[lustre-discuss] quota problem on lustre 2.5.3 when doing lfs_migrate

[Philippe Weill]

Le 16/04/2015 08:41, Dilger, Andreas a écrit :
> On 2015/04/15, 11:38 PM, "Philippe Weill" <Philippe.Weill at latmos.ipsl.fr>
> wrote:
>> we have problem with quota on a 2.5.3 filesystem (ext4 based on
>> scientific linux 6 )when using lfs_migrate after desactivating an ost on
>> mds file seem to be well migrated but allocated space stay in quota on the
>> desactivated device
>
> The deactivated OST isn't able to update the quota.  If you reactivate the
> OST and then restart the MDS, the objects on the OST will be destroyed and
> the quota will be updated.
>
> Cheers, Andreas
>

Thanks Andreas
OK so we have just  unmount the mdt and remount after it seem to be ok
for the size quota but not for the inode quota

for sample with another account with  14288 files migrated

after restarting the mds quota size is accurate

Disk quotas for user oneuser (uid 901):
      Filesystem  kbytes   quota   limit   grace   files   quota   limit 
   grace
           /data 118429360  838860800 1048576000       -   29941  300000 
  310000       -
datafs-MDT0000_UUID
                    3768       -       0       -   29941       -   65538 
       -
datafs-OST0000_UUID
                       0       - 16075692       -       -       - 
-       -
datafs-OST0001_UUID
                       0       - 16023952       -       -       - 
-       -
datafs-OST0002_UUID
                       0       - 16020528       -       -       - 
-       -
datafs-OST0003_UUID
                       0       - 15960772       -       -       - 
-       -
datafs-OST0004_UUID
                       0       - 15933760       -       -       - 
-       -
datafs-OST0005_UUID
                       0       - 15769424       -       -       - 
-       -
datafs-OST0006_UUID
                 16997428       - 27068640       -       -       - 
  -       -
datafs-OST0007_UUID
                 29532356       - 43761424       -       -       - 
  -       -
datafs-OST0008_UUID
                 20113180       - 31903688       -       -       - 
  -       -
datafs-OST0009_UUID
                 51782628       - 67237844       -       -       - 
  -       -
[root ~]# chown -R admtest /data/oneuser

after giving all files to another user


[root at ciclad-ng ~]# lfs quota -uv oneuser
Disk quotas for user oneuser (uid 901):
      Filesystem  kbytes   quota   limit   grace   files   quota   limit 
   grace
           /data       0  838860800 1048576000       -   14288  300000 
310000       -
datafs-MDT0000_UUID
                       0       -       0       -   14288       -   65538 
       -
datafs-OST0000_UUID
                       0       - 16075692       -       -       - 
-       -
datafs-OST0001_UUID
                       0       - 16023952       -       -       - 
-       -
datafs-OST0002_UUID
                       0       - 16020528       -       -       - 
-       -
datafs-OST0003_UUID
                       0       - 15960772       -       -       - 
-       -
datafs-OST0004_UUID
                       0       - 15933760       -       -       - 
-       -
datafs-OST0005_UUID
                       0       - 15769424       -       -       - 
-       -
datafs-OST0006_UUID
                       0       - 16143188       -       -       - 
-       -
datafs-OST0007_UUID
                       0       - 16179636       -       -       - 
-       -
datafs-OST0008_UUID
                       0       - 16480220       -       -       - 
-       -
datafs-OST0009_UUID
                       0       - 16002064       -       -       - 
-       -

I'm still having 14288 files but no block on ost

server with rpm for rhel6 on sl6
[root at mds2-ipsl ~]# rpm -qa|egrep 'lustre|e2fs'
lustre-modules-2.5.3-2.6.32_431.23.3.el6_lustre.x86_64.x86_64
e2fsprogs-1.42.12.wc1-7.el6.x86_64
lustre-osd-ldiskfs-2.5.3-2.6.32_431.23.3.el6_lustre.x86_64.x86_64
lustre-iokit-2.5.3-2.6.32_431.23.3.el6_lustre.x86_64.x86_64
kernel-2.6.32-431.23.3.el6_lustre.x86_64
e2fsprogs-libs-1.42.12.wc1-7.el6.x86_64
lustre-2.5.3-2.6.32_431.23.3.el6_lustre.x86_64.x86_64


lfs migrate has been executed on a client  sl6 with rpm from site

[root at nfs-lustre ~]# rpm -qa|grep lustre
lustre-modules-2.5.3-2.6.32_431.23.3.el6_lustre.x86_64.x86_64
lustre-client-2.5.3-2.6.32_431.23.3.el6.x86_64.x86_64
kernel-2.6.32-431.23.3.el6_lustre.x86_64
lustre-client-modules-2.5.3-2.6.32_431.23.3.el6.x86_64.x86_64


cluster client nodes are 1.8.9wc1 because we still have some filesystems 
still in 1.8.9wc1
-- 
Weill Philippe -  Administrateur Systeme et Reseaux
CNRS/UPMC/IPSL   LATMOS (UMR 8190)