Cluster moving VM's around

[lucas2000]

Hello All i currently have a proxmox cluster, that im trying to figure out how to best temporarily remove a node without downtime.

a) For example if i want to do maintenance in node A, is there a way to move all of its VMs to node B/C? i know i can power it off and HA will handle it but i rather do it without turning it off since it takes more time for HA to notice.

b) Also i have some VMs that i want to keep on different nodes, is there a way to prevent HA from putting them on the same node?

c) Is there a way to automatically balance my cluster? sometimes when i turn off some nodes to do maintenance or whatever, I endup with one node full of VMs and the other one empty.

 

[grepler]

a) You should be able to use the 'Migrate' button. I believe there is also a Bulk Migrate option available. Generally you will want to setup a replication schedule for each VM is you are using ZFS local storage, or use shared storage like NFS or Ceph to allow faster live migrations. We do it all the time, super useful.

b) you can setup different HA groups with specific priorities if you want to 'pin' particular groups of machines to specific systems, but preserve HA.

There's no built-in method for anti-affinity, like you'd do in Kubernetes.

c) there's no automatic load balancer, but I think I have seen some third-party scrips to it using the APIs. Can't help you here.

 

[lucas2000]

a) You should be able to use the 'Migrate' button. I believe there is also a Bulk Migrate option available. Generally you will want to setup a replication schedule for each VM is you are using ZFS local storage, or use shared storage like NFS or Ceph to allow faster live migrations. We do it all the time, super useful.

b) you can setup different HA groups with specific priorities if you want to 'pin' particular groups of machines to specific systems, but preserve HA.
View attachment 27904
There's no built-in method for anti-affinity, like you'd do in Kubernetes.

c) there's no automatic load balancer, but I think I have seen some third-party scrips to it using the APIs. Can't help you here.

Thanks!!! i will try this now, what is a replication schedule??? (im using ceph)

 

[ph0x]

With Ceph as backing storage you don't need replication. For a planned downtime I would always go for migrating VMs away manually. Otherwise you still have some downtime.