Backup serialization and rotation

[thheo]

I run this 2-node HA cluster and I have set up a backup job for all the VMs across the 2 nodes ( I run 4 VMs on one-node and 4 on the other in the normal state of the cluster ).
The backup is done over the same NFS share mounted on both nodes ( attached via the gui ).
What I have seen is the moment the job starts both nodes start vzdump so the NFS is accessed simultaneously which is something I would like to avoid
since the backend is just a single sata drive capable of doing about 100M/s linear, but in this case the performance drops significantly because of constant seeking.
What I would like to happen is to get one VM at a time per cluster, not per node. Is this possible?
Then a second question would be how this job handles rotation of backups? I have checked for 2 backups when I configured the NFS storage in gui. Is this done
automatically?
And a third question: If all VMs get on a single node ( because of HA or live migration done manually ) or simply sometime later I'll rearrange them to be 6 on a node and 2 on a node,
will the backup job perform complete backup? or does it expect to get the VMs like it was when I originally set up the job?

 

[dietmar]

What I would like to happen is to get one VM at a time per cluster, not per node. Is this possible?

no, but you can define 2 different backup jobs.

Then a second question would be how this job handles rotation of backups? I have checked for 2 backups when I configured the NFS storage in gui. Is this done
automatically?

yes

And a third question: If all VMs get on a single node ( because of HA or live migration done manually ) or simply sometime later I'll rearrange them to be 6 on a node and 2 on a node,
will the backup job perform complete backup? or does it expect to get the VMs like it was when I originally set up the job?

if you do not pin the backup job to a specific node, the job is started on all nodes.

 

[wahmed]

On your NFS server, if you put another NIC with separate IP address and if possiblesetup a second pool on a 2nd hdd for backup, then you can easily do simultaneously backup without losing performance. This way 2 nodes will not share any bandwidth and both backup will have same speed. Not sure what kind of setup you have for NFS server, simple Ubuntu with NFS server installed, Freenas or anything else. But this technique should work in all situation depending on the hardware you have. I have a 3rd backup server on a Pentium 4 (14 years old pc) with Ubuntu, NFS server, 2 NICs setup this way.

 

[thheo]

Thank you Dietmar for your reply, splitting the job in 2 makes sense in my case and it seems the script will do exactly what I needed.
Symmcom, you are right that's something I could do, however I only have a server available with 6 hdd slots and they are all occupied for this backup purpose
so I cannot get additional throughput or I/O performance. 1 Gigabit NIC it's enough for me at this stage.