Dear All,

du (and quota) report that files are using double the apparent-size, we suspect this is due to DOM being enabled.

This is a new cluster, running 2.16-1 with 3 x MDT and 7 x OST, using ZFS for both MDT and OST

looking at a couple of files created with dd:
fmb384_gollum> du -sh *
734M    big_file
390G    huge_file

compare to:
fmb384_gollum> du -sh --apparent-size *
319M    big_file
170G    huge_file

DOM is enabled:
root@dino-s1 11:22 [~]: lfs getstripe /scratch/gollum/huge_file
/scratch/gollum/huge_file
  lcm_layout_gen:    3
  lcm_mirror_count:  1
  lcm_entry_count:   2
    lcme_id:             1
    lcme_mirror_id:      0
    lcme_flags:          init
    lcme_extent.e_start: 0
    lcme_extent.e_end:   1048576
      lmm_stripe_count:  0
      lmm_stripe_size:   1048576
      lmm_pattern:       mdt
      lmm_layout_gen:    0
      lmm_stripe_offset: 0

    lcme_id:             2
    lcme_mirror_id:      0
    lcme_flags:          init
    lcme_extent.e_start: 1048576
    lcme_extent.e_end:   EOF
      lmm_stripe_count:  1
      lmm_stripe_size:   1048576
      lmm_pattern:       raid0
      lmm_layout_gen:    0
      lmm_stripe_offset: 3
      lmm_objects:
      -   0: { l_ost_idx:   3, l_fid: [0x380000402:0xa8e72:0x0] }

Whereas folder without DOM:
# lfs setstripe /scratch/jog/noDOM
# dd if=/dev/zero of=noDOM_100M_file bs=1M count=100
# du -h noDOM_100M_file
92M     noDOM_100M_file

compare to:
# du -h --apparent-size noDOM_100M_file
100M    noDOM_100M_file

quota agrees:
root@dino-s1 14:13 [scratch]: lfs quota -h -u jog
Disk quotas for usr jog (uid 1022):
Filesystem used quota limit grace files quota limit grace /scratch 92.05M 0k 0k - 4 0 0

Quota agrees with du, unless we use apparent-size:
fmb384_gollum> du -h /scratch/gollum/*
229M    /scratch/gollum/DOM_100M_file

fmb384_gollum> du -h --apparent-size /scratch/gollum/*
100M    /scratch/gollum/DOM_100M_file

fmb384_gollum> lfs quota -h -u gollum /scratch
Disk quotas for usr gollum (uid 1084):
Filesystem used quota limit grace files quota limit grace /scratch 228.3M 1.5T 2T - 5 0 0 -

Any thoughts gratefully received...

Jake

--
Dr Jake Grimmett
Head Of Scientific Computing
MRC Laboratory of Molecular Biology
Francis Crick Avenue, Cambridge CB2 0QH, UK.
_______________________________________________
lustre-discuss mailing list
lustre-discuss@lists.lustre.org
http://lists.lustre.org/listinfo.cgi/lustre-discuss-lustre.org

Reply via email to