fix hostnames in kops openstack

[zetaab]

fixes #6441

% openstack server list
+--------------------------------------+-------------------------------+--------+-----------------------------------------+--------+----------+
| ID                                   | Name                          | Status | Networks                                | Image  | Flavor   |
+--------------------------------------+-------------------------------+--------+-----------------------------------------+--------+----------+
| 83df9825-22cd-49c0-8f0e-ff90a7372832 | bastions-1.oma.k8s.local      | ACTIVE | oma.k8s.local=10.1.1.196, 192.168.1.206 | ubuntu | ds1G     |
| a83c3663-a07e-4cca-a8d2-53e4905d1259 | nodes-1.oma.k8s.local         | ACTIVE | oma.k8s.local=10.1.1.156                | ubuntu | m1.small |
| e3845789-1c6c-40ac-bd51-dbfd29df9051 | master-nova-2-1.oma.k8s.local | ACTIVE | oma.k8s.local=10.1.2.218                | ubuntu | m1.small |
| 0ca71cd0-b0fb-41fc-8628-296f3007e5b8 | master-nova-1-1.oma.k8s.local | ACTIVE | oma.k8s.local=10.1.1.12                 | ubuntu | m1.small |
| d4fed7f6-21aa-484a-8e1b-733eadb14060 | master-nova-3-1.oma.k8s.local | ACTIVE | oma.k8s.local=10.1.1.225                | ubuntu | m1.small |
+--------------------------------------+-------------------------------+--------+-----------------------------------------+--------+----------+

Kubernetes: (these will be fixed if openstack cloudprovider will start parsing hostnames correctly)

% kubectl get nodes
NAME                            STATUS     ROLES     AGE       VERSION
master-nova-1-1.oma.k8s.local   Ready      master    48s       v1.11.6
master-nova-2-1.oma.k8s.local   Ready      master    27s       v1.11.6
master-nova-3-1.oma.k8s.local   Ready      master    45s       v1.11.6
nodes-1.oma.k8s.local           NotReady   node      2s        v1.11.6

and hostnames are now in machines like bastions-1, nodes-1, master-nova-2-1 and so on.

[chrisz100]

/lgtm
/assign @mikesplain @justinsb

As discussed on slack that’s a rather high prio pr as it allows external ccm to function as addon.

[zetaab]

/hold

Waiting for kubernetes/cloud-provider-openstack#467

Also I am thinking that could we just replace . with - in hostname.

Then the hostnames in linux should work, and in openstack also. Example: master-nova-1-1.oma.k8s.local -> master-nova-1-1-oma-k8s-local

@chrisz100 what you think? Then that PR to upstream openstack is not needed if we just replace dots

[zetaab]

/hold cancel

[drekle]

Been testing with a variety of helm charts. There are a number which have podAntiAffinity rules which require this PR to work correctly.

[zetaab]

and this PR is must if we want to get external CCM working. (I am working with that in another PR)

[zetaab]

/test pull-kops-e2e-kubernetes-aws

[zetaab]

/sig openstack

[chrisz100]

/lgtm

[zetaab]

@dims could you approve if ok?

[dims]

/approve

[k8s-ci-robot]

[APPROVALNOTIFIER] This PR is APPROVED

This pull-request has been approved by: dims, zetaab

The full list of commands accepted by this bot can be found here.

The pull request process is described here

Needs approval from an approver in each of these files:

• pkg/model/openstackmodel/OWNERS [dims,zetaab]
• pkg/resources/openstack/OWNERS [dims,zetaab]

Approvers can indicate their approval by writing /approve in a comment
Approvers can cancel approval by writing /approve cancel in a comment

[justinsb]

FYI on AWS we use kubernetes.io/cluster/<cluster-name>=shared and kubernetes.io/cluster/<cluster-name>=owned. <cluster-name> can be anything but for kops we use the name of the cluster you create.

We originally had KubernetesCluster=<cluster-name> but this made it hard to have shared resources.

[justinsb]

(and it looks like GCP is using kubernetes.io/cluster-id=<cluster-name>, but only in the description field for GCLBs https://github.com/kubernetes/kubernetes/blob/1c557b9ce866d67ec6088f37058e8594b89606ee/pkg/cloudprovider/providers/gce/gce_loadbalancer_external.go#L862