Re: Switching from group to project quota's

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On 2016-12-19 06:07 AM, Eric Sandeen wrote:
On 12/18/16 7:03 PM, ncoulson-ml@xxxxxxxxxxxx wrote:
We are currently using user and group quota's on multiple xfs
filesystem.  When trying to switch to Project Quota's, it attempts to
perform a quotacheck.
I assume that you mean you switched mount -o uquota,gquota for
mount -o uquota,pquota or something like that?

We are, (that or alternatively in testing were using xfs_quota -x, off -g).
Tested on a lvm snapshot of a live volume,  it was still not mounted
after about a week (and couldn't find a way to cancel the attempted
mount.  dmesg did mention quotacheck, but I can't say it was doing
the quotacheck for the full week).

Is there a way to convert from Group Quota's to Project Quota's while
avoiding the quotacheck?  (There are currently no project quota's
defined).  Outside of that, I was curious what scenario's would cause
quotacalc to recalculate.
If any quota is on at mount time, and the "checked" flag is not present
in the on-disk superblock, then quotacheck gets run.

In general, we do accounting even if we don't do enforcement; i.e.
having no quotas defined does not man that quota activity isn't
happening - it is, because accounting happens regardless.

Thank You



1TB Volumes, Reviewing one of them, we are using 67G of 934GB (they
vary from 400GB free to 50GB Free).

Inode usage seems to range from 5000000 to 8000000.

Using the 2.6.32 kernel from Centos 6.8, but was doing my tests with
a snapshot of the volume on a centos 7.2 host (kernel 3.10)


I thought I may be able to work around this by removing the group
quota information,  but xfs_quota -x, remove -g with and without
group quota's on failed to work.    Everytime I received XFS_QUOTARM:
Invalid argument
This was likely fixed upstream in 3.16

commit 9da93f9b7cdf8ab28da6b364cdc1fafc8670b4dc
Author: Eric Sandeen <sandeen@xxxxxxxxxxx>
Date:   Mon May 5 17:25:50 2014 +1000

     xfs: fix Q_XQUOTARM ioctl

and should be fixed in sufficiently up to date centos7 as well.

but removing quota inodes won't obviate the need for a quotacheck,
I'm afraid.

-Eric


With the long quotacheck time (+1wk, I think one test I left running for a few weeks) with our filesystem (not sure if that is due to the number of files we have, or if something else is at play), I was hoping to find an alternative. Best idea I have so far is to rsync to a 2nd volume with project quota's enabled (a few days per server most likely), and some symlink magic to keep things appearing in the same location.


In testing, it looks like disabling group quota support will not cause a QuotaCheck.

In the case of mounting a filesystem with pquotas, that previously used gquotas does certainly seem to trigger it.

What I was hoping (falsely), was mounting a volume with pquotas where there is no group quota information would not need a quotacheck (or at least not one that goes on as long as our tests have since there would be no projects setup yet). I noticed it says in dmesg when it does a quotacheck, and the following certainly points to it being triggered.


root@test [/mnt]# mount /dev/sys.vg/test.lv z -o uqnoenforce
XFS (dm-8): Mounting Filesystem
XFS (dm-8): Ending clean mount

root@test [/mnt]# mount /dev/sys.vg/test.lv z -o uqnoenforce,pquota
XFS (dm-8): Mounting Filesystem
XFS (dm-8): Ending clean mount
XFS (dm-8): Quotacheck needed: Please wait.
XFS (dm-8): Quotacheck: Done.

But thank you Eric for your time.


Nathan Coulson

User quota state on /test (/dev/sda) Accounting: ON Enforcement: OFF
Inode: #131 (1663 blocks, 1639 extents) Group quota state on /test
(/dev/sda) Accounting: ON Enforcement: OFF Inode: #132 (1664 blocks,
1648 extents) Project quota state on /test (/dev/sda) Accounting:
OFF Enforcement: OFF Inode: #132 (1664 blocks, 1648 extents) Blocks
grace time: [7 days 00:00:30] Inodes grace time: [7 days 00:00:30]
Realtime Blocks grace time: [7 days 00:00:30]


Nathan Coulson. -- To unsubscribe from this list: send the line
"unsubscribe linux-xfs" in the body of a message to
majordomo@xxxxxxxxxxxxxxx More majordomo info at
http://vger.kernel.org/majordomo-info.html


--
To unsubscribe from this list: send the line "unsubscribe linux-xfs" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html



[Index of Archives]     [XFS Filesystem Development (older mail)]     [Linux Filesystem Development]     [Linux Audio Users]     [Yosemite Trails]     [Linux Kernel]     [Linux RAID]     [Linux SCSI]


  Powered by Linux