Uploaded image for project: 'Lustre'
  1. Lustre
  2. LU-7906

osd-zfs should not set blocksize on MDT

Details

    • Improvement
    • Resolution: Fixed
    • Critical
    • Lustre 2.9.0
    • None
    • None
    • 9223372036854775807

    Description

      osd-zfs should not set blocksize on MDT. currently it does as the check is broken.

      Attachments

        Issue Links

          Activity

            [LU-7906] osd-zfs should not set blocksize on MDT

            Landed to master for 2.9.0

            jgmitter Joseph Gmitter (Inactive) added a comment - Landed to master for 2.9.0

            Oleg Drokin (oleg.drokin@intel.com) merged in patch http://review.whamcloud.com/19286/
            Subject: LU-7906 osd: no blocksize on non-OST objects
            Project: fs/lustre-release
            Branch: master
            Current Patch Set:
            Commit: aa84fbc8165f526dae4bd824a48c186c3ac2f639

            gerrit Gerrit Updater added a comment - Oleg Drokin (oleg.drokin@intel.com) merged in patch http://review.whamcloud.com/19286/ Subject: LU-7906 osd: no blocksize on non-OST objects Project: fs/lustre-release Branch: master Current Patch Set: Commit: aa84fbc8165f526dae4bd824a48c186c3ac2f639

            Alex Zhuravlev (alexey.zhuravlev@intel.com) uploaded a new patch: http://review.whamcloud.com/19286
            Subject: LU-7906 osd: no blocksize on non-OST objects
            Project: fs/lustre-release
            Branch: master
            Current Patch Set: 1
            Commit: 12370e8871c19618b20b3af3293b72a19617a9c4

            gerrit Gerrit Updater added a comment - Alex Zhuravlev (alexey.zhuravlev@intel.com) uploaded a new patch: http://review.whamcloud.com/19286 Subject: LU-7906 osd: no blocksize on non-OST objects Project: fs/lustre-release Branch: master Current Patch Set: 1 Commit: 12370e8871c19618b20b3af3293b72a19617a9c4

            in my testing it doesn't improve performance a lot, but except for large dnodes and dnode accounting I haven't found any silver bulltet. it's rather a lot of small improvements closing the gap with ldiskfs.

            bzzz Alex Zhuravlev added a comment - in my testing it doesn't improve performance a lot, but except for large dnodes and dnode accounting I haven't found any silver bulltet. it's rather a lot of small improvements closing the gap with ldiskfs.

            This seems like it could be a major cause of performance loss on the MDT? It would be worthwhile to get a patch and test this sooner rather than later.

            adilger Andreas Dilger added a comment - This seems like it could be a major cause of performance loss on the MDT? It would be worthwhile to get a patch and test this sooner rather than later.

            Assigning to Alex as part of the ZFS metadata performance work

            jgmitter Joseph Gmitter (Inactive) added a comment - Assigning to Alex as part of the ZFS metadata performance work

            People

              utopiabound Nathaniel Clark
              bzzz Alex Zhuravlev
              Votes:
              0 Vote for this issue
              Watchers:
              6 Start watching this issue

              Dates

                Created:
                Updated:
                Resolved: