# Operations — <Project>

```
Status: Current
Owner: <who maintains this>
Last reviewed: <YYYY-MM-DD>
Governs: what watches this in production, and what to do when it stops
Review trigger: A new alert or a new place errors are sent; a change to the
  backup schedule or the restore procedure; any incident whose first ten minutes
  were spent working out what to run
```

> *(only for a deployed service.)* A library, a local tool or a static site with
> no backend deletes this file. Do not keep it with the headings unanswered — an
> empty runbook reads as one nobody wrote, which is worse than one that never
> applied.

## Why this exists

Every other document here describes what the project *is*. This one describes
what happens when it stops being that at three in the morning, and it is the
only document written for somebody who is not thinking clearly.

The failure it prevents is specific: **the first ten minutes of an incident
spent working out which host, which container, and which command**, while the
thing everybody actually wants to know — is it down, or is it just unreachable
from here — goes unanswered.

## Where errors go

<The service, the project or DSN, and who can see it. Name it, do not describe
it: "Sentry, project <name>, alerts to <where>" is useful at 3am and "we have
error tracking" is not.>

**A liveness tick is not this.** `scripts/healthcheck.sh` answers *is it up*.
Error tracking answers *is it working*, and the two disagree constantly: a
service returning 500 to every request is up. If this project has no error
tracking, say so here in one line rather than leaving the section blank — a gap
somebody chose is a different thing from a gap nobody noticed.

## What alerts, and to whom

| Signal | Where it goes | Who acts on it |
| --- | --- | --- |
| <> | <> | <> |

<Name a person or a rota, not a channel nobody owns.>

**An alert nobody acts on trains everybody to ignore the ones that matter.** If
a signal here has fired more than twice without anyone doing anything, it is
either wrong or it is not an alert — fix the threshold or move it to a
dashboard, and say which happened.

## Backups

| | |
| --- | --- |
| Schedule | <when, and by what — cron, systemd timer, platform feature> |
| What is captured | <database, uploads, secrets — and what is deliberately not> |
| Where it lands | <path or bucket, and who else can read it> |
| Retention | <how many, how old, what prunes them> |
| **Last verified restore** | **<YYYY-MM-DD>, by <who>** |

`scripts/backup.sh` proves a dump is readable before trusting it.
`scripts/restore-check.sh` proves it can be restored, which is a different
claim: a file that parses is not a database you can get back.

**A backup nobody has restored is a guess.** The date above is the only line in
this table that says otherwise, and it is a fact with an owner — if it is empty
or a year old, this project does not have backups, it has files.

## Rate limits and cost ceilings

*(precautionary — no incident here has yet come from this.)*

| Endpoint or job | What it costs per call | What bounds it |
| --- | --- | --- |
| <> | <> | <> |

<Anything that calls a paid API, sends mail, or runs a model belongs in this
table. So does authentication, which costs nothing per call and is the one an
attacker will hit hardest.>

**Where does the spend alert go, and at what number?** <> A ceiling nobody is
told about is a bill, not a limit.

## It is down — what now

In this order. Each step is a command that answers one question, and none of
them change anything.

1. **Is it actually down, or unreachable from here?**
   `bash scripts/status.sh` — it distinguishes those two, and reports the
   restart count. A climbing restart count with a short uptime is a crash loop,
   not a healthy service, and reads as "running" in every tool that shows only a
   state.
2. **What is it saying?** `bash scripts/status.sh --logs 200`.
3. **Is the deployed version the one you think?**
   `bash scripts/status.sh --deployed-version`. Non-zero means it could not be
   read — stop, rather than assuming.
4. **Did something change recently?** The tracker and `git log` — a deploy, a
   migration, a config edit.
5. **Roll forward or back:** `python3 scripts/deploy.py` to move the stack to a
   published image. Publishing and deploying are separate acts; this is the
   second one.

<Add the steps this project actually needs — a queue to drain, a cache to warm,
a partner to notify — and delete any of the above it does not have.>

**Write the incident down afterwards**, in `docs/history/DEVELOPMENT_LOG.md`
with the date, and file what broke as an issue with a severity label. An
incident nobody recorded happens again with the same surprise.