# The operations manual

<!-- module: s13-operations · deck: One live page tells the Service whether the installation is ready to run: thirteen checks every 10 minutes and on demand, the jobs and e-mails of the last day, the queues, the schedule, failed jobs to retry or forget, and an e-mail the moment a check starts failing. -->

**Table of contents · 24 October 2026 · S13 of plan CORE. The page, its checks, failed jobs, and the two commands every
deployment runs (`app:deploy`, `app:check`).**

The operations page is the Service's view of the installation itself, not of any workspace: whether the queue worker
and the scheduler are running, whether mail, the file vault and the scanner work, whether production is set up safely,
whether the database and the option lists are complete. Each check says **Passing**, **Needs a look**, **Off** or
**Failing**, in words that say what to do. The schedule checks every 10 minutes and e-mails the Service when a check
starts failing; the page checks again on demand and refreshes itself every 10 seconds.

The one sentence to carry through it: **a red line on this page is something users will notice soon; the words beside
it say what to run.**

**Who meets what**

| Seat | Who | What they do here |
|---|---|---|
| Service | The platform's owner | Reads the page, checks again, retries or forgets failed jobs, gets the e-mail |
| The server's operator | Whoever deploys | Runs `app:deploy` and `app:check`; sets up the cron line, the worker, backups |
| Owner, Admin, Member | A workspace's people | Never see it (the page is the platform's) |

**As shipped**

| | Shipped |
|---|---|
| Checks | 13: queue worker, scheduler, mail, file vault, malware scanner, debug mode, cookies and headers, backups, languages, migrations, option lists, notifications, the brand |
| The schedule checks | every 10 minutes (`app:check --record`) |
| The scheduler's heartbeat | every minute; stale after 15 minutes |
| The page refreshes | every 10 seconds (pause the feed to read it) |
| A job waiting is stuck after | 10 minutes |
| The queue-wait alert | 5 seconds (p95) |
| Backups are stale after | 26 hours (one night missed) |
| Measurements kept | jobs 48 hours, readiness runs 31 days |

---

## Part 1 · Operations

- [1.1 Reading the operations page](p01-operations.md#11-reading-the-operations-page)
- [1.2 Checking again](p01-operations.md#12-checking-again)
- [1.3 When a check starts failing](p01-operations.md#13-when-a-check-starts-failing)
- [1.4 Retrying or forgetting a failed job](p01-operations.md#14-retrying-or-forgetting-a-failed-job)
- [1.5 Deploying a release (the server's operator)](p01-operations.md#15-deploying-a-release-the-servers-operator)
- [1.6 Setting up the server for the checks (the server's operator)](p01-operations.md#16-setting-up-the-server-for-the-checks-the-servers-operator)