Production Checklist

What to check before your nodes serve production traffic.

• A namespace and a created node pool matching the node endpoint and RIVET_POOL. See Nodes & Node Pools.
• RIVET_ENDPOINT in a secret on every node. See Nodes.
• At least 60 seconds between SIGTERM and SIGKILL, delivered to the node process itself, not a wrapper such as npm start.
• At least two nodes per node pool, so Rivet Actors have somewhere to wake during a rollout.
• No idle sleeping or scale to zero on the platform. A node holds an outbound connection, and a suspended node cannot receive Rivet Actors.
• A one hour idle timeout on any proxy between your nodes and the control plane. See Ports.
• Health and metrics routes mounted if your platform probes the process. See Health checks.

Next steps

• Control plane production checklist