Pr v1alpha2 kubeadm bootstrapper

[sbueringer]

Migrated to kubeadm bootstrapper as described here: #380 (comment)

Also rewrote the examples (which are probably still far from perfect) and moved them to the examples folder.

Only the operating systems supported by kubeadm bootstrapper are supported now (currently has a dependency to cloud-init >= 18.4 (see kubernetes-retired/cluster-api-bootstrap-provider-kubeadm#41). Ignition is not supported yet.

I tested it with the Ubuntu images from https://github.com/kubernetes-sigs/image-builder/tree/master/images/capi. I tested a 2 control plane node 1 worker node setup with LB and a 1 control plane node 1 worker node without LB scenario.

The user-data must be adjusted if kubelet/kubeadm is not already installed (can be added by further PRs, I justed tested it with an image with pre-installed kubelet/kubeadm)

Note: Examples generation might be a little bit slow because kustomize seems to download the whole cluster-api folder and the vendor folder there hasn't been removed yet.

[jichenjc]

@sbueringer I will try this but not sure how long will it take due to some other works in my daily job :) @hidekazuna if you have time , can you help test as well ? :)

[sbueringer]

All rebased. I'll do some work later today and compare the examples with my own test files

[sbueringer]

All rebased. I'll do some work later today and compare the examples with my own test files

Done. I diffed it against my own test yamls at it's very similar. I think the biggest problem will be that you're maybe not using a OS image which already contains kubelet/kubeadm/docker (or containerd)

I think there are two options:

• Adjust the os image
• Add an install script to the KubeadmConfig and add a kubeadmPreCommand to execute it

I also changed the examples to deploy a cloud-config into the cluster namespace (which has the same name as the cluster). So we're now flexible to deploy multiple clusters. The controller doesn't have a mounted cloud-config anymore.

P.S. if there are some problems with ca cert or authentication to OpenStack, take a look at the changes in provider.go (should still work, but I couldn't test it for all scenarios)

NAME                                               PROVIDERID                                           PHASE
machine.cluster.x-k8s.io/os1pi020-controlplane-0   openstack:////4e72b47b-8d67-4d8a-a1c1-671961e461f4   running
machine.cluster.x-k8s.io/os1pi020-controlplane-1   openstack:////b3b3140b-e97c-46d8-8a3e-f28253ae9d3a   running
machine.cluster.x-k8s.io/os1pi020-node-01          openstack:////742b9f21-07bc-4dfd-bf74-f16ba5e23402   running

NAME                                                                       AGE
openstackmachine.infrastructure.cluster.x-k8s.io/os1pi020-controlplane-0   11m
openstackmachine.infrastructure.cluster.x-k8s.io/os1pi020-controlplane-1   8m56s
openstackmachine.infrastructure.cluster.x-k8s.io/os1pi020-node-01          8m12s

NAME                                PHASE
cluster.cluster.x-k8s.io/os1pi020   provisioned

NAME                                                        AGE
openstackcluster.infrastructure.cluster.x-k8s.io/os1pi020   33m
+--------------------------------------+-------------------------+--------+----------------------------------------------------+--------------------------+--------+
| ID                                   | Name                    | Status | Networks                                           | Image                    | Flavor |
+--------------------------------------+-------------------------+--------+----------------------------------------------------+--------------------------+--------+
| 742b9f21-07bc-4dfd-bf74-f16ba5e23402 | os1pi020-node-01        | ACTIVE | k8s-clusterapi-cluster-os1pi020-os1pi020=10.6.0.10 | ubuntu-1804-kube-v1.15.0 | large  |
| b3b3140b-e97c-46d8-8a3e-f28253ae9d3a | os1pi020-controlplane-1 | ACTIVE | k8s-clusterapi-cluster-os1pi020-os1pi020=10.6.0.16 | ubuntu-1804-kube-v1.15.0 | large  |
| 4e72b47b-8d67-4d8a-a1c1-671961e461f4 | os1pi020-controlplane-0 | ACTIVE | k8s-clusterapi-cluster-os1pi020-os1pi020=10.6.0.5  | ubuntu-1804-kube-v1.15.0 | large  |
+--------------------------------------+-------------------------+--------+----------------------------------------------------+--------------------------+--------+
NAME                      STATUS     ROLES    AGE     VERSION
os1pi020-controlplane-0   NotReady   master   9m22s   v1.15.0
os1pi020-controlplane-1   NotReady   master   49s     v1.15.0
os1pi020-node-01          NotReady   <none>   5m49s   v1.15.0

[sbueringer]

I updated the doc for the usage of clusterctl. I'm not aware if there are released images for the cluster-api-controller & cluster-api-bootstrap-provider-kubeadm, so the doc doesn't work out of the box. If somebody wants to find out if these images are already their, please do :)

For now I will continue using Tilt for development. At the latest when the other providers are starting to use released images for those two controller we can also start using them

Edit: More info about the current status here: https://kubernetes.slack.com/archives/C8TSNPY4T/p1566928917004400

[jichenjc]

as we are moving forward on this, let's quick merge and make changes later..

/lgtm
/approve

[k8s-ci-robot]

[APPROVALNOTIFIER] This PR is APPROVED

This pull-request has been approved by: jichenjc, sbueringer

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 [jichenjc]

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

[sbueringer]

as we are moving forward on this, let's quick merge and make changes later..

/lgtm
/approve

Thx. I'll update the v1alpha2 issue later.

I would also go over our existing issues & PRs and check if they are still valid.

I also have a few suggestions for followup work