[LU-10865] quota reintegration reported version and index count errors.

• Type: Bug
• Resolution: Won't Fix
• Priority: Minor
• Labels: patch
• Environment: Lustre master branch; osd-ldiskfs backend.
• Severity: 3
• Assignee: Wang Shilong (Inactive)
• Reporter: Wang Shilong (Inactive)

Description

One of Customers reported following errors when force quota reintegrations:

Mar 29 01:30:20 xssds31 kernel: Lustre: 2007:0:(qsd_reint.c:246:qsd_reint_index()) lustre3-MDT0000: index version for fid [0x200000005:0x1016:0x0] is 0, but index isn't empty (1)

This is easy to reproduce:

    1.format Lustre and mount MDT/MGS, OST.

    2.lctl conf_param lustre.quota.mdt=ug

    3.lctl set_param osd-ldiskfs.*.quota_slave.force_reint=1

 

 Then you will see some dmesg like above.

 Problem is osd-ldiskfs IAM will try to init one

 zero entry initially, so for newly created slave index

 file, there will be at least one entry for it, let's

 bump version properly to make reintegration check happy. 

Issue Links

• is related to: LU-10850 empty indexes don't iterate correctly on ldiskfs

Comments

[Peter Jones]

Thanks wangshilong

[Gerrit Updater]

Wang Shilong ([e-mail]) uploaded a new patch: https://review.whamcloud.com/31834
Subject: LU-10865 osd-ldiskfs: bump quota index version initially
Project: fs/lustre-release
Branch: master
Current Patch Set: 1
Commit: 8b408368b679293b5bcde3854be356f7cc8c8ea9