Key Parameters Evaluated During a Cluster Upgrade

Problem

• It is expected that the master nodes should upgrade one by one.
• When the master node upgrade is triggered, the process should complete the upgrade of one master node and ensure it is in a Ready state before moving on to the next node.
• However, in some cases, this sequential upgrade logic breaks and all master nodes report NotReady status due to ongoing upgrade on them.

Environment

• Platform9 Managed Kubernetes - v5.9 and Higher

Answer

As per the current cluster upgrade workflow, the following parameters are considered for a node upgrade to proceed:

• qbert checks if the sunpike hostState is set to ok. This data is stored on the node at the location /var/opt/pf9/kube_status, where a field calledpf9_kube_node_stateindicates the status. If pf9_kube_node_state is set to ok, qbert considers the host as converged.
• Checks if the API server on the master node is responsive.

Additional Information

• An RFE with the ID PMK-6607 is already in place to update the cluster upgrade workflow to respect node status [Ready/NotReady] before proceeding with upgrade.
• The RFE is targeted for future releases.