[LU-3396] Disable quota by default for 2.4 Created: 24/May/13  Updated: 21/Jun/13  Resolved: 21/Jun/13

Status: Closed
Project: Lustre
Component/s: None
Affects Version/s: None
Fix Version/s: None

Type: Bug Priority: Major
Reporter: Andreas Dilger Assignee: Andreas Dilger
Resolution: Won't Fix Votes: 0
Labels: None

Issue Links:
Blocker
is blocking LU-3305 Quotas affect Metadata performance Resolved
Severity: 3
Rank (Obsolete): 8408

 Description   

The change to enable quota by default at format time still appears to have significant metadata performance impact (25% or more), per recent testing in LU-3305.

We are now looking at whether we should disable quota by default for new filesystems, as was the case in the past Lustre releases, and we are wondering what the side effects of this will be?

It is already possible to disable quota via "tune2fs -O ^quota" after format time, and this matches the upgrade case where the quota feature is not already enabled. Is there anything beyond changing mkfs.lustre to not set the quota feature that would need to be done?

The other issue is about enabling quota on such a filesystem. For upgraded filesystems, or new filesystems that do not have quota enabled by default, this would involve "tune2fs -O quota", and then running e2fsck? Is the distributed quota state on the master automatically rebuilt at this time?



 Comments   
Comment by Alex Zhuravlev [ 26/May/13 ]

Andreas, are you suggesting to disable accounting by default in ZFS as well?

Comment by Andreas Dilger [ 27/May/13 ]

No, tune2fs is only for ldiskfs.

Comment by Andreas Dilger [ 21/Jun/13 ]

Since 2.4.0 was shipped without this change, and we are instead pursuing a path to patch the quota code to fix the locking bottlenecks, I'm closing this bug as "Won't Fix".

Generated at Sat Feb 10 01:33:33 UTC 2024 using Jira 9.4.14#940014-sha1:734e6822bbf0d45eff9af51f82432957f73aa32c.