A new tutorial is now available for deploying Spegel, a stateless peer-to-peer OCI registry mirror, on CoreWeave Kubernetes Service (CKS). Spegel speeds up container image pulls, reduces external registry dependencies, and improves cluster reliability by sharing images across nodes using a Kademlia-based Distributed Hash Table (DHT). The Deploy Spegel, a P2P registry mirror on CKS guide walks through verifying containerd configuration, deploying Spegel as a DaemonSet with Helm, and confirming that peer-to-peer image sharing is working across your cluster nodes.Documentation Index
Fetch the complete documentation index at: https://docs.coreweave.com/llms.txt
Use this file to discover all available pages before exploring further.
Overview
Spegel runs as a DaemonSet on every node and acts as a local OCI registry mirror. When containerd pulls an image, it checks Spegel first. If a peer node already has the requested image layers, those layers stream directly from the peer instead of the upstream registry. Spegel is stateless and requires no persistent storage, as it leverages containerd’s existing image cache on each node. CKS clusters are pre-configured with the containerd settings that Spegel requires, so no additional node configuration is needed after installation.Key features
- P2P image distribution: Uses a Kademlia DHT to find and pull image layers from cluster peers, reducing external registry traffic.
- Stateless deployment: Requires no persistent storage. Spegel uses the existing containerd image cache on each node.
- SUNK GPU node support: Includes configuration for running Spegel on SUNK GPU nodes with the correct tolerations and Burstable QoS memory settings.
- Built-in debug interface: The Spegel debug web interface shows peer statistics and lets you measure image pull performance directly.
- Benchmark tooling: The tutorial covers the official Spegel benchmark tool for quantitative performance validation with standardized test images.