Node name is not always the same as kubernetes.io/hostname

[mpitt]

Describe the bug
The havener-usage-retriever pods fail to be scheduled because the node selector assumes that the kubernetes.io/hostname label contains the node name, which is not always the case.

To Reproduce
Steps to reproduce the behavior:

1. Run havener top
2. Look at the events (e.g. havener events)
3. There are events like this:

Warning kube-system/havener-usage-retriever-ip-111-222-333-444.xx.compute.internal  FailedScheduling  0/11 nodes are available: 11 node(s) didn't match Pod's node affinity.

Expected behavior
The pod should be correctly scheduled.

Additional context
In my case, the cluster is on EKS and it was created with eksctl.

The kubernetes.io/hostname label contains the hostname (ip-111-222-333-444) while the node name is the FQDN.

[edwardstudy]

Could we set nodeName for this case?

[HeavyWombat]

First of all, thanks for bringing it up. I have to take a look at the options, i.e. fields in the node resource YAML we can use to address it.