Introduction
Kubernetes does a great job at distributing workloads on multiple nodes. Allowing node failures to occur without affecting uptime. A critical component for this to work is that each node has to be able to pull the workload images before they can start. Each replica running on a node will incur a pull operation. The images may be pulled from geographically close registries within the cloud provider, public registries, or self-hosted registries. This process has a flaw in that each node has to make this round trip separately. Why can’t the nodes share the image among themselves?
What is Spegel?
Spegel enables each node in a Kubernetes cluster to act as a local registry mirror, allowing nodes to share images between themselves. Any image already pulled by a node will be available for any other node in the cluster to pull. This has the benefit of reducing workload startup times and egress traffic as images will be stored locally within the cluster. On top of that it allows the scheduling of new workloads even when external registries are down.