Uploaded image for project: 'Lustre'
  1. Lustre
  2. LU-4294

lctl dl does not show inactive state for deactivated OST

Details

    • Bug
    • Resolution: Duplicate
    • Minor
    • None
    • Lustre 2.4.1
    • Lustre 2.4.1 RHEL6 2.6.32-358.18.1.el6_lustre.x86_64
    • 3
    • 11778

    Description

      I deactivate an OST with

      lctl --device 56 deactivate

      56 being the first column number of "lctl dl" output.

      After this "lctl dl" still shows the OST as "UP":

      56 UP osp vegas-OST0059-osc-MDT0000 vegas-MDT0000-mdtlov_UUID 5

      Doing a

      cat /proc/fs/lustre/lov/vegas-MDT0000-mdtlov/target_obd

      however shows the OST set to INACTIVE:

      89: vegas-OST0059_UUID INACTIVE

      Attachments

        Issue Links

          Activity

            [LU-4294] lctl dl does not show inactive state for deactivated OST
            emoly.liu Emoly Liu added a comment -

            Yes, this issue can be marked as a duplicate of LUDOC-218. And in the future if needed, we can file a new ticket to modify OSP code.

            emoly.liu Emoly Liu added a comment - Yes, this issue can be marked as a duplicate of LUDOC-218 . And in the future if needed, we can file a new ticket to modify OSP code.
            pjones Peter Jones added a comment -

            Do I understand correctly that this issue is essentially a duplicate of LUDOC-218?

            pjones Peter Jones added a comment - Do I understand correctly that this issue is essentially a duplicate of LUDOC-218 ?
            emoly.liu Emoly Liu added a comment -

            Yes, LUDOC-218 is created for doc update.

            emoly.liu Emoly Liu added a comment - Yes, LUDOC-218 is created for doc update.

            Since "lctl dl" output has changed in newer Lustre versions this should be reflected in the Lustre 2.x manual. Can you create a documentation bug?

            rlaifer Roland Laifer added a comment - Since "lctl dl" output has changed in newer Lustre versions this should be reflected in the Lustre 2.x manual. Can you create a documentation bug?
            emoly.liu Emoly Liu added a comment -

            I have a check at Lustre 2.x manual and find it does say "lctl dl" reports "IN". But I think the example should be older than b2_4, because since b2_4 OSP has been added to replace OSC connected from MDS.

            The following values are part of "lctl dl" from b2_3 and master branch respectively.

            b2_3:
             10 UP mdc lustre-MDT0000-mdc-ffff88003c798000 aff4df3d-483c-c37a-bed4-19f17dd103ed 5
             11 UP osc lustre-OST0000-osc-ffff88003c798000 aff4df3d-483c-c37a-bed4-19f17dd103ed 5
             12 UP osc lustre-OST0001-osc-ffff88003c798000 aff4df3d-483c-c37a-bed4-19f17dd103ed 5
             13 UP osc lustre-OST0000-osc-MDT0000 lustre-MDT0000-mdtlov_UUID 5
             14 UP osc lustre-OST0001-osc-MDT0000 lustre-MDT0000-mdtlov_UUID 5
            
            master:
             18 UP mdc lustre-MDT0000-mdc-ffff88002379b800 67d79455-bc1e-4856-d567-68656b793b34 5
             19 UP osc lustre-OST0000-osc-ffff88002379b800 67d79455-bc1e-4856-d567-68656b793b34 5
             20 UP osc lustre-OST0001-osc-ffff88002379b800 67d79455-bc1e-4856-d567-68656b793b34 5
             21 UP osp lustre-OST0000-osc-MDT0000 lustre-MDT0000-mdtlov_UUID 5
             22 UP osp lustre-OST0001-osc-MDT0000 lustre-MDT0000-mdtlov_UUID 5
            

            And the code shows that although OSP can receive de/activated setting by "lctl --device", osp_import_event() doesn't call obd_notify() to update obd_inactive status used by "lctl dl". That's why we didn't see "IN" change by deactivating osp device.

            emoly.liu Emoly Liu added a comment - I have a check at Lustre 2.x manual and find it does say "lctl dl" reports "IN". But I think the example should be older than b2_4, because since b2_4 OSP has been added to replace OSC connected from MDS. The following values are part of "lctl dl" from b2_3 and master branch respectively. b2_3: 10 UP mdc lustre-MDT0000-mdc-ffff88003c798000 aff4df3d-483c-c37a-bed4-19f17dd103ed 5 11 UP osc lustre-OST0000-osc-ffff88003c798000 aff4df3d-483c-c37a-bed4-19f17dd103ed 5 12 UP osc lustre-OST0001-osc-ffff88003c798000 aff4df3d-483c-c37a-bed4-19f17dd103ed 5 13 UP osc lustre-OST0000-osc-MDT0000 lustre-MDT0000-mdtlov_UUID 5 14 UP osc lustre-OST0001-osc-MDT0000 lustre-MDT0000-mdtlov_UUID 5 master: 18 UP mdc lustre-MDT0000-mdc-ffff88002379b800 67d79455-bc1e-4856-d567-68656b793b34 5 19 UP osc lustre-OST0000-osc-ffff88002379b800 67d79455-bc1e-4856-d567-68656b793b34 5 20 UP osc lustre-OST0001-osc-ffff88002379b800 67d79455-bc1e-4856-d567-68656b793b34 5 21 UP osp lustre-OST0000-osc-MDT0000 lustre-MDT0000-mdtlov_UUID 5 22 UP osp lustre-OST0001-osc-MDT0000 lustre-MDT0000-mdtlov_UUID 5 And the code shows that although OSP can receive de/activated setting by "lctl --device", osp_import_event() doesn't call obd_notify() to update obd_inactive status used by "lctl dl". That's why we didn't see "IN" change by deactivating osp device.
            pjones Peter Jones added a comment -

            Emoly

            Could you please look into this one?

            Thanks

            Peter

            pjones Peter Jones added a comment - Emoly Could you please look into this one? Thanks Peter

            This seems to have changed between Lustre 1.8 and Lustre 2.x: With Lustre 1.8 "lctl dl" reported "IN".
            Therefore, the question is if this change was intended since it might confuse administrators (like me).
            Also, if the change was intended the Lustre 2.x manual has to be adapted, e.g. chapter 18.8.3 and chapter 19.1.2.

            rlaifer Roland Laifer added a comment - This seems to have changed between Lustre 1.8 and Lustre 2.x: With Lustre 1.8 "lctl dl" reported "IN". Therefore, the question is if this change was intended since it might confuse administrators (like me). Also, if the change was intended the Lustre 2.x manual has to be adapted, e.g. chapter 18.8.3 and chapter 19.1.2.

            "UP" means the device is configured, it is not the same thing as INACTIVE.

            adilger Andreas Dilger added a comment - "UP" means the device is configured, it is not the same thing as INACTIVE.

            People

              emoly.liu Emoly Liu
              wbaudler Wolfgang Baudler
              Votes:
              0 Vote for this issue
              Watchers:
              7 Start watching this issue

              Dates

                Created:
                Updated:
                Resolved: