Start1 min read

Plan a production installation

Choose and validate an OpsKnight deployment topology.

Choose from operational ownership, HA, and isolation requirements. No topology currently has a certified numeric production envelope, so do not choose from an uncertified requests-per-second claim.

Choose a topology

  • Single host and simplest operation: Integrated Compose.
  • Single host with worker isolation: Split Compose.
  • Split runtime with web/database connection pressure: calculate the connection budget, then add supported PgBouncer transaction pooling.
  • Docker-native multi-node HA: Docker Swarm.
  • Existing Kubernetes platform: Kubernetes.
  • Packaged, schema-validated Kubernetes: Helm.
  • GitOps or owned overlays: Kustomize.

Start with the deployment landing page for the current decision and acceptance path.

Use the complete deployment decision table and capacity methodology before converting test observations into a production plan.

Promotion checklist

  1. Pin images by immutable tag or digest and run migrations as a one-shot step.
  2. Store database, authentication, encryption, provider, and signing secrets in the platform secret store.
  3. Configure TLS, public URLs, outbound policy, and least-privilege database access.
  4. Validate liveness, readiness, metrics authentication, worker health, and scheduler ownership.
  5. Exercise backup and restore, an upgrade, and rollback with representative data.
  6. Run a synthetic incident through every delivery channel before accepting production alerts.
  7. Establish queue, provider, PostgreSQL, SSE, and projector alerts from scaling signals.

Last updated for v2.0.0

Edit this page on GitHub