xfs user quota differs from filesystem data

xfs user quota differs from filesystem data

[Marc Mertes]

Hi everybody,
I run into a problem and I don´t know how to solve that.

First my system infos:

xfs_repair version 2.9.8
Kernel: 2.6.26-2-amd64 on debian lenny (5.0.9)
CPU: 2x Quad-Core AMD Opteron(tm) Processor 2378
RAM: 16GB
Hardisk:
Volume /dev/drbd0 /data xfs rw,noatime,attr2,nobarrier,usrquota 0 0
DRBD Version 8.3.7 (api:88/proto:86-91)
Hardware RAID 5 with 6 SAS2 Seagate Cheetah (450GB) disks (5+1xHS) on 
LSI 9260-8i SAS Controller

xfs_info /dev/drbd0
meta-data=/dev/drbd0             isize=256    agcount=4, 
agsize=109744873 blks
          =                       sectsz=512   attr=2
data     =                       bsize=4096   blocks=438979490, imaxpct=5
          =                       sunit=0      swidth=0 blks
naming   =version 2              bsize=4096
log      =internal               bsize=4096   blocks=32768, version=2
          =                       sectsz=512   sunit=0 blks, lazy-count=0
realtime =none                   extsz=4096   blocks=0, rtextents=0


Now my problem:
I defined a userquota (for each, it´s our login server) with a 
soft-/hardlimit of 10/12GB
Now I have a few users where the listet used quota is different from the 
real amount of data in ther folders.

Example1: xfs_quota -x -c "quota -uh bthoma" /data
Disk quotas for User bthoma (343)
Filesystem   Blocks  Quota  Limit Warn/Time    Mounted on
/dev/drbd0    10,1G    12G    15G  00 [------] /data

but:
du -csh /data/user/bthoma
8,7G    /data/user/bthoma
8,7G    insgesamt

Example2:
xfs_quota -x -c "quota -uh lindau" /data
Disk quotas for User lindau (320)
Filesystem   Blocks  Quota  Limit Warn/Time    Mounted on
/dev/drbd0    13,5G    20G    22G  00 [------] /data

but:
du -csh /data/user/lindau
17G     /data/user/lindau
17G     insgesamt

I have no clue how to "refresh" the quota database,
the quick n dirty solution was to set a higher quota for the affected 
users, that they are able to continue with working.
Some reached their quota limit (like in Example one)

at least:
xfs_db frag -v:
actual 3996075, ideal 3118453, fragmentation factor 21.96%

Any Ideas?
Best regards

Marc

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

Re: xfs user quota differs from filesystem data

[Dave Chinner]

On Fri, Jun 29, 2012 at 01:07:45PM +0200, Marc Mertes wrote:
> Hi everybody,
> I run into a problem and I don´t know how to solve that.
> 
> First my system infos:
> 
> xfs_repair version 2.9.8
> Kernel: 2.6.26-2-amd64 on debian lenny (5.0.9)

That's old (both kernel and userspace).

> CPU: 2x Quad-Core AMD Opteron(tm) Processor 2378
> RAM: 16GB
> Hardisk:
> Volume /dev/drbd0 /data xfs rw,noatime,attr2,nobarrier,usrquota 0 0
> DRBD Version 8.3.7 (api:88/proto:86-91)
> Hardware RAID 5 with 6 SAS2 Seagate Cheetah (450GB) disks (5+1xHS)
> on LSI 9260-8i SAS Controller
> 
> xfs_info /dev/drbd0
> meta-data=/dev/drbd0             isize=256    agcount=4,
> agsize=109744873 blks
>          =                       sectsz=512   attr=2
> data     =                       bsize=4096   blocks=438979490, imaxpct=5
>          =                       sunit=0      swidth=0 blks
> naming   =version 2              bsize=4096
> log      =internal               bsize=4096   blocks=32768, version=2
>          =                       sectsz=512   sunit=0 blks, lazy-count=0
> realtime =none                   extsz=4096   blocks=0, rtextents=0
> 
> 
> Now my problem:
> I defined a userquota (for each, it´s our login server) with a
> soft-/hardlimit of 10/12GB
> Now I have a few users where the listet used quota is different from
> the real amount of data in ther folders.
> 
> Example1: xfs_quota -x -c "quota -uh bthoma" /data
> Disk quotas for User bthoma (343)
> Filesystem   Blocks  Quota  Limit Warn/Time    Mounted on
> /dev/drbd0    10,1G    12G    15G  00 [------] /data
> 
> but:
> du -csh /data/user/bthoma
> 8,7G    /data/user/bthoma
> 8,7G    insgesamt

Does the user own files anywhere else on the filesystem? User quota
accounting is not limited to a single sub-tree, but if this is not
the case, then I'm not sure how this can occur.

> Example2:
> xfs_quota -x -c "quota -uh lindau" /data
> Disk quotas for User lindau (320)
> Filesystem   Blocks  Quota  Limit Warn/Time    Mounted on
> /dev/drbd0    13,5G    20G    22G  00 [------] /data
> 
> but:
> du -csh /data/user/lindau
> 17G     /data/user/lindau
> 17G     insgesamt

And that is apparently under-reported. Which implies that there are
delayed allocation reservations - on kernels that old they are not
reported in quota, and depending on how long the quota sync took
before the report was issued and the user behaviour it is possible
to see this temporarily.

How widespread are these issues?

> I have no clue how to "refresh" the quota database,

Quota checks are done at mount time.  You need to unmount, mount
without the quota mount options, then unmount and mount again with
the quota options. Alternatively, you can unmount, run this:

# xfs_db -x -c 'sb 0' -c 'write qflags 0' $device

and then mount again.

> the quick n dirty solution was to set a higher quota for the
> affected users, that they are able to continue with working.
> Some reached their quota limit (like in Example one)

If it is affecting all users, then I'm wondering if, for some
reason, quotas have not been enabled at some time, and when
re-enabled a quotacheck was not executed.....

> at least:
> xfs_db frag -v:
> actual 3996075, ideal 3118453, fragmentation factor 21.96%

Totally irrelevant.

Cheers,

Dave.
-- 
Dave Chinner
david@fromorbit.com

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs