[LU-16055] old file version returned on some client

• Status: Open
• Project: Lustre
• Type: Bug
• Priority: Minor
• Reporter: Götz Waschk
• Assignee: WC Triage
• Resolution: Unresolved

Environment

Lustre 2.12.6 on the client, Lustre 2.12.8_6_g5457c37 on the servers
both on Scientific Linux 7.9

• Severity: 3

Description

On some clients (9 out of 242), the wrong version of a file is read, unless direct I/O is used. The file is small and completely stored on the mdt:

[wgs34] /root # lfs getstripe /lustre/fs24/files/test.py 
/lustre/fs24/files/test.py
  lcm_layout_gen:    2
  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:   524288
      lmm_stripe_count:  0
      lmm_stripe_size:   524288
      lmm_pattern:       mdt
      lmm_layout_gen:    0
      lmm_stripe_offset: 0

    lcme_id:             2
    lcme_mirror_id:      0
    lcme_flags:          0
    lcme_extent.e_start: 524288
    lcme_extent.e_end:   EOF
      lmm_stripe_count:  1
      lmm_stripe_size:   524288
      lmm_pattern:       raid0
      lmm_layout_gen:    0
      lmm_stripe_offset: -1

If I copy the file on one of the affected machines, it is the old overwritten version. If I use direct I/O, the file has the correct content:

[pax9-02] /root # dd if=/lustre/fs24/files/test.py of=bla  iflag=direct bs=4096
0+1 records in
0+1 records out
642 bytes (642 B) copied, 0.000977625 s, 657 kB/s

 

There are no Lustre errors on the client or server. sync or lflush had no effect, but umounting and remounting the file system fixed this.

Comments

[Götz Waschk]

The correct version of the file is returned after dropping caches with:

echo 1 > /proc/sys/vm/drop_caches