DRBD9 failover pool / HA

[hjjg]

Hi forum!

Since DRBD9 doesn't support dual primary as in pve.proxmox.com/wiki/DRBD I'd like to ask if there are alternative setups. I never used one LVM volume group read-write on two nodes in parallel. Instead I followed the recommendation and created two volume groups and use the first one on node1 and the second one on node2. The coolest feature with this is that I can reboot node1 after live-migrating all VMs from the first DRBD pool to the second machine.

So - do I need to wait until DRBD9.1 is released with full dual-primary support or is it possible to do this with my current setup?

Thanks and kind regards,

hjjg

 

[dietmar]

Since DRBD9 doesn't support dual primary as in pve.proxmox.com/wiki/DRBD

What problem do you have exactly? We use the new auto-promote feature, so we switch to dual-primary automatically when needed.
So live migration works out of the box.

 

[hjjg]

Maybe I did not understand the new concept of DRBD9. If I understand correctly then the virtual machine images are created as resources so every disk image is replicated. That means that we do not need a big DRBD device that is replicated and dual primary.
I still need to check the documentation. Is split brain still possible?

 

[dietmar]

Is split brain still possible?

should not happen.

 

[tatyrza]

Hello! I don't understand what is really doing new HA-manager? When I reboot node, all vm machines freeze. When node stop, all vms stop. When node reset, vm got error, and never go online "Could not open '/dev/drbd/by-res/vm-104-disk-1/0': Permission denied". I'm testing PVE4 with DRBD9, got subscription.




I am considering a return to the version 3.4 PVE.