Problems with migrating an OpenVZ container

[Didelo]

We are experiencing problems with migrating an OpenVZ container from one Proxmox server in a cluster to the other. Both machines (Dell 2850's) are equiped with the same hardware, except for the amount of memory; the machine that we are migrating from has 12 GB of memory, the machine that we are migrating to has 5 GB.

The 5 GB machine has no virtual machines running. We first tried to move the troublesome 4 GB virtual machine while running (even though we didn't attempt to do it with the 'live' option, since we know that doesn't work). This resulted in the following output:

/usr/sbin/vzmigrate 82.98.XXX.XXX 101
OPT:82.98.XXX.XXX
Starting migration of CT 101 to 82.98.XXX.XXX
Preparing remote node
Initializing remote quota
Syncing private
Stopping container
Syncing 2nd level quota
Starting container
Error: Failed to start container
Starting container ...
mount.nfs: /mnt/wcwXXXXXX/data is busy or already mounted
Container is mounted
Setting CPU units: 5000
Setting CPUs: 1
Configure meminfo: 1048576
Set hostname: wcwXXXXXX.XXXXXXX.nl
File resolv.conf was modified
Setting quota ugidlimit: 0
Configure veth devices: veth101.0 veth101.1
Adding interface veth101.0 to bridge vmbr2 on CT0 for CT101
Adding interface veth101.1 to bridge vmbr1 on CT0 for CT101
Container start in progress...
vzquota : (error) Quota is not running for id 101
VM 101 migration failed -

After this, we tried to migrate another 1 GB OpenVZ machine from the 12 GB to the 5 GB machine, which worked without a hitch. Therefore, we decided to stop the first (4 GB) virtual and try to migrate it again. Unfortunately, this resulted in the following output:

/usr/sbin/vzmigrate 82.98.XXX.XXX 101
OPT:82.98.XXX.XXX
Starting migration of CT 101 to 82.98.XXX.XXX
Preparing remote node
Initializing remote quota
Syncing private
Unmounting container
Syncing 2nd level quota
Mounting container
Error: Failed to mount container
mount.nfs "data is busy or already mounted"
Container is mounted
vzquota : (error) Quota is not running for id 101
VM 101 migration failed -

We don't think it is a memory related problem, but we also don't understand how it could be a mounting problem. Does anybody has any ideas?

 

[dietmar]

mount.nfs: /mnt/wcwXXXXXX/data is busy or already

Proxmox does not use nfs (because quotas does not work with nfs), nor does it use directory /mnt/wcwXXXXXX/data.

So please can you explain what you are doing first.

- Dietmar

 

[Didelo]

What we were trying to do was to migrate a server that has an NFS mount configured. In OpenVZ this is done bij adding an <VPS_ID>.mount file in the /etc/vz/conf/ directory for the VPS that needs the NFS mount.

Apparently this file isn't correctly migrated alongside with the config file, which results in the error as described above. The solution we found was to:

1) Move the <VPS_ID>.mount file out of the /etc/vz/conf/ directory on the 'source' Proxmox server. Don't delete it, because you will need it later!
2) On the 'destination' server, create the directory to which the NFS mount is mounted, as defined by the <VPS_ID>.mount file.
3) Stop the VPS on the source server.
4) Migrate the server to the destination server, using the Proxmox GUI.
5) Copy the <VPS_ID>.mount file to the /etc/vz/conf/ directory on the destination server.
6) Start the VPS on the destination server. If all went well, the server should be up and running, using the mount and all.

I hope this helps, just in case anybody else in the future encounters the same problem.