Charmed Kubernetes on OpenStack

Charmed Kubernetes will run seamlessly on OpenStack. With the addition of the openstack-integrator, your cluster will also be able to directly use OpenStack native features.

OpenStack integrator

The openstack-integrator charm simplifies working with Charmed Kubernetes on OpenStack. Using the credentials provided to Juju, it acts as a proxy between Charmed Kubernetes and the underlying cloud, granting permissions to dynamically create, for example, Cinder volumes.

Prerequisites

OpenStack integration requires Octavia to be available in the underlying OpenStack cloud, both to support Kubernetes LoadBalancer services and to support the creation of a load balancer for the Kubernetes API.

Installing

When installing Charmed Kubernetes using the Juju bundle, you can add the openstack-integrator at the same time by using the appropriate overlay file (Versions >= 1.29, Versions <= 1.28):

description: Charmed Kubernetes overlay to add native OpenStack support.
applications:
  kubernetes-control-plane:
    options:
      allow-privileged: "true"
  openstack-integrator:
    charm: openstack-integrator
    num_units: 1
    trust: true
  openstack-cloud-controller:
    charm: openstack-cloud-controller
  cinder-csi:
    charm: cinder-csi
relations:
  - [openstack-cloud-controller:certificates,            easyrsa:client]
  - [openstack-cloud-controller:kube-control,            kubernetes-control-plane:kube-control]
  - [openstack-cloud-controller:external-cloud-provider, kubernetes-control-plane:external-cloud-provider]
  - [openstack-cloud-controller:openstack,               openstack-integrator:clients]
  - [easyrsa:client,                                     cinder-csi:certificates]
  - [kubernetes-control-plane:kube-control,              cinder-csi:kube-control]
  - [openstack-integrator:clients,                       cinder-csi:openstack]

To use the overlay with the Charmed Kubernetes bundle, specify it during deploy like this:

juju deploy charmed-kubernetes --overlay ~/path/openstack-overlay.yaml --trust

...and remember to fetch the configuration file!

juju ssh kubernetes-control-plane/leader -- cat config > ~/.kube/config

For more configuration options and details of the permissions which the integrator uses, please see the charm docs.

Note:

Resources allocated by Kubernetes or the integrator are usually cleaned up automatically when no longer needed. However, it is recommended to periodically, and particularly after tearing down a cluster, use the OpenStack administration tools to make sure all unused resources have been successfully released.

Using Octavia Load Balancers

There are two ways in which Octavia load balancers can be used with Charmed Kubernetes: load balancers automatically created by Kubernetes for Services which sit in front of Pods and are defined with type=LoadBalancer, and as a replacement for the load balancer in front of the API server itself.

In either case, the load balancers can optionally have floating IPs (FIPs) attached to them to allow for external access.

Note:

For security reasons, the security groups automatically managed by Juju will not by default allow traffic into the nodes from external networks which can otherwise reach the FIPs. The easiest way to allow this is to add a rule to the model security group (named `juju-`) to allow ingress traffic from the FIP network, according to your security and network traffic policy and needs. Alternatively, you could create a separate security group to manage the rule(s) across multiple models or controllers.

Configuring or creating a security group will also be necessary if you wish to have the Amphora instances in a different subnet from the node instances, since you will need to allow at least traffic on the NodePort range (30000-32767) from the Amphorae into the nodes.

LoadBalancer-type Pod Services

To use Octavia for LoadBalancer-type services in the cluster, you will need to set the subnet-id config to the appropriate tenant subnet where your nodes reside, and if desired, the floating-network-id config to whatever network you want FIPs created in. See the Charm config docs for more details.

As an example of this usage, this will create a simple application, scale it to five pods, and expose it with a LoadBalancer-type Service:

kubectl create deployment hello-world --image=gcr.io/google-samples/node-hello:1.0
kubectl scale deployment hello-world --replicas=5
kubectl expose deployment hello-world --type=LoadBalancer --name=hello --port 8080

You can verify that the application and replicas have been created with:

kubectl get deployments hello-world

...which should return output similar to:

NAME              READY   UP-TO-DATE   AVAILABLE   AGE
hello-world      5/5               5                            5             2m38s

To check that the service is running correctly:

kubectl get service hello

...which should return output similar to:

NAME    TYPE           CLUSTER-IP       EXTERNAL-IP   PORT(S)          AGE
hello   LoadBalancer   10.152.183.136   202.49.242.3  8080:32662/TCP   2m

You can see that the External IP is now in front of the five endpoints of the example deployment. You can test the ingress address:

curl  http://202.49.242.3:8080

Hello Kubernetes!

