[LU-5806] lustre 2.4.3 quotacheck Created: 24/Oct/14  Updated: 16/Oct/15  Resolved: 16/Oct/15

Status: Resolved
Project: Lustre
Component/s: None
Affects Version/s: Lustre 2.4.3
Fix Version/s: None

Type: Question/Request Priority: Minor
Reporter: Mahmoud Hanafi Assignee: Niu Yawei (Inactive)
Resolution: Won't Fix Votes: 0
Labels: None
Environment:

lustre 2.4.3 servers


Rank (Obsolete): 16287

 Description   

The quota on our OST slaves are not correct and we would like to do a quotacheck. How do we do this without taking down the filesystem.

For example here we see that root has 2.9TB on OST0. We have check the file at the filesystem and the objects on the ost and this value is wrong.
ldan6 ~ # lfs quota -h -u root -v /nobackupp7
Disk quotas for user root (uid 0):
Filesystem used quota limit grace files quota limit grace
/nobackupp7 227.7T 0k 0k - 20842 0 0 -
nbp7-MDT0000_UUID
12.47G - 0k - 20842 - 0 -
nbp7-OST0000_UUID
2.926T - 0k - - - - -
nbp7-OST0001_UUID
2.98T - 0k - - - - -
nbp7-OST0002_UUID



 Comments   
Comment by Peter Jones [ 25/Oct/14 ]

Niu

Can you please advise?

Thanks

Peter

Comment by Niu Yawei (Inactive) [ 27/Oct/14 ]

For example here we see that root has 2.9TB on OST0. We have check the file at the filesystem and the objects on the ost and this value is wrong.

How did you check to find this value is wrong?

Is your system upgraded from 2.1 or it's an fresh 2.4 system? If it's upgraded from 2.1, that's could be dup of LU-3784, and you need to upgrade your e2fsprogs and trigger quota check on the problematic OSTs. I always suggest you upgrade your e2fsprogs to the latest version (1.42.12.wc1-7)

There is no way to do quotacheck online. You need to umount the OST and tirgger quotacheck by "tune2fs -O ^quota $ostdev", then "tune2fs -O quota $ostdev" (disable, then re-enable quota feature on OST device).

Comment by Mahmoud Hanafi [ 27/Oct/14 ]

I think there is a need for a better way to do a force quotacheck without unmounting the OST.

Comment by Niu Yawei (Inactive) [ 28/Oct/14 ]

I think there is a need for a better way to do a force quotacheck without unmounting the OST.

Well, the purpose of making quota feature as an primary feature in ext4/ldiskfs is to eliminate the needs for qutoacheck. I don't think we should implement an online quotacheck, instead, we'd fix the bug which can cause inconsistency.

Comment by Mahmoud Hanafi [ 16/Oct/15 ]

please close this issue

Comment by Peter Jones [ 16/Oct/15 ]

ok Mahmoud

Generated at Sat Feb 10 01:54:41 UTC 2024 using Jira 9.4.14#940014-sha1:734e6822bbf0d45eff9af51f82432957f73aa32c.