Managing Node Taints

Taints

A taint is a key-value pair associated with an effect. The following effects are available:

• NoSchedule: No pod will be scheduled onto the node unless it has a matching toleration. Existing pods will not be evicted from the node.
• PreferNoSchedule: Kubernetes prevents pods that cannot tolerate this taint from being scheduled onto the node.
• NoExecute: If the pod has been running on a node, the pod will be evicted from the node. If the pod has not been running on a node, the pod will not be scheduled onto the node.

To add a taint to a node, run the kubectl taint node nodename command as follows:

$ kubectl get node
NAME             STATUS   ROLES    AGE    VERSION
192.168.10.170   Ready    <none>   73d    v1.19.8-r1-CCE21.4.1.B003
192.168.10.240   Ready    <none>   4h8m   v1.19.8-r1-CCE21.6.1.2.B001
$ kubectl taint node 192.168.10.240 key1=value1:NoSchedule
node/192.168.10.240 tainted

To view the taint configuration, run the describe and get commands as follows:

$ kubectl describe node 192.168.10.240
Name:               192.168.10.240
...
Taints:             key1=value1:NoSchedule
...
$ kubectl get node 192.168.10.240 -oyaml
apiVersion: v1
...
spec:
  providerID: 06a5ea3a-0482-11ec-8e1a-0255ac101dc2
  taints:
  - effect: NoSchedule
    key: key1
    value: value1
...

To remove a taint, run the following command with a hyphen (-) added after NoSchedule:

$ kubectl taint node 192.168.10.240 key1=value1:NoSchedule-
node/192.168.10.240 untainted
$ kubectl describe node 192.168.10.240
Name:               192.168.10.240
...
Taints:             <none>
...

On the CCE console, you can also manage taints of a node in batches.

1. Log in to the CCE console.
2. Click the cluster name, access the cluster details page, and choose Nodes in the navigation pane. On the page displayed, select a node and click Labels and Taints above the list.
3. In the displayed dialog box, click Add Operation under Batch Operation, choose Add/Update, and select Taint.

   Enter the key and value of the taint to be added, select the taint effect, and click OK.

4. After the taint is added, check the added taint in node data.

System Taints

When some issues occurred on a node, Kubernetes automatically adds a taint to the node. The built-in taints are as follows:

• node.kubernetes.io/not-ready: The node is not ready. The node Ready value is False.
• node.kubernetes.io/unreachable: The node controller cannot access the node. The node Ready value is Unknown.
• node.kubernetes.io/memory-pressure: The node memory is approaching the upper limit.
• node.kubernetes.io/disk-pressure: The node disk space is approaching the upper limit.
• node.kubernetes.io/pid-pressure: The node PIDs are approaching the upper limit.
• node.kubernetes.io/network-unavailable: The node network is unavailable.
• node.kubernetes.io/unschedulable: The node cannot be scheduled.
• node.cloudprovider.kubernetes.io/uninitialized: If an external cloud platform driver is specified when kubelet is started, kubelet adds a taint to the current node and marks it as unavailable. After a controller of cloud-controller-manager initializes the node, kubelet will delete the taint.

Node Scheduling Settings

To configure scheduling, log in to the CCE console, click the cluster, choose Nodes in the navigation pane, and click More > Disable Scheduling in the Operation column of a node in the node list.

In the dialog box that is displayed, click OK to set the node to be unschedulable.

This operation will add a taint to the node. You can use kubectl to view the content of the taint.

$ kubectl describe node 192.168.10.240
...
Taints:             node.kubernetes.io/unschedulable:NoSchedule
...

On the CCE console, remove the taint and set the node to be schedulable.

Tolerations

Tolerations are applied to pods, and allow (but do not require) the pods to schedule onto nodes with matching taints.

Taints and tolerations work together to ensure that pods are not scheduled onto inappropriate nodes. One or more taints are applied to a node. This marks that the node should not accept any pods that do not tolerate the taints.

Example:

apiVersion: v1
kind: Pod
metadata:
  name: nginx
  labels:
    env: test
spec:
  containers:
  - name: nginx
    image: nginx
    imagePullPolicy: IfNotPresent
  tolerations:
  - key: "key1"
    operator: "Equal"
    value: "value1"
    effect: "NoSchedule"  

In the preceding example, the toleration label of the pod is key1=value1 and the taint effect is NoSchedule. Therefore, the pod can be scheduled onto the corresponding node.

You can also configure tolerations similar to the following information, which indicates that the pod can be scheduled onto a node when the node has the taint key1:

tolerations:
- key: "key1"
  operator: "Exists"
  effect: "NoSchedule"