Skip to content

Operations and Deployment

Marmitakus edited this page Aug 5, 2026 · 1 revision

Operations and Deployment

Jul.IA supports two durable operating shapes.

Shape Configuration ownership Best fit
Editable Jul.IA can persist managed changes and configuration history. Console-led operations and quick rollback.
Read-only A provisioning or GitOps pipeline owns the configuration. Immutable infrastructure and controlled promotion.

Filesystem model

Keep writable state explicit and separate from the binary:

  • configuration;
  • configuration history, when managed apply is enabled;
  • cache and certificate state;
  • file-based access logs, when selected.

The canonical Linux paths, Windows layout, permissions, and container volumes are maintained in the deployment guide.

Service principles

  • Run as an unprivileged service account.
  • Keep the host filesystem read-only except for required state directories.
  • Persist ACME and cache state where appropriate.
  • Keep the admin listener on loopback unless it is deliberately protected.
  • Treat configuration and history as sensitive because they may contain credentials.
  • Use atomic, controlled delivery for configuration changes.

Health checks

Enable the admin listener to expose liveness and readiness endpoints, then use the built-in probe:

jul healthcheck -config /etc/jul/server.toml
jul healthcheck -config /etc/jul/server.toml -ready

This works in minimal or distroless environments without relying on curl or a shell.

Observability

Plan for three different outputs:

  • process and security logs;
  • request access records;
  • metrics and traces.

They have different volume, retention, and privacy characteristics. Use the observability guide and metric contract rather than copying names or labels into this wiki.

Safe change procedure

  1. Validate the candidate with jul check.
  2. Apply or stage it using the correct lifecycle mode.
  3. Confirm the correlated apply result.
  4. Confirm readiness and serving version.
  5. Watch logs and key metrics for regression.
  6. Roll back or discard a staged candidate when necessary.

Platform guides

Clone this wiki locally