Hi All, I am using Terraform Provider for Harvester to bootstrap nodes for Rancher controlplane/workers. Is there a way to setup anti-affinity / scheduling rules for the VM, to make sure that vms are not scheduled on one machine?

If you set labels, e.g.

<http://topology.kubernetes.io/zone=zone1|topology.kubernetes.io/zone=zone1>

on the Harvester nodes, you can define the zone in the VM definition (in the node scheduling section (which is the same as nodeAffinity), e.g.

<http://topology.kubernetes.io/zone|topology.kubernetes.io/zone> In zone1

. The zone info is even populated into the node/guest cluster, so you can use it there to further split your workload.

Yes, but the question is about Terraform provider, which does not expose scheduling options. I mean it does, but it's not released yet

Not sure what you exactly mean. With the current provider you can set zones... which is a scheduling option, even though no nodeselector.

@bumpy-portugal-40754 I am using this provider, the latest version (1.6.0): https://registry.terraform.io/providers/harvester/harvester/latest/docs/resources/virtualmachine The problem is that If I use it, it may happen that all three VMs will e scheduled on one physical machine

Ok. Didn't know that.

The UI options are not present in TF:

Interesting that it's working. Normally there is a version match with Harvester... and Harvester 1.7 is not released yet.

missing this feature as well, any idea when the new provider version will be available ?

Together with 1.7 Harvester? Due date is mid Dec?

I really hope so, also I’ve discovered a bug in the Harvester provider. When node scheduling is enabled via UI, Terraform seems to lose track of the PVC resources, which results in a disk claim failure during deletion.