two nodes without clustering ( only using DRBD replication)

[djahida]

Hi guys,


I really need some help please

I have some trouble about my configuration of cluster Proxmox 4.3 (3 nodes: two physical machines and the third one is a virtual machine, used as quorum) and RDBD 9. At several times, the nodes are rebooted at the same time, probably because of multicast. But I am sure that it is enabled on the switch!!!!

So, I'm looking for other solution, maybe I can use only the two physical machines separately without configuring cluster, and just keep the DRBD configuration.

So I lost the automatic migration to the other node (HA) because there is no cluster, but I keep the replication with DRBD.


I can use only the first node with all VM, and use the second node as backup. So all VM are configured in the two nodes, but only started at the first one.


If I have troubles on the first node, I can just start all VM on the second one.


Can you tell me if this solution is correct or if there are some negative points to use it on a production environment?


Thanks a lot

 

[djahida]

Is there someone, who can tell me if that solution is correct ?
thanks

 

[remark]

Instead of separate nodes, you can use two-node cluster configuration.
I know, that it is not recommended, but it works.

 

[remark]

Also,as I remember, there were issues with several switches.
Please upgrade your switch firmware.
I have had some troubles too, remember, with VLAN configuration and multicast. As a result, I have had to move cluster network to dedicated network interfaces on each node and connect it to separate dumb switch, and this worked.
Please search the forum for related topics and, maybe, try to use separate unmanaged switch for cluster network.

 

[djahida]

Remark, thanks for the replay

I am in product environment and I don’t think that the two-node cluster configuration can be best solution (I use it with the pve 3 and I have some troubles)

About the switch, I think that the multicast configuration is correct but I don’t know about the firmware of my witch. So I will see to upgrade it.

But what do you think about using separate node without cluster and only with DRBD. Do you think that can be correct?

 

[remark]

The best solution for you is add third hardware node, not VM node.
I think configuration you want to use can be correct, but you lose all cluster features, i. e. centralized nodes control, migration, etc.