Production QA
Production QA follows your work after it merges. It records each release of a project and each incident your monitoring reports, points to the pull requests most likely behind an incident, and, for a project that opts in, files a task to fix it or revert the change. Each release has its own page, opened from a project's pipeline on the Tasks page, and the Incidents page lists incidents for the projects you can see. It is included on some plans only.
Where Production QA fits
Recording runs for every connected project; acting (filing tasks, reverts and probes) needs the project's Production QA switch on. Monitoring alerts arrive through an automation source, and a fix or revert is an ordinary task whose pull request goes through review and the merge lane like any other. What happens after a merge also feeds the command center's quality figures.
A release moves through Queued and Releasing to Succeeded, Failed, Cancelled or Skipped. An incident is Open when first seen, Resolved when it stops and Reopened if it returns; each repeat raises its count instead of adding a row.
Pages in this section
- How Production QA works — releases, incidents, suspects, and when an incident becomes a task.
- Turn on Production QA for a project — connect releases, switch it on and add its automations.
- Record monitoring alerts as incidents — connect Sentry, or another monitoring tool, so its alerts arrive on Incidents under a project.
- Find the pull request behind an incident — open an incident, confirm or dismiss its suspects, add one Jaah missed and see its tasks.
- Check whether a release broke production — open a release, see where it failed, what it shipped, its metrics and its probes.
- Check that a fix resolved an incident — see when an incident stopped, whether it came back, and whether its fix is credited to its task.
- Revert a release that breaks production — set the Revert policy and what counts as a regression, then follow the revert task Jaah files.
- Check production after every release — add health URLs and smoke commands that run after each release and record an incident when one fails.
- Find where an incident happened — read a monitoring incident's environment, release, commit and stack, and open it in your monitoring tool.
- Respond to a failed release — open a Release failed incident, find the failing job and its log, follow its task and see it resolve on the next good release.
- Open the incident behind a task — from a task filed for an incident, open it and read what broke, its suspects and whether it is still happening.
- Respond to a merge lane slowdown — open a Lane slowdown incident, find the change that slowed a gate stage and see it resolve.
- Stop Production QA filing tasks for a project — clear a project's Production QA checkbox so it stops filing tasks and running probes, while incidents are still recorded.
- Confirm a pull request reached production — find the release that carried a task, open its run and check it succeeded with your pull request in it.
- Find a release run by environment or status — list every run of a project under Runs, narrow it to one environment or status and open one.
- Find a project's open incidents — narrow Incidents to one project's open or reopened incidents, worst first, and share the view.
- Respond to a failed health check — open a Health incident, read the failed probe's output, fix it and see the next release resolve it.
- Release runs — the page of one release: range, pull requests, status, metrics and probes.
- Incidents — the Incidents page and an incident's own page: severity, state, suspects, history, tasks and ledger.
- How projects work — where a project's Production QA tab sits.
- How automations work — the sources that bring monitoring alerts in.