Re-provision a Kubernetes (k8s) cluster

Products

• CloudHealth

Issue/Introduction

When a cluster is initially provisioned, the platform retrieves the UID from the kube-system default namespace. This UID is created once when the cluster is provisioned, and then not changed through its existence.

If the cluster is re-provisioned, basically by deleting and recreating it from scratch, it would spin up with a new kube-system namespace w/ a new UID, which is expected since its technically a “new” cluster.

Resolution

CloudHealth k8s FlexReports (Containers/Nodes Usage + K8s Cost History) have a concept of never re-writing history.  Once data has been collected and generated for a date, it is never changed even if the cluster is deleted from the system.  If you were to remove the cluster from Clusters UI, these reports would not be affected.

However, that is not the same for other k8s reports (Kubernetes Resources/Allocation).  In these cases, if a cluster is removed from the UI all of its data is automatically removed from those usage + cost reports.

Original clusters can be safely removed from https://apps.cloudhealthtech.com/containers_clusters without impacting historical data.  However, this ONLY applies to the following Flexreport based reports:

https://apps.cloudhealthtech.com/ui/reports/flexreports/view/template-qname/Kubernetes_Containers_Usage
https://apps.cloudhealthtech.com/ui/reports/flexreports/view/template-qname/Kubernetes_Nodes_Usage
https://apps.cloudhealthtech.com/ui/reports/flexreports/view/template-qname/AWS_Kubernetes_Cost_History