Rolling upgrade - cluster
By incrementally upgrading nodes within the system, downtime is minimized, ensuring continuous availability and service to users throughout the process.
Prerequisite
A cluster configuration should include at least two nodes with a load balancer positioned in front. The load balancer must have the capability to be configured to redirect traffic selectively to one node at a time during the upgrade process.
Usage of a test environment is always a best practice. Even though it may be hard maintaining replicas of production setup it is a strong recommendation.
Steps of execution
Configure load balancer to redirect traffic to one node only.
Split the cluster - On all other nodes, configure them as stand alone.
Configure load balancer to redirect traffic to one of the node configured to be stand alone from previous stage.
On the remaning node, configure to be stand alone.
Execute upgrade on on all nodes not receiving traffic.
Verify functionality on upgraded nodes.
If using more than two node, create cluster including all nodes but the one receiving traffic. Include the remaining node, still working stand alone, in configuration
Redirect traffic the upgraded node(s).
Upgrade remaining node and include into cluster.
Configure load balancer to include last node in worker list.
Last updated