Dynamically rebalancing workloads and optimizing resource utilization in Kubernetes
Kubernetes does not rebalance your workloads — if you have five pods on one node and zero in another, Kubernetes will not act to balance out the cluster. This is unfortunate as there will be times when your cluster isn’t perfectly balanced, exposing you to disruptions and wasting resources. But you can fix this using the Descheduler — an official Kubernetes project designed to intelligently force pods to be redeployed elsewhere.
In this session, you will learn the following:
- How the Kubernetes scheduler works and why assigning the correct requests is essential.
- How the descheduler can reallocate pods to nodes that have less utilization.
- How to define policies in the descheduler to optimize placements even further.