CODEWITHSUNDEEP
kubernetesgoogle-cloud-platform
Here_2_learn
Here_2_learn

Reputation: 5451

Understanding the nodes of Kubernetes cluster in Google Cloud Platform

Being new to Google Cloud Platform, I have a basic question on the nodes that are part of kubernetes cluster.

With a free-tier access I have launched a kubernetes cluster with 3 nodes.

kubectl get nodes
NAME                                       STATUS    ROLES     AGE       VERSION
gke-cluster-1-default-pool-5ac7520f-097t   Ready     <none>    20h       v1.7.8-gke.0
gke-cluster-1-default-pool-5ac7520f-9lp7   Ready     <none>    20h       v1.7.8-gke.0
gke-cluster-1-default-pool-5ac7520f-vhs2   Ready     <none>    20h       v1.7.8-gke.0

While trying to explore the cluster got to know that the nodes which are launched are just pods but not VMs or servers. 

$kubectl --namespace=kube-system get pods 
NAME                                                 READY     STATUS    RESTARTS   AGE 
kube-proxy-gke-cluster-1-default-pool-5ac7520f-097t   1/1       Running   0          20h 
kube-proxy-gke-cluster-1-default-pool-5ac7520f-9lp7   1/1       Running   0          20h 
kube-proxy-gke-cluster-1-default-pool-5ac7520f-vhs2   1/1       Running   0          20h

(NOTE : removed info about other pods for readability)

The confusion here is how does the high availability of the nodes in the cluster is achieved(as the no of instances of the pod is 1/1)?

Upvotes: 0

Views: 545

Answers (1)

Rob Davis
Rob Davis

Reputation: 46

I think you're missing how the nodes are actually setup.

Each of the nodes is a separate Virtual machine in your network. They host your Kubernetes network between them. Kubeproxy is just one aspect of the node, which allows the pods, services and deployments to communicate between the nodes. There are also a bunch of extra deployments runnning on those nodes. I normally ignore all the system namespaces as it just works, but you can view them in the GCP web interface, under Kubernetes Engine, Workloads. Then delete the Is System Object flag.

Kubeproxy is started as a pod on each node, should it fail, Kubernetes would restart it on the node, which will (hopefully) heal the node and enable it to communicate correctly again.

Upvotes: 2

Related Questions