openstack: update the machine templates

[flaper87]

Note that we're passing a version for kubelet, which the openstack
actuator checks on. This version, however, is completely ignored by the
CoreOS image. We'll work on removing this requirement from the openstack
actuator but, in the meantime, we should just set it to something to
avoid breaking.

[tomassedovic]

Thanks for doing this!

I've added a few more things that we need to add to get the actuator to create the workers. These are all things I've had to add in my env before the workers could be created.

[tomassedovic]

We also need all of these (using my own hardcoded values):

      cloudName: standalone
      cloudsSecret: openstack-creds
      networks:
      - uuid: c24f8bc2-4d8b-4ab3-a6ed-1be816113bf6
      placement:
        region: regionOne
      securityGroups:
      - worker
      subnet: worker

And basically the same thing for the master template.

[flaper87]

I knew I was forgetting something. Will update

[flaper87]

From your comment, I'm guessing that using tags didn't work for you. Right?

[tomassedovic]

In what way? If you mean for worker deletion, than no they didn't.

Btw thanks for the update. As far as I can see, it's still missing the networks array and fixing up the subnet name and securityGroups. Will you add those as well please? I've had to fix all three of them in my testing.

[flaper87]

@tomassedovic sorry, I meant for filtering subnets, networks, and securityGroups. The keys exist in the template but I'm assuming that the tag filter didn't work for you.

[tomassedovic]

@flaper87 ah yes, sorry.

You're correct: the filters did not work. The actuator kept complaining that it could not be deserialised properly and that it expected a string (or an array of strings for SG).

As far as I can see, this is the validation it uses:

https://github.com/kubernetes-sigs/cluster-api-provider-openstack/blob/master/config/crd/openstackproviderconfig_v1alpha1.yaml#L19-L28

(or at least something consistent with those expectations)

[flaper87]

@tomassedovic right, that's what I thought. I think this should be fixed in the actuator CRD since we've added support for filters already: https://github.com/kubernetes-sigs/cluster-api-provider-openstack/blob/master/pkg/cloud/openstack/clients/machineservice.go#L255

What do you think about getting this patch in as is and check these filters one by one in the actuator?

[abhinavdahiya]

Please try to use https://github.com/openshift/installer/blob/master/CONTRIBUTING.md#commit-message-format

[tomassedovic]

I'm getting:

I0111 12:13:56.221433       1 controller.go:114] Running reconcile Machine for ostest-worker-8rwch
E0111 12:13:56.225632       1 controller.go:160] Error checking existence of machine instance for machine object ostest-worker-8rwch; error unmarshaling JSON: while decoding JSON: json: cannot unmarshal object into Go struct field OpenstackProviderSpec.networks of type []v1alpha1.NetworkParam

I think this is because networks expects an array of filter objects.

Something like this got my actuator to create the VMs:

networks:
- filter:
    tag: openshiftClusterID=9a6f0a69-3033-4adb-8e7a-91c8f4b75665

That instance is stuck in ERROR, but I think that may be unrelated (altohugh maybe this syntax isn't fully right either).

[flaper87]

oh, ok. So, I failed at defining the filter. Lemme fix that

[flaper87]

actually, did you check it had the right network attached?

[tomassedovic]

It did not. Which may have been the ERROR I was seeing. But my environment was busted for unrelated reasons, so I can't tell for sure if that was it.

I'm running it from scratch again now.

[flaper87]

/test e2e-aws

[tomassedovic]

@flaper87 let me paste the exact diff. Yours is still a little off:

diff --git a/pkg/asset/machines/openstack/worker.go b/pkg/asset/machines/openstack/worker.go
index b73fb3cd6..d6e749387 100644
--- a/pkg/asset/machines/openstack/worker.go
+++ b/pkg/asset/machines/openstack/worker.go
@@ -54,15 +54,10 @@ spec:
           flavor: {{.Machine.FlavorName}}
           placement:
             region: {{.Region}}
-          subnet:
-            filters:
-            - name: "tag:Name"
-              values:
-              - "{{.ClusterName}}-worker-*"
-          tags:
+          networks:
 {{- range $key,$value := .Tags}}
-            - name: "{{$key}}"
-              value: "{{$value}}"
+            - filter:
+                tags: "{{$key}}={{$value}}"
 {{- end}}
           securityGroups:
             - worker

So for each tag we create a filter object in the networks array.

[flaper87]

@tomassedovic I think there was a small race between your comment and my latest push 😄 The current PS should match what you expect to have for the filters field.

[tomassedovic]

/lgtm

[openshift-ci-robot]

[APPROVALNOTIFIER] This PR is APPROVED

This pull-request has been approved by: flaper87, tomassedovic

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:

• OWNERS [flaper87,tomassedovic]

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

[flaper87]

/test e2e-aws