Skip to content

Latest commit

 

History

History
54 lines (44 loc) · 2.77 KB

static-pods.md

File metadata and controls

54 lines (44 loc) · 2.77 KB
Raw

Kubernetes Components as Static Pods

Starting with Kismatic v1.3.0, the Kubernetes control plane components are deployed as static pods instead of systemd services. These include the API server, the controller manager and the scheduler. The kube-proxy and calico-node components also run as a pod on each node. These static pods are deployed into the kube-system namespace.

The Kubelet is the only Kubernetes component that continues to be deployed as a systemd service, along with docker and etcd.

The following is a listing of pods that are running in the kube-system namespace. As you can see, the API server, controller manager, scheduler and kube-proxy are all running as pods on the cluster. Each static pod has the node's hostname appended to it's name (e.g. kube-proxy-node001 is running on node001).

# kubectl get pods -n kube-system
NAME                                    READY     STATUS    RESTARTS   AGE
calico-node-rxp2b                       2/2       Running   0          2m
default-http-backend-4kpt3              1/1       Running   0          1m
ingress-q65t5                           1/1       Running   0          1m
kube-apiserver-node001                  1/1       Running   0          3m
kube-controller-manager-node001         1/1       Running   0          3m
kube-dns-b0d29                          3/3       Running   0          2m
kube-proxy-node001                      1/1       Running   0          2m
kube-scheduler-node001                  1/1       Running   0          3m
kubernetes-dashboard-1389325151-90dzp   1/1       Running   0          1m

The fact that most of the Kubernetes components are now being hosted by Kubernetes itself has implications when it comes to troubleshooting and getting logs. If the API Server is running, you may use kubectl logs to get logs for a component.

For example, if you are interested in the controller manager's logs, you may use the following command (Replace node001 with the node's hostname):

kubectl logs -n kube-system kube-controller-manager-node001

In the case that the API Server is down or unavailable, you must use docker logs to get logs from the components.

For example, if you are interested in diagnosing why the API server is down, you may use the following command:

# This gets the container ID of the API server container.
CONTAINER_ID=$(docker ps -a | grep kube-apiserver-amd64 | cut -f 1 -d " ")
docker logs $CONTAINER_ID

Upgrading to Static Pods

When upgrading to v.1.3.0, the systemd services that will be replaced by static pods are stopped, and the packages removed. Once done, the necessary container images are downloaded and static pod manifests are created in /etc/kubernetes/manifests. The kubelet is configured to watch this directory for static pod manifests.