Ops Console Install

Monitor and Playbooks

Monitor is where the agent starts from a failure and ends at a pull request, with you deciding in between.

From log line to pull request

  1. Scan. The console scans each Service's logs through Datadog on a check interval and raises an Alert for each kind of failure. An Alert has a fingerprint, a severity and a route.
  2. Verification. You start one. The agent tries to explain the Alert from the code, in a scratch copy: it follows the log through the code, replays the request, tests with mock data when it can, and cross-checks your Connections.
  3. Report. The Verification ends in a Report: the steps, the replayed request and its result, the root cause and the proposed fix.
  4. Review. You confirm the Report as a bug or an improvement, or dismiss it. A dismissal can be undone.
  5. Fix. After you confirm, the agent works on a branch. By default that means a code change, mocked tests, the full test suite and a diff review. If Jira is connected and the Playbook asks for one, it makes a Ticket first.
  6. Pull request. The console pushes the branch and opens the pull request. A merged pull request resolves the Alert. You merge it yourself.

An open Alert that no Scan has seen for 14 days is resolved by the console. A Verification works in a scratch copy at the newest commit of the branch your Playbook names, and the scratch copy is removed when the Report is ready.

Playbooks

A Playbook holds the rules and ordered steps the agent follows when a Trigger starts it, the Skills it holds, the branches Verify and Fix read from, and whether a Fix makes a Ticket.

  • Triggers are the places that start the agent: Verify and Fix in Monitor, and Review PR, Check reply, Address comments, Reply in thread and Session message in Workspace.
  • Built-in Playbooks run until you attach your own. They are read-only. New Playbook gives you an editable copy.
  • Steps can be marked scratch copy, so a step like running the tests happens away from your checkout.
  • Skills are instructions the agent can load. The console ships built-in skills, and you can add your own or use a project's.

Whatever the Playbook, the console fetches origin before every run and tells the agent to work from the newest production branch.

Agent runs

The Agent runs page lists each run with its health: where it is, why it is slow or ended, where its time went, and every call it made.

Edit this page on GitHub