Problem with migration and/or Openvz config with disk_quota

[jleg]

Hi,

we just moved a 2-node proxmox 1.9 to a 3-node proxmox 2.1 cluster - and things look really pretty. Currently, we only see an oddity which we do not really understand: migration of containers do not work - unless the container's config manually is modified with the line
DISK_QUOTA=no

Otherwise, migration alwas tries to call "vzdqdump", which fails because of "Can't open quota file". OpenVZ file system is not extn, but xfs; i guess vzquota still does not support xfs.
However, DISK_QUOTA is set to "no" in /etc/vz/vz.conf - which does not seem to help.

In OpenVZMigrate.pm i can see that vzdqdump should only be called when "DISK_QUOTA" is != 0:

if (!defined($disk_quota) || ($disk_quota != 0)) {
...
        $cmd = "vzdqdump $vmid -U -G -T > " . PVE::Tools::shellquote($self->{quotadumpfile});

But this thing seems to read the container's config - and there is no "DISK_QUOTA" parameter at all, if not added manually.
Of course, GUI says "user quota disabled", seems to relate to "QUOTAUGIDLIMIT"...

Are we doing something wrong, or is this a little flaw in the migration's code?

 

[dietmar]

Are we doing something wrong, or is this a little flaw in the migration's code?

yes, we do not test xfs because quota does not work.

 

[jleg]

yes, we do not test xfs because quota does not work.

i understand; however, looks more like OpenVZMigrate.pm operating on a non-existing parameter, which imo is "non-clean" anyway ;-)

so question would be - is there a way to get "DISK_QUOTA=no" into $VEID.conf automatically besides "hard editing" API2/OpenVZ.pm?

 

[dietmar]

is there a way to get "DISK_QUOTA=no" into $VEID.conf

Simply add it to the VEID.conf file?

 

[jleg]

Simply add it to the VEID.conf file?

well, key word was 'automatically'. Having to do this manually would disqualify for productive usage in a team of admins (for us), which then would have to follow a 'soft policy' to manually enter cli to add that line...

Hm, perhaps some cronjob could do/check this, let's see...

 

[jleg]

i found a workaround for this - since we're using vps.mount/umount scripts anyway to activate xfs project quota, those scripts now also simply check for "DISK_QUOTA" being "no" in $VEID.conf. Looks good so far.

Btw, there seems to be a little problem with proxmox creating fresh containers: after the creation, the container's config has these lines:

VE_ROOT="/var/lib/vz/root/$VEID"
VE_PRIVATE="/var/lib/vz/private/107"

looks like $VEID is not expanded?

 

[dietmar]

looks like $VEID is not expanded?

see 'man vzctl':

      --root path
           Sets the path to root directory (VE_ROOT) for  this  container.   This  is
           essentially  a  mount  point for container's root directory.  Argument can
           contain literal string $VEID, which will be substituted with  the  numeric
           CT ID.

But I guess we should expand all, or nothing.