Multi-master vSphere or Openstack Tenant Cluster Fails to Upgrade
During the upgrade of a multi-master vSphere or Openstack tenant cluster, the upgrade fails because the etcd leader is reported missing.
The following error is displayed in the cloud-init logs of one of the master nodes:
Error: etcdserver: leader changed
Recommended Solution
Before upgrading a multi-master tenant cluster, run the following scripts to ensure that the etcd leader is available during the upgrade process:
Procedure
Step 1 |
Run the get-etcd-pod-name.sh script on the control plane to get the name of the etcd pod.
Sample output of this script for a multi-master vSphere cluster: control-plane-output
|
||
Step 2 |
Copy the following script move-etcd-leader.sh and the job template move-leader-job-ccp.yaml to the same directory on the master node.
move-leader-job-ccp.yaml
|
||
Step 3 |
Run the move-etcd-leader.sh script on one of the master nodes of the tenant cluster.
Sample output of the script executed on the chosen master node of the tenant cluster: tenant-cluster-master-node
|