Cluster unavailable in Rancher dashboard after cross-cluster restore

Summary

This article explains why a cluster may be unavailable in the Rancher dashboard after a cross-cluster restore has completed.

Description

When performing a cross-cluster Rancher restore, please ensure that Rancher system namespaces (with the “cattle-” prefix) are excluded from the restore. The configuration stored in Rancher namespaces is specific to each cluster, and restoring these namespaces to a different cluster may result in issues. For example, the cluster might no longer be accessible from the Rancher dashboard due to connectivity problems.

Here is a sample log message from Rancher agent indicating that its connection is broken:

time="2023-07-17T08:25:02Z" level=error msg="Failed to dial steve aggregation server: read tcp 192.168.212.94:57366->xx.xxx.xxx.xxx:443: use of closed network connection"

If you face this issue and still have access to the cluster, you can try restarting the Rancher agent pods to correct it. Each pod should start a new connection, and the cluster should be accessible again in the Rancher dashboard.