Clustering PVE 2.0 with 2.1 and existing Containers

[webhodnetworks]

Hi all,

currently we are running some Proxmox VE HN´s and want to set up a PVE Cluster in order to migrate containers between physical nodes.
On one HN Proxmox VE 2.0 is running, on the others PVE 2.1.

My questions are:
1. Do you see any problems in clustering PVE 2.0 with PE 2.1? => The HN with PVE 2.0 should be the master server
2. On all HN´s a minimum of 10 OpenVZ containers are running. The VEIDS differ from each other (e.g on the master Server all Containers have VEIDS staring with 200, on the others starting with 300, 400 and 500). Will we get problems when clustering HN´s with existing and running containers?
3. In the tutorial I saw that all HN´s should be in the same network. Is is it possible to cluster physical servers over different class-c networks?

Many thanks for your replies!
Greetings

Chris

 

[Toshik]

Answer for 3rd question - yes, one of options you should setup private network between physical nodes and use it as network for cluster (tutorial: http://www.nedproductions.biz/wiki/...-2.x-cluster-running-over-an-openvpn-intranet)

 

[dietmar]

Will we get problems when clustering HN´s with existing and running containers?

Yes, the tools will reject to add nodes with existing VMs.

 

[Toshik]

Yes, the tools will reject to add nodes with existing VMs.

Well, as an option - he could backup all VMs at one of the PM servers, restore them at other one and then form cluster

 

[webhodnetworks]

Hi all,

many thanks for your replies!

Yes, the tools will reject to add nodes with existing VMs.

Okay, that is poor... So, backing up all VM´s send them to another machine and build the cluster is the only option?

Maybe this is an option for further releases...
Is the only reason why this will be disabled by the tools probably doublicate VMID´s on the HN´s or are there any other system related reasons?

Greetings!
Chris

 

[dietmar]

Is the only reason why this will be disabled by the tools probably doublicate VMID´s on the HN´s or are there any other system related reasons?

A cluster join overwrites the complete configurateion (storage configuration for example).