Uploaded image for project: 'Lustre'
  1. Lustre
  2. LU-2587

Quota error after upgrade from 2.1.4 to 2.4

Details

    • Bug
    • Resolution: Fixed
    • Blocker
    • Lustre 2.4.0
    • Lustre 2.4.0
    • before upgrade server and client: 2.1.4
      after upgrade server and client: lustre-master tag-2.3.58 build#1142
    • 3
    • 6030

    Description

      The test steps is:
      1.on lustre-2.1.4, set quota limit for quota_2usr and run IOZONE to use up the limit
      2.stop lustre, upgrade the whole system to 2.4, enable quota on the new system
      3.run IOZONE again on lustre-2.4 with the same user, should expect EDQUOT but the user can still write data exceed quota limit a lot.

      Lustre: DEBUG MARKER: upgrade-downgrade : @@@@@@ FAIL: iozone did not fail with EDQUOT

      Quota info on lustre-2.1.4 filesystem before running IOZONE:

      Listing quotas for user "quota_2usr" and group "quota_2usr" on client-5,client-6...
      client-5: Disk quotas for user quota_2usr (uid 60001):
      client-5:      Filesystem  kbytes   quota   limit   grace   files   quota   limit   grace
      client-5:     /mnt/lustre       0       0 1048576       -       0       0    5120       -
      client-5: lustre-MDT0000_UUID
      client-5:                       0       -  131072       -       0       -    1280       -
      client-5: lustre-OST0000_UUID
      client-5:                       0       -   65536       -       -       -       -       -
      
      client-5: Disk quotas for group quota_2usr (gid 60001):
      client-5:      Filesystem  kbytes   quota   limit   grace   files   quota   limit   grace
      client-5:     /mnt/lustre       0       0 1048576       -       0       0    5120       -
      client-5: lustre-MDT0000_UUID
      client-5:                       0       -  131072       -       0       -    1280       -
      client-5: lustre-OST0000_UUID
      client-5:                       0       -   65536       -       -       -       -       -
      

      Quota info on lustre-2.1.4 filesystem after running IOZONE :

      Listing quotas for user "quota_2usr" and group "quota_2usr" on client-5,client-6...
      client-5: Disk quotas for user quota_2usr (uid 60001):
      client-5:      Filesystem  kbytes   quota   limit   grace   files   quota   limit   grace
      client-5:     /mnt/lustre 1049108*      0 1048576       -       3       0    5120       -
      client-5: lustre-MDT0000_UUID
      client-5:                       4       -    1024       -       3       -     640       -
      client-5: lustre-OST0000_UUID
      client-5:                 1049104*      - 1047552       -       -       -       -       -
      
      client-5: Disk quotas for group quota_2usr (gid 60001):
      client-5:      Filesystem  kbytes   quota   limit   grace   files   quota   limit   grace
      client-5:     /mnt/lustre 1049108*      0 1048576       -       3       0    5120       -
      client-5: lustre-MDT0000_UUID
      client-5:                       4       -    1024       -       3       -     640       -
      client-5: lustre-OST0000_UUID
      client-5:                 1049104*      - 1047552       -       -       -       -       -
      

      Quota info after upgrade from 2.1.4 to 2.4, before running IOZONE:

      Listing quotas for user "quota_2usr" and group "quota_2usr" on client-5,client-6...
      client-5: Disk quotas for user quota_2usr (uid 60001):
      client-5:      Filesystem  kbytes   quota   limit   grace   files   quota   limit   grace
      client-5:     /mnt/lustre 1049108*      0 1048576       -       3       0    5120       -
      client-5: lustre-MDT0000_UUID
      client-5:                       4       -       0       -       3       -       0       -
      client-5: lustre-OST0000_UUID
      client-5:                 1049104       -       0       -       -       -       -       -
      
      client-5: Disk quotas for group quota_2usr (gid 60001):
      client-5:      Filesystem  kbytes   quota   limit   grace   files   quota   limit   grace
      client-5:     /mnt/lustre 1049108*      0 1048576       -       3       0    5120       -
      client-5: lustre-MDT0000_UUID
      client-5:                       4       -       0       -       3*      -       3       -
      client-5: lustre-OST0000_UUID
      client-5:                 1049104       -       0       -       -       -       -       -
      

      Quota info on lustre-2.4 filesystem after running IOZONE:

      Listing quotas for user "quota_2usr" and group "quota_2usr" on client-5,client-6...
      client-5: Disk quotas for user quota_2usr (uid 60001):
      client-5:      Filesystem  kbytes   quota   limit   grace   files   quota   limit   grace
      client-5:     /mnt/lustre 3146264*      0 1048576       -       0       0    5120       -
      client-5: lustre-MDT0000_UUID
      client-5:                       0       -       0       -       0       -    1024       -
      client-5: lustre-OST0000_UUID
      client-5:                 3146264       -       0       -       -       -       -       -
      

      Attachments

        Activity

          [LU-2587] Quota error after upgrade from 2.1.4 to 2.4
          sarah Sarah Liu added a comment -

          verified, close.

          sarah Sarah Liu added a comment - verified, close.

          landed for 2.4

          niu Niu Yawei (Inactive) added a comment - landed for 2.4
          sarah Sarah Liu added a comment -

          Sarah, did you enable quota on the MGS via lctl conf_param? If not, enforcement won't be enabled and this behavior is expected.

          I did lctl conf_param lustre.quota.mdt/ost=ug

          sarah Sarah Liu added a comment - Sarah, did you enable quota on the MGS via lctl conf_param? If not, enforcement won't be enabled and this behavior is expected. I did lctl conf_param lustre.quota.mdt/ost=ug
          niu Niu Yawei (Inactive) added a comment - http://review.whamcloud.com/4969

          Seems there is a defect: after the global index & global index copy were created, they both set the initial version as 1, so even if the global index migration happened, slave will not reintegrate the global index...

          niu Niu Yawei (Inactive) added a comment - Seems there is a defect: after the global index & global index copy were created, they both set the initial version as 1, so even if the global index migration happened, slave will not reintegrate the global index...

          Sarah, did you enable quota on the MGS via lctl conf_param? If not, enforcement won't be enabled and this behavior is expected.

          johann Johann Lombardi (Inactive) added a comment - Sarah, did you enable quota on the MGS via lctl conf_param? If not, enforcement won't be enabled and this behavior is expected.

          People

            niu Niu Yawei (Inactive)
            sarah Sarah Liu
            Votes:
            0 Vote for this issue
            Watchers:
            4 Start watching this issue

            Dates

              Created:
              Updated:
              Resolved: