Skip to content

Latest commit

 

History

History
161 lines (122 loc) · 5.05 KB

real-cluster.md

File metadata and controls

161 lines (122 loc) · 5.05 KB
Raw

Installing Virtlet on a real cluster

For Virtlet to work, the following prerequisites have to be fulfilled on the nodes which will run them:

  1. Node names must be resolvable via DNS configured on the nodes
  2. AppArmor and SELinux must be disabled on the nodes
  3. Kubernetes 1.8 and 1.9 must have the MountPropagation=true feature gate enabled in API server and on all kubelet instances

Virtlet deployment consists of preparing the nodes and then deploying the Virtlet DaemonSet.

Installing CRI Proxy

Virtlet requires CRI Proxy package to be able to run as DaemonSet on the nodes and support runnings system pods like kube-proxy there. To install CRI Proxy, please follow the steps from its documentation. Repeat it on each node that's going to run Virtlet.

Deploying Virtlet DaemonSet

First, you need to apply extraRuntime=virtlet label to each node that will run Virtlet DaemonSet (replace XXXXXX with the node name):

kubectl label node XXXXXX extraRuntime=virtlet

Then you need to install image translations configmap. You can use the default one:

curl https://raw.githubusercontent.com/Mirantis/virtlet/master/deploy/images.yaml >images.yaml
kubectl create configmap -n kube-system virtlet-image-translations --from-file images.yaml

After that, you need to get virtletctl command line tool (replace N.N.N in the command below accordingly):

curl -SL -o virtletctl https://github.com/Mirantis/virtlet/releases/download/vN.N.N/virtletctl

In case if you're using Mac OS X, you need to use this command instead:

curl -SL -o virtletctl https://github.com/Mirantis/virtlet/releases/download/vN.N.N/virtletctl.darwin

You can also use virtletctl from Virtlet image, see below.

Then you can deploy Virtlet:

./virtletctl gen | kubectl apply -f -

If you want to use the latest image, you can use virtletctl from that image:

docker run --rm mirantis/virtlet:latest virtletctl gen --tag latest | kubectl apply -f -

You can also use other image tag instead of latest, just replace it in both places in the above command.

By default it has KVM enabled, but you can configure Virtlet to disable it. In order to do so, create a configmap named virtlet-config in kube-system prior to creating Virtlet DaemonSet that contains key-value pair disable_kvm=y:

kubectl create configmap -n kube-system virtlet-config --from-literal=disable_kvm=y

After completing this step, you can look at the list of pods to see when Virtlet DaemonSet is ready:

kubectl get pods --all-namespaces -o wide -w

Testing the installation

Checking basic pod startup

To test your Virtlet installation, start a sample VM:

kubectl create -f https://raw.githubusercontent.com/Mirantis/virtlet/master/examples/cirros-vm.yaml
kubectl get pods --all-namespaces -o wide -w

And then connect to console:

$ kubectl attach -it cirros-vm
If you don't see a command prompt, try pressing enter.

Press enter and you will see:

login as 'cirros' user. default password: 'gosubsgo'. use 'sudo' for root.
cirros-vm login: cirros
Password:
$

Escape character is ^]

Verifying ssh access to a VM pod

You can also ssh into the VM using virtletctl tool (available as part of each Virtlet release on GitHub starting from Virtlet 1.0).

virtletctl ssh cirros@cirros-vm -- -i examples/vmkey

Verifying accessing services from a VM pod

After connecting to the VM using one of the above methods you can check access from the VM to cluster services. To check DNS resolution of cluster services, use the following command:

nslookup kubernetes.default.svc.cluster.local

The following command may be used to check service connectivity (note that it'll give you an authentication error):

curl -k https://kubernetes.default.svc.cluster.local

You can also verify Internet access from the VM:

curl -k https://google.com
ping -c 1 8.8.8.8

If you have Kubernetes Dashboard installed (it's present in kubeadm-dind-cluster installations for example), you can check dashboard access using this command:

curl http://kubernetes-dashboard.kube-system.svc.cluster.local

This should display some html from the dashboard's main page.

Removing Virtlet

In order to remove Virtlet, first you need to delete all the VM pods.

You can remove Virtlet DaemonSet with the following command:

kubectl delete daemonset -R -n kube-system virtlet

After doing this, remove CRI proxy from each node by reverting the changes in Kubelet flags, e.g. by removing /etc/systemd/system/kubelet.service.d/20-virtlet.conf in case of kubeadm scenario described above. After this you need to restart kubelet and remove the CRI Proxy binary (/usr/local/bin/criproxy) and its node configuration file (/etc/criproxy/node.conf).

Customizing Virtlet per-node configuration

It's possible to specify per-node configuration options for Virtlet. See this document for more information.