Can a Pod tolerate one of a set of taints - kubernetes

Question

Consider a cluster in which each node has a given taint (let's say NodeType) and a Pod can tolerate a set of NodeType. For example, there are nodes tainted NodeType=A, NodeType=B and NodeType=C.
I'd like to be able to specify for example that some Pods tolerate NodeType=A or NodeType=C, but not NodeType=B. Other Pods (in different Deployments) would tolerate different sets. Is there a way to do this?

Answer 1

Yes, it appears it is possible to do so by adding multiple tolerations with the same key on the pod's spec. An example of the same is given in the official docs.
Here is a demo I tried which works to produce the desired result.
The cluster has three nodes:
kubectl get nodes
NAME STATUS AGE VERSION
dummy-0 Ready 3m17s v1.17.14
dummy-1 Ready 26m v1.17.14
dummy-2 Ready 26m v1.17.14
I tainted them as mentioned in the question using the kubectl taint command:
kubectl taint node dummy-0 NodeType=A:NoSchedule
kubectl taint node dummy-1 NodeType=B:NoSchedule
kubectl taint node dummy-2 NodeType=C:NoSchedule
Created a Deployment with three replicas with the matching tolerations:
apiVersion: apps/v1
kind: Deployment
metadata:
name: nginx-deployment
labels:
app: nginx
spec:
replicas: 3
selector:
matchLabels:
app: nginx
template:
metadata:
labels:
app: nginx
spec:
containers:
- image: nginx
name: nginx-nfs
tolerations:
- key: "NodeType"
operator: "Equal"
value: "A"
effect: "NoSchedule"
- key: "NodeType"
operator: "Equal"
value: "B"
effect: "NoSchedule"
From the kubectl get pods command, we can see that the pods of the Deployment were scheduled only on the nodes dummy-0 and dummy-1 and not on dummy-2 which has a different taint:
kubectl get pod -o wide
NAME READY STATUS RESTARTS AGE IP NODE
nginx-deployment-5fc8f985d8-2pfvm 1/1 Running 0 8s 100.96.2.11 dummy-0
nginx-deployment-5fc8f985d8-hkrcz 1/1 Running 0 8s 100.96.6.10 dummy-1
nginx-deployment-5fc8f985d8-xfxsx 1/1 Running 0 8s 100.96.6.11 dummy-1
Further, it is important to understand that the taints and tolerations are useful to make sure that the pods don't get scheduled to a particular node.
We should use the concepts of node affinities namely affinity and anti-affinity to make sure that the pods are scheduled to a particular node.