DEPLOYMENT WORKFLOW

Make every deployment a repeatable sequence.

Prepare, build, release, verify, and observe. Each step has a purpose and a result that the next step can check.

Before you start

Identify the PHP version, document root, extensions, build command, persistent directories, health endpoint, and rollback procedure. Do this in staging before changing production traffic.

  1. Connect a fixed source revision

    Choose the repository revision or release archive. Record the revision so a later rebuild can be traced to the same application source.

  2. Resolve dependencies and build

    Install locked dependencies, produce the application artifact, and reject a build when required files or runtime requirements are missing.

  3. Apply runtime and resource policy

    Select the web engine, document root, writable mounts, resource limits, and secret references. Never include secret values in the release artifact.

  4. Start and verify readiness

    Start the isolated runtime and check the health endpoint. Keep the current release serving until readiness and domain checks pass.

  5. Switch traffic and observe

    Record the deployment result, inspect logs and runtime state, and retain the previous artifact for rollback. Database changes may require a separate recovery procedure.

Deployment acceptance checklist

Before switching traffic, verify the actual application rather than only a process exit code. Exercise a public route, login and authorization, a database read and write, static assets, and an upload if the application accepts files. Test scheduled jobs and queue workers separately from web requests.

For persistent data, record the source and target, copy method, expected write freeze or synchronization period, and recovery point. Keep the original environment available until the new application and its data have passed the agreed checks.

  • Record the source revision and configuration identity
  • Confirm writable directories and external service access
  • Agree on rollback conditions and the person who can approve them

What happens when a step fails?

A failed build should not replace the currently serving application. A runtime that starts but fails readiness should not receive production traffic. If a health check succeeds while a business-critical path fails, investigate the application checks before continuing.

A repeated request must not create conflicting operations. Retain the request reference, outcome, and runtime evidence so an operator can distinguish a retry from a separate deployment. Production execution remains subject to the node activation gate.

YOUR APPLICATION. YOUR WORKFLOW.

Bring your PHP application, choose the runtime it needs, and keep releases, domains, and operations together in your Orisite panel.