Taints and Tolerations in Kubernetes

              We all know that Kubernetes is powerful orchestration tool in the world of containers. The whole complexity of managing, distributing multiple containers across the cluster is being taken care Kubernetes OOB. In shorts it takes care of all the heavy and complex lifting for us.

Since, its K8S who takes care of all distribution and scheduling of pods across different nodes in the cluster. So, what we if we want to run specific pod on specific node only. Luckily we have option to manage this as well. In K8S its called "taint and toleration".

In general terms:

        - Taint is the capability of the node which makes node to do not let any pod to be scheduled on it.

        - On the other hand, Toleration is another capability makes that particular pod to be tolerated by specific Node.

To summarise, Taint and Toleration are used to set restrictions on the what pods can be scheduled on a node.

Let us suppose we have 3 node cluster as below and below is the state when we have pods running in normal scenarios.

Now suppose, we got a requirements where we want to schedule only specific pods on Node1 and nothing should be scheduled on that. For this now lets add a taint called "taint=blue" on Node1. After this no pod will be able to schedule on this Node, until we add tolerations to specific pod to get scheduled on Node1.

Below, we added "blue" toleration to pod "D" and then after below will be the status.

Demo -
             In below, fresh setup we'll see we don't have any taint set on worker node, though we have a taint set on Master node. That is the reason that by default nothing will be scheduled on master node.

Now, lets add a taint "Taint=Dog" to worker node and try to schedule a pod on this.

Create a pod and see the status of the pod-

You noticed that status is pending, let's see what logs say. If you see the last line it says "0/2 nodes are available: 2 node(s) had taints that the pod didn't tolerate".

Now lets create a new pod "dog" which should be tolerate to Node1.

You'll see that after adding tolerance to the pod it got scheduled to Node01 and the other pod is still in pending state.

I hope this clears out the concept of taint and toleration in K8S.