by Andreas Dilger [ 13/Sep/12 ]

Chris reverted to the earlier e2fsprogs-1.42.3.wc3 at 12:30 MT, but it may be that any tests already running had the old e2fsprogs installed.

I've since pushed an updated version of e2fsprogs-1.42.5.wc2 that should resolve the problem, but the e2fsprogs builds need to be re-enabled in order to test this out.

Coincidentally, I found a much more serious defect in the filefrag/FIEMAP handling in Lustre, LU-1923, for which I've submitted a separate lustre patch.

Comment by Peter Jones [ 14/Sep/12 ]

This will not occur in the next upgrade to e2fsprogs

Comment by Andreas Dilger [ 15/Sep/12 ]

Unfortunately, there was another instance of filefrag causing intermittent (but relatively frequent) errors, when being run on a file that was just written. A fix has been pushed as e2fsprogs-1.42.5.wc3, and should be seen on new tests that are starting with this version.

Comment by Andreas Dilger [ 15/Sep/12 ]

Link to TT-881, since that will allow proper e2fsprogs + Lustre testing before it affects all test runs.

Generated at Sat Feb 10 01:20:46 UTC 2024 using Jira 9.4.14#940014-sha1:734e6822bbf0d45eff9af51f82432957f73aa32c.