Proxmox with DRBD in primary/secondary with manual intervention failover and DR

[asdasdasdasdasfrvcwdcdsq]

Hello,

I have been struggling with a proxmox cluster configuration for about a week now. I understand that Proxmox suggest using DRBD in dual primary mode with fencing for protection but for someone who is NOT looking for automatic redundancy this is an overly dangerous setup. All I want is a cluster that can perform the following:

NodeA fails
Administrator is alerted (via Nagios)
Administrator assesses the situation and determines a failover IS appropriate
Administrator performs a single or selection of few commands to trigger NodeB to spin up VMs that were previously running on NodeA


Now here is the configuration that I currently have in order to try to produce that cluster.
Two identical hardware hosts in PVE Cluster
Each with 1 drive for Proxmox
Each with 1 drive for DRBD to store things

I am open to nearly any solution or configuration that would allow me to do manual recovery of a collection of VMs. I have workable solutions for migrations and recover-from-backup but I just need "node goes down, admin starts vms on other node" functionality.

Thanks

 

[udo]

Hello,

I have been struggling with a proxmox cluster configuration for about a week now. I understand that Proxmox suggest using DRBD in dual primary mode with fencing for protection but for someone who is NOT looking for automatic redundancy this is an overly dangerous setup.

Hi,
why it's dangerous? pve handle the locking for the LVs on the lvm-storage (which are on the drbd-devices). And you can't run primary/seconday because pve need the VG active on all nodes, which are use the storage.

All I want is a cluster that can perform the following:

NodeA fails
Administrator is alerted (via Nagios)
Administrator assesses the situation and determines a failover IS appropriate
Administrator performs a single or selection of few commands to trigger NodeB to spin up VMs that were previously running on NodeA

this is the normal behavior, with an primary/primary drbd-cluster without HA - esp. with only 2 nodes. If one node fail you must do something by hand (because you lost the quorum).

Now here is the configuration that I currently have in order to try to produce that cluster.
Two identical hardware hosts in PVE Cluster
Each with 1 drive for Proxmox
Each with 1 drive for DRBD to store things

I am open to nearly any solution or configuration that would allow me to do manual recovery of a collection of VMs. I have workable solutions for migrations and recover-from-backup but I just need "node goes down, admin starts vms on other node" functionality.

Thanks

The solution is simple - if one pve-host go down (and the VMs are on the drbd-volume):
Enable quorum (with an 3-node cluster not nessesary).
Move the VM-configs ( mv /etc/pve/node-which-is-down/qemu-server/*.conf /etc/pve/node-which-is-up/qemu-server/ )
Start the VMs ( qm start nnn ).

That is the normal behavior without any trouble... but you should use two drbd-ressources ( one for each server).

Udo

 

[asdasdasdasdasfrvcwdcdsq]

Oh god. You just dispersed an enormous cloud of emotional pressure that had built up from smashing my head against this thing and trying to cp VM config files instead of mv them. It makes sense to me know that cp would not work.

copy vs move is literally the lynch pin between our company using Proxmox or not and me keeping my sanity or losing it.

Thank you so much.

Also I do have two drbd resources but omitted on in my original post for simplicity of explanation.