Aller au contenu

Node-exporter - k8s 1.25 post-upgrade steps#

This procedure needs to be applied after a cluster upgrade in 1.25, due to an incompatibility between K8s 1.25 version and the Caascad current versions of node-exporter.

In order to fix node-exporter on a cluster that has been upgraded in 1.25, you need to force remove the old version of node-exporter and install a more recent version.

Steps#

  1. Remove node-exporter secrets: 

    kswitch ${ZONE}
kubectl get secrets -n <namespace> | grep sh.helm.release.v1.caascad-node-exporter # to identify the secrets names
kubectl delete secrets sh.helm.release.v1.caascad-node-exporter.v1 -n <namespace>
kubectl delete secrets sh.helm.release.v1.caascad-node-exporter.v2 -n <namespace>
...
    Where namespace can be:

    • monitoring for cloud and infra zones;
    • caascad-monitoring for client zones.

  2. Sectets deletion causes the uninstall of node-exporter helm chart. If it's not the case, force te removal: 

    kswitch ${ZONE}
helm list -a -n  <namespace> | grep node-exporter # to identify the chart name
helm uninstall caascad-node-exporter -n <namespace> --force

  3. Install node-exporter helm chart: 

    cd ${ENVS_NG}/contexts/caascad
trackbone apply -c node-exporter -z ${ZONE}

  4. Check node-exporter pod status and logs.