A Helm chart for deploying NVIDIA Dynamo Snapshot checkpoint/restore infrastructure on Kubernetes.
Overview
A Helm chart for deploying NVIDIA Dynamo Snapshot checkpoint/restore infrastructure on Kubernetes. This chart works alongside the Dynamo Operator to enable CRIU-based checkpoint and restore for GPU inference workloads, reducing cold-start times for large models from minutes to seconds. It installs the namespace-scoped snapshot agent DaemonSet, checkpoint storage, RBAC, and seccomp profile required by CRIU.
Experimental Feature: Dynamo Snapshot is currently in preview. The DaemonSet runs in privileged mode to perform CRIU checkpoint and restore operations. For more information, see the Snapshot guide.
License
NVIDIA Dynamo is released under an open-source license, Apache-2.0, making it freely available for development, research, and deployment. Disclaimer: Installing this helm chart might result in pulling and running docker images that contain software licensed under GPL-2.0, GPL-3.0+, and LGPL-2.1+. Please make sure to comply with the software license requirements.
Technical Support
- GitHub Issues: Dynamo GitHub Issues