Uploaded image for project: 'Lustre'
  1. Lustre
  2. LU-15740

runtests test_1: 'Space not all freed

Details

    • 3
    • 9223372036854775807

    Description

      This issue was created by maloo for Cliff White <cwhite@whamcloud.com>

      This issue relates to the following test suite run: https://testing.whamcloud.com/test_sets/6363612b-d4a1-4780-9248-a54e1bf4bfea

      test_1 failed with the following error:

      'Space not all freed: now 8932kB, was 8788kB'
      

      May be a repeat of LU-10106, LU-12807, LU-12579

      VVVVVVV DO NOT REMOVE LINES BELOW, Added by Maloo for auto-association VVVVVVV
      runtests test_1 - 'Space not all freed: now 8932kB, was 8788kB'

      Attachments

        Issue Links

          Activity

            [LU-15740] runtests test_1: 'Space not all freed

            "Oleg Drokin <green@whamcloud.com>" merged in patch https://review.whamcloud.com/c/fs/lustre-release/+/51606/
            Subject: LU-15740 tests: scale fs_log_size by OSTCOUNT
            Project: fs/lustre-release
            Branch: b2_15
            Current Patch Set:
            Commit: 34e1409cad412086d509349f64dc9d77911d2fb8

            gerrit Gerrit Updater added a comment - "Oleg Drokin <green@whamcloud.com>" merged in patch https://review.whamcloud.com/c/fs/lustre-release/+/51606/ Subject: LU-15740 tests: scale fs_log_size by OSTCOUNT Project: fs/lustre-release Branch: b2_15 Current Patch Set: Commit: 34e1409cad412086d509349f64dc9d77911d2fb8

            "Andreas Dilger <adilger@whamcloud.com>" uploaded a new patch: https://review.whamcloud.com/c/fs/lustre-release/+/51606
            Subject: LU-15740 tests: scale fs_log_size by OSTCOUNT
            Project: fs/lustre-release
            Branch: b2_15
            Current Patch Set: 1
            Commit: 3d2bcc99c9b3deb296f2659f65daedea710dec10

            gerrit Gerrit Updater added a comment - "Andreas Dilger <adilger@whamcloud.com>" uploaded a new patch: https://review.whamcloud.com/c/fs/lustre-release/+/51606 Subject: LU-15740 tests: scale fs_log_size by OSTCOUNT Project: fs/lustre-release Branch: b2_15 Current Patch Set: 1 Commit: 3d2bcc99c9b3deb296f2659f65daedea710dec10

            "Oleg Drokin <green@whamcloud.com>" merged in patch https://review.whamcloud.com/c/fs/lustre-release/+/50419/
            Subject: LU-15740 tests: scale fs_log_size by OSTCOUNT
            Project: fs/lustre-release
            Branch: master
            Current Patch Set:
            Commit: fabec6f2cb39950a2f208567dac716e21880fa9f

            gerrit Gerrit Updater added a comment - "Oleg Drokin <green@whamcloud.com>" merged in patch https://review.whamcloud.com/c/fs/lustre-release/+/50419/ Subject: LU-15740 tests: scale fs_log_size by OSTCOUNT Project: fs/lustre-release Branch: master Current Patch Set: Commit: fabec6f2cb39950a2f208567dac716e21880fa9f

            In all of the cases where fs_log_size() is checked (runtests included) the actual space compared was based on OST space usage. However, it looks like runtests was failing because fs_log_size() was scaling the returned value based on the MDT count, and not on the OST count. It looks like the main "leakage" of space on the OSTs was only due to blocks being allocated to the object directories O/<seq>/dN, and this was very obvious when runtests was run in isolation and failed repeatedly, but passed if it was run after other tests.

            Change fs_log_size() to scale the "slop" by OSTCOUNT rather than MDTCOUNT.

            adilger Andreas Dilger added a comment - In all of the cases where fs_log_size() is checked ( runtests included) the actual space compared was based on OST space usage. However, it looks like runtests was failing because fs_log_size() was scaling the returned value based on the MDT count, and not on the OST count. It looks like the main "leakage" of space on the OSTs was only due to blocks being allocated to the object directories O/<seq>/dN , and this was very obvious when runtests was run in isolation and failed repeatedly, but passed if it was run after other tests. Change fs_log_size() to scale the "slop" by OSTCOUNT rather than MDTCOUNT .

            "Andreas Dilger <adilger@whamcloud.com>" uploaded a new patch: https://review.whamcloud.com/50419
            Subject: LU-15740 tests: scale fs_log_size by OSTCOUNT
            Project: fs/lustre-release
            Branch: master
            Current Patch Set: 1
            Commit: 56edc654cc035af4e5f60c821892b93655ed5c5e

            adilger Andreas Dilger added a comment - "Andreas Dilger <adilger@whamcloud.com>" uploaded a new patch: https://review.whamcloud.com/50419 Subject: LU-15740 tests: scale fs_log_size by OSTCOUNT Project: fs/lustre-release Branch: master Current Patch Set: 1 Commit: 56edc654cc035af4e5f60c821892b93655ed5c5e

            "Oleg Drokin <green@whamcloud.com>" merged in patch https://review.whamcloud.com/47065/
            Subject: LU-15740 tests: add more stats to runtests
            Project: fs/lustre-release
            Branch: master
            Current Patch Set:
            Commit: 156073a2914145e2b029a658aeec04a54c524e5b

            gerrit Gerrit Updater added a comment - "Oleg Drokin <green@whamcloud.com>" merged in patch https://review.whamcloud.com/47065/ Subject: LU-15740 tests: add more stats to runtests Project: fs/lustre-release Branch: master Current Patch Set: Commit: 156073a2914145e2b029a658aeec04a54c524e5b

            "Andreas Dilger <adilger@whamcloud.com>" uploaded a new patch: https://review.whamcloud.com/47065
            Subject: LU-15740 tests: add more stats to runtests
            Project: fs/lustre-release
            Branch: master
            Current Patch Set: 1
            Commit: a9c290bc420028de274ccfc2c52363bde6208f49

            gerrit Gerrit Updater added a comment - "Andreas Dilger <adilger@whamcloud.com>" uploaded a new patch: https://review.whamcloud.com/47065 Subject: LU-15740 tests: add more stats to runtests Project: fs/lustre-release Branch: master Current Patch Set: 1 Commit: a9c290bc420028de274ccfc2c52363bde6208f49

            It seems unlikely that this is a straight duplicate of LU-12807, since that was dealing with a difference of only 64KB, while this is 144KB difference or more.

            There have been 15 similar failures in the past 4 weeks, but it goes back a long time and it could just be a slow bleed from 64KB to 144KB of space usage.

            adilger Andreas Dilger added a comment - It seems unlikely that this is a straight duplicate of LU-12807 , since that was dealing with a difference of only 64KB, while this is 144KB difference or more. There have been 15 similar failures in the past 4 weeks, but it goes back a long time and it could just be a slow bleed from 64KB to 144KB of space usage.

            People

              adilger Andreas Dilger
              maloo Maloo
              Votes:
              0 Vote for this issue
              Watchers:
              3 Start watching this issue

              Dates

                Created:
                Updated:
                Resolved: