4 Benefits of Pod Descheduling

Eric Muccino
Mindboard
Published in
3 min readDec 21, 2022

Is your original scheduling decision no longer valid? Are new nodes added to your cluster? Are your current nodes under or over-utilized? Pod Descheduling (automatic pod eviction) could be your solution! Descheduling is a way of moving already running pods to different nodes in a cluster. This process is performed by a Kubernetes controller, known as the Descheduler. The Descheduler’s behavior is determined by its configurable policy, which is a set of rules, known as predicates and priorities. These rules guide the Descheduler’s decisions about which pods to move and where to move them.

Image: Descheduling Pods — Moving Pod 3 from Node A to Node B

Advantages of Pod Descheduling

1 - Improved resource utilization

By moving pods from heavily utilized nodes to less utilized ones, the Descheduler can help ensure that all nodes are being used effectively. This can help improve the overall performance of the cluster and prevent situations where some nodes are overwhelmed while others are underutilized. In a cluster with nodes that have specialized hardware, such as GPUs or FPGAs, the Descheduler can automatically move pods that require access to these resources to the appropriate nodes, as they become available, ensuring that the specialized hardware is used effectively.

2 - Maintenance of the desired state of the cluster

If new nodes are added to the cluster, the Descheduler can automatically move existing pods to the new nodes to ensure that the cluster remains balanced. Similarly, if taints or labels are added or removed from nodes, or if pod/node affinity requirements are no longer satisfied, the Descheduler can move affected pods to new nodes to maintain the desired state of the cluster.

3 - Enhanced scalability

By automatically evicting pods when new nodes are added to the cluster, the Descheduler can create an opportunity for the pods to be rescheduled on nodes with more resource availability. This ensures that the cluster remains balanced and scalable as it grows, making it easier to add new nodes and expand the cluster without encountering issues with resource utilization or affinity requirements.

4 - Enhanced security

In the event that a node in the cluster becomes compromised or is otherwise deemed insecure, the Descheduler can automatically move any pods that are running on that node to other nodes in the cluster. This can help prevent any sensitive data or workloads from being exposed to potential security threats.

The Descheduler is a valuable tool for managing and optimizing Kubernetes clusters. By evicting the appropriate pods, the Descheduler can help improve resource utilization, maintain the desired state of the cluster, and enhance the scalability and security of the cluster in the face of node failures and vulnerabilities. While simple Deschedulers can be implemented using heuristics, more sophisticated Descheduling can be performed with artificial intelligence and machine learning. Read more to understand how AIOps can improve the performance of a Descheduler.

--

--