API Server Load Balancer

If desired, the openstack-integrator can also replace kubeapi-load-balancer and create a native OpenStack load balancer for the Kubernetes API server, which simplifies the model and is properly HA, which kubeapi-load-balancer on its own is not. To enable this, use the appropriate overlay ( Versions >= 1.29, Versions <= 1.28):

applications:
  kubeapi-load-balancer: null                            # excludes the kubeapi-load-balancer
  kubernetes-control-plane:
    options:
      allow-privileged: "true"
  openstack-integrator:
    charm: openstack-integrator
    num_units: 1
    trust: true
  openstack-cloud-controller:
    charm: openstack-cloud-controller
  cinder-csi:
    charm: cinder-csi
relations:
  - [openstack-cloud-controller:certificates,            easyrsa:client]
  - [openstack-cloud-controller:kube-control,            kubernetes-control-plane:kube-control]
  - [openstack-cloud-controller:external-cloud-provider, kubernetes-control-plane:external-cloud-provider]
  - [openstack-cloud-controller:openstack,               openstack-integrator:clients]
  - [easyrsa:client,                                     cinder-csi:certificates]
  - [kubernetes-control-plane:kube-control,              cinder-csi:kube-control]
  - [openstack-integrator:clients,                       cinder-csi:openstack]
  - [kubernetes-control-plane:loadbalancer-external,     openstack-integrator:lb-consumers]

You will also need to set the lb-subnet config to the appropriate tenant subnet where your nodes reside, and if desired, the lb-floating-network config to whatever network you want the FIP created in. See the Charm config docs for more details.

Using Cinder Volumes

Many pods you may wish to deploy will require storage. Although you can use any type of storage supported by Kubernetes (see the storage documentation), you also have the option to use Cinder storage volumes, if supported by your OpenStack.

A csi-cinder-default storage class will be automatically created when the cinder-csi charm is used. This storage class can then be used when creating a Persistent Volume Claim:

kubectl create -f - <<EOY
kind: PersistentVolumeClaim
apiVersion: v1
metadata:
  name: testclaim
spec:
  accessModes:
    - ReadWriteOnce
  resources:
    requests:
      storage: 1Gi
  storageClassName: csi-cinder-default
EOY

This should finish with a confirmation. You can check the current PVCs with:

kubectl get pvc

...which should return something similar to:

NAME        STATUS   VOLUME                                     CAPACITY   ACCESS MODES   STORAGECLASS   AGE
testclaim   Bound    pvc-54a94dfa-3128-11e9-9c54-028fdae42a8c   1Gi        RWO            cinder         9s

This PVC can then be used by pods operating in the cluster. As an example, the following deploys a busybox pod:

kubectl create -f - <<EOY
apiVersion: v1
kind: Pod
metadata:
  name: busybox
  namespace: default
spec:
  containers:
    - image: busybox
      command:
        - sleep
        - "3600"
      imagePullPolicy: IfNotPresent
      name: busybox
      volumeMounts:
        - mountPath: "/pv"
          name: testvolume
  restartPolicy: Always
  volumes:
    - name: testvolume
      persistentVolumeClaim:
        claimName: testclaim
EOY

Using Keystone Authentication / Authorisation

The openstack-integrator also provides an interface for authentication and authorisation using Keystone. This is covered in detail in the Keystone and LDAP documentation.

Upgrading the integrator charm

The openstack-integrator has not specifically been tied to the version of Charmed Kubernetes installed and may generally be upgraded at any time with the following command:

juju refresh openstack-integrator

The 1.29/stable release of openstack-integrator replaces the relation for using Octavia as a loadbalancer for the API Service. The 1.29/stable release of kubernetes-control-plane drops the responsibility of deploying cinder-csi and the openstack-controller-manager. In order to upgrade the control-plane and worker charms, follow this process:

1. Upgrade the openstack-integrator charm:

juju refresh openstack-integrator --switch --channel=1.29/stable

2. Integrate the kubernetes-control-plane application:

juju integrate openstack-integrator:lb-consumers kubernetes-control-plane:loadbalancer-external

3. Deploy and migrate to the openstack-cloud-controller charm (See its charm docs for details).

4. Deploy and migrate to the cinder-csi charm (See its charm docs for details).

5. Remove the loadbalancer relation to the control-plane:

juju remove-relation openstack-integrator:loadbalancer kubernetes-control-plane:loadbalancer

Troubleshooting

If you have any specific problems with the openstack-integrator, you can report bugs on Launchpad.

For logs of what the charm itself believes the world to look like, you can use Juju to replay the log history for that specific unit:

juju debug-log --replay --include openstack-integrator/0

We appreciate your feedback on the documentation. You can edit this page or file a bug here.

See the guide to contributing or discuss these docs in our public Mattermost channel.