Consider consolidating workloads that utilize the same instance type
In the migration of Kubernetes, we've taken workloads from VM's into their own node pools in Kubernetes. Some of these node pools are precisely the same instance type. Consider consolidating these workloads onto the same node pool. This requires investigation into the node pool workload and excellent knowledge of when GKE will decide when to scale that particular instances. We should decide if a failure to scale the node pool will inflict too much pain on the sidekiq shards or any workloads for that matter, running on said node pool.
Doing so has a few benefits:
- Less node pools to manage
- More efficient use of the nodes that are active