Hello,
i have a 8 node proxmox cluster located over two datacenters, 4 on each side. We are still running Proxmox 5.4. Currently all network communication (corosync and normal network communication) is done over one switch per datacenter. I know that is not optimal and i could add a second network interface for the corosync communication over an other switch. Sadly this can't be done in time for the change that is coming.
Long story short: one of my colleagues wants to reconfigure the switch ports where my proxmox cluster is connected to and all of the proxmox communications is running over. He said nothing should happen because its a simple change and in the worst case my network interfaces are gone for 30 seconds. What does happen when the proxmox hosts aren't able to talk to each other for 30 seconds? Is there a maintenance mode or something to disable all fencing and split brain scenarios? I'm currently using ha-manager and would disable this before the change. Shutting down all VMs running on the cluster is not an option. Are there any other steps i can do to reduce the impact if something goes wrong?
Thanks & best regrads,
Daniel
i have a 8 node proxmox cluster located over two datacenters, 4 on each side. We are still running Proxmox 5.4. Currently all network communication (corosync and normal network communication) is done over one switch per datacenter. I know that is not optimal and i could add a second network interface for the corosync communication over an other switch. Sadly this can't be done in time for the change that is coming.
Long story short: one of my colleagues wants to reconfigure the switch ports where my proxmox cluster is connected to and all of the proxmox communications is running over. He said nothing should happen because its a simple change and in the worst case my network interfaces are gone for 30 seconds. What does happen when the proxmox hosts aren't able to talk to each other for 30 seconds? Is there a maintenance mode or something to disable all fencing and split brain scenarios? I'm currently using ha-manager and would disable this before the change. Shutting down all VMs running on the cluster is not an option. Are there any other steps i can do to reduce the impact if something goes wrong?
Thanks & best regrads,
Daniel
