Production readiness checklist
Use this checklist before a cluster takes production traffic. The items are usually done from top to bottom, but skip any that don't apply to your deployment. Each item links to the guide that owns the procedure.
- Choose the worker node model. Shared or private nodes are fixed at creation, and the choice is a security boundary. See Decide whether tenants share hardware.
- Use a production backing store. Replace the default embedded SQLite, which can't run more than one replica. See Choose a backing store.
- Run three control plane replicas. See Size for high availability and scale.
- Set control plane resources. Start with the P3 profile and adjust for your workload. See Control plane sizing and performance.
- Harden the deployment. Apply the security baseline, and the CIS hardening guide for a pod-based control plane. See Secure vCluster deployments and the hardening guide. On shared nodes, also apply shared-node hardening.
- Restrict network traffic. Enable a default network policy, and confirm the control plane cluster's CNI enforces it. Some CNIs accept NetworkPolicy objects without enforcing them. See Network policy.
- Schedule backups and test a restore. Take regular snapshots, then restore one to prove the backup works. Deployed etcd needs etcd's own tooling instead. See Snapshots and Restore snapshots.
- Plan upgrades. Know which versions are supported and how to move between them. See Upgrade vCluster and the lifecycle policy.
- Plan certificate rotation. See Rotate certificates.
- Collect logs and set alerts. Ship control plane logs somewhere durable, and alert on API latency, errors, and etcd health. See Logging and the alert thresholds in Operational guidance.
- Know how outages behave. Understand what tenants see when a control plane or the control plane cluster fails. See What happens during control plane outages?.
Next steps​
- Deploy the cluster with the choices you made. See Deploy vCluster.