suspend / resume / restart

Three commands for managing SPCS service state without tearing anything down. Suspending drops compute cost to near-zero while preserving state on the stage-backed volume.

suspend

snowclaw suspend

Suspends the service first, then the compute pool. The persistent volume (stage) keeps all skills, workspace files, and config. Cortex will not serve traffic while suspended.

Use this between work sessions if you want to pause the compute cost but keep state.

resume

snowclaw resume

Resumes the compute pool first, then the service. Once the service is READY, traffic starts flowing again through the same public ingress URL.

restart

snowclaw restart

Restarts the service (without touching the compute pool) so it picks up fresh config from the stage. Use this after a snowclaw push to hot-reload openclaw.json or skills without rebuilding the image.

When to use what