I find it really frustrating that certain attributes in Deployments or StatefulSets are read-only, which means I have to recreate those objects to clean things up. This often leads to downtime unless my cluster has a solid failover setup. I'm wondering if there's a specific patch command that might help with this situation?
1 Answer
To avoid downtime when requiring a recreation, you might consider using a blue-green deployment strategy. This involves having a single Service that points to the Pods' labels for both the current and new versions, allowing you to roll down the old version smoothly. However, it requires careful planning around your resources and capacity.
In my case, the previous team's incorrect labeling made the process tricky. Also, bundling services in Helm charts complicates blue-green deployments, as I'd need to do it at the top load balancer level. Having an additional EKS cluster solves some issues, but it's not always feasible.