Skip to content

Latest commit

 

History

History
193 lines (156 loc) · 4.81 KB

tutorial6.md

File metadata and controls

193 lines (156 loc) · 4.81 KB
Raw

Tutorial 6: scheduler

The scheduler is responsible for allocating pods to nodes. We only have one node right now, but the scheduler can scale to hundreds or thousands of nodes. It can handle arbitrary constraints (e.g. only schedule on GPU nodes, don't schedule in this region), quality of service, resource requirements, and many other things. You can think of it as a special case of the controller-manager, but it's the basis on which most other kubernetes primitives (including most controllers) rely.

Credentials

Like the controller-manager, the scheduler gets all its information through the API server. We'll need to give it credentials:

$ sudo kubeadm init phase kubeconfig scheduler
[kubeconfig] Writing "scheduler.conf" kubeconfig file

This is written out to /etc/kubernetes/scheduler.conf.

scheduler pod

The scheduler will be the last static pod we set up, so once more with feeling. The kubeadm cheat is sudo kubeadm init phase control-plane scheduler

apiVersion: v1
kind: Pod
metadata:
  labels:
    component: kube-scheduler
    tier: control-plane
  name: kube-scheduler
  namespace: kube-system
spec:
  containers:
  - image: k8s.gcr.io/kube-scheduler:v1.13.4
    name: kube-scheduler

hostNetwork

Just like the controller-manager, we need hostNetwork to access the API server.

apiVersion: v1
kind: Pod
metadata:
  labels:
    component: kube-scheduler
    tier: control-plane
  name: kube-scheduler
  namespace: kube-system
spec:
  containers:
  - image: k8s.gcr.io/kube-scheduler:v1.13.4
    name: kube-scheduler
  hostNetwork: true

Volume mounts

We only need one mount this time: our credentials.

apiVersion: v1
kind: Pod
metadata:
  labels:
    component: kube-scheduler
    tier: control-plane
  name: kube-scheduler
  namespace: kube-system
spec:
  containers:
  - image: k8s.gcr.io/kube-scheduler:v1.13.4
    name: kube-scheduler
  hostNetwork: true
  volumeMounts:
    - mountPath: /etc/kubernetes/scheduler.conf
      name: kubeconfig
      readOnly: true
  volumes:
  - hostPath:
      path: /etc/kubernetes/scheduler.conf
      type: FileOrCreate
    name: kubeconfig

Arguments

Refreshingly, the scheduler requires only one argument.

apiVersion: v1
kind: Pod
metadata:
  labels:
    component: kube-scheduler
    tier: control-plane
  name: kube-scheduler
  namespace: kube-system
spec:
  containers:
  - image: k8s.gcr.io/kube-scheduler:v1.13.4
    name: kube-scheduler
    volumeMounts:
    - mountPath: /etc/kubernetes/scheduler.conf
      name: kubeconfig
      readOnly: true
    command:
    - kube-scheduler
    - --kubeconfig=/etc/kubernetes/scheduler.conf
  hostNetwork: true
  volumes:
  - hostPath:
      path: /etc/kubernetes/scheduler.conf
      type: FileOrCreate
    name: kubeconfig

Write the scheduler

Write the pod out to /etc/kubernetes/manifests/scheduler.yaml

If all's gone well, your pod should be running properly:

$ kubectl get pods --namespace kube-system
NAME                                   READY   STATUS    RESTARTS   AGE
etcd-k8s-tutorial                      1/1     Running   0          5h39m
kube-apiserver-k8s-tutorial            1/1     Running   0          5h39m
kube-controller-manager-k8s-tutorial   1/1     Running   0          52m
kube-scheduler-k8s-tutorial            1/1     Running   0          47s

Apply a pod

Now that we've got a scheduler, we've got everything that we need to run a pod!

Here's an example pod from the kubernetes documentation. Save it as a file in your local directory, i.e. example.yaml.

apiVersion: v1
kind: Pod
metadata:
  name: myapp-pod
  labels:
    app: myapp
spec:
  containers:
  - name: myapp-container
    image: busybox
    command: ['sh', '-c', 'echo Hello Kubernetes! && sleep 3600']

Now, we can use kubectl to apply it to our cluster:

$ kubectl apply -f example.yaml
pod/myapp-pod created

Let's see how it's doing:

$ kubectl get po
NAME                     READY   STATUS    RESTARTS   AGE
myapp-pod                1/1     Running   0          12h
myapp-pod-k8s-tutorial   1/1     Running   1          18h

Looks like it's working!

We can check the logs:

$ kubectl logs myapp-pod
Hello Kubernetes!

Conclusion

At this point we have a (minimally) viable kubernetes cluster! We can connect to the API, start and stop pods, and view logs.

For a real production cluster, there are a number of additional steps. We don't have DNS, or networking, and we have a number of components that won't scale past a single instance.

But I hope at this point you're empowered to see Kubernetes as more than a black box that commands go into and pods come out of. And hopefully next time something goes wrong, you'll know where to start.