Skip to content

Latest commit

 

History

History
60 lines (48 loc) · 2.25 KB

infrastructure-taint-and-toleration.md

File metadata and controls

60 lines (48 loc) · 2.25 KB

Taint and Toleration

Pod Eviction

Pod eviction from node behavior can be configured by adding toleration to pod. This default value is 5 minutes in case of node unreachable or not-ready

Node Unreachable

Configure pod to evict from node in case node is unreachable.

  • Check pod default toleration

    • Check toleration

      oc describe pod $(oc get pods | grep Running | tail -n 1 | awk '{print $1}') | \
grep -A2 -i toleration

    • Output example.

      • node.kubernetes.io/not-ready is 300s (5 minutes)
      • node.kubernetes.io/unreachable is 300s (5 minutes)
      Tolerations:     node.kubernetes.io/memory-pressure:NoSchedule op=Exists
                 node.kubernetes.io/not-ready:NoExecute op=Exists for 300s
                 node.kubernetes.io/unreachable:NoExecute op=Exists for 300s

  • Set unreachable to 1 minute by adding toleration for node unreachable to deployment or deployment config. This toleration mean that Pods remain bound to nodes for 60s after unreachable of these problems is detected.

    Remark: Default value of 5 minutes is reasonable in the assumption for preventing false posivites

    • Example of deployment with unreachable toleration set to 1 minute.

      template:
metadata:
    labels:
    app: backend-native
    deploymentconfig: backend-native
spec:
    containers:
    - image: image-registry.openshift-image-registry.svc:5000/demo/backend-native@sha256:5b76fdf7113c0db6d7fddea54997dd648a55b2a04383effb82f55cdbb0419dd5
    ...
    ...
    tolerations:
    - key: node.kubernetes.io/unreachable
    operator: Exists
    effect: NoExecute
    tolerationSeconds: 60

  • Check pod toleration node.kubernetes.io/unreachable is chaged to 60s. Pod will be recreate on another node if node is unreachable for 60s

    Tolerations:     node.kubernetes.io/memory-pressure:NoSchedule op=Exists
                 node.kubernetes.io/not-ready:NoExecute op=Exists for 300s
                 node.kubernetes.io/unreachable:NoExecute op=Exists for 60s