Skip to content


Nodes can be grouped for provisioning, isolating workloads from one another. For example, Kubernetes cluster might have a dedicated node group solely for database services. To achieve this, nodes can be configured with specific node taints. Pods that lack specified tolerations for those taints will avoid scheduling on such node groups.

For more information on toleration and taints, please follow the official documentation.

Kubedeploy offers the ability to define custom tolerations for Pods with tolerations value configuration.


By default tolerations is undefined.

Let's assume we have a specific node group running the c6 instance family, intended only for database workloads. Nodes have a specific label set: workload-type: database.

Additionally, there's a taint named workload-type with the value database and the effect NoSchedule. If we wish to target this specific node group, we can't rely solely on Karpenter's known labels. Using those labels would provision new nodes of the desired instance family, where other workloads could be scheduled if free capacity is available.

To target this specific group, we need to use a nodeSelector with tolerations.

Define custom tolerations

nameOverride: my-app
  repository: nginx
  tag: latest

  workload-type: database
  - key: workload-type
    value: database
    effect: NoSchedule
Deploy command
helm install webapp sysbee/kubedeploy -f values.yaml

This configuration targets the specific workload-type: database node group. Kubernets cluster administrators can then control which instance family is required for this workload on a system level, rather than defining the desired instance family in each deployment. The toleration key ensures that our Pods will tolerate the taints set on the node group.

See also: