MMWatchdog documentation
Start with Get started, the path through your first hour.
Do a job
| Page | Read it when you want to |
|---|---|
| Install MMWatchdog on one node | Install MMWatchdog on a server with MMX, and prove that an alert reaches you |
| Add a check | Watch one more service, page, port, disk, or database on a node |
| See what is happening | Read the state of each check, what happened and when, and a fresh run |
| Decide when a check alerts | Choose how soon a failure alerts you, how serious the alert is, and when the problem counts as over |
| Restart a service automatically | Restart a service, or run a command of your own, when a check fails |
| Send alerts where your team sees them | Add Slack, email, a webhook, or another destination, and choose which alerts each one gets |
| Pause alerts during planned work | Do planned work without alerts, and without forgetting to turn them back on |
| Report from a scheduled job | Get an alert when a backup or another scheduled job stops succeeding |
| Stop alerts that follow another failure | Get one alert, not many, when one failure, such as a database, breaks other checks |
| Look back at measurements | See how memory, CPU, disk, or another number changed over time |
| Watch for security updates | Get an alert when security updates wait, a known exploited flaw affects the node, or it needs a reboot |
| Set up a fleet | Make one server the hub, join the others, and see them all in one place |
| Watch a fleet | See the whole fleet, one node's checks, and which node went silent |
| Change a fleet | Add or forget a node, use node groups, rotate the join code, or move the hub |
| Update MMWatchdog | Move a node to a new release and keep its checks, destinations, vault, and history |
| Remove MMWatchdog | Remove MMWatchdog from a node completely, and prove nothing is left |
| Write your own check or destination | Check something no module covers, or send alerts to a service no module serves |
Look something up
| Page | Read it when you want to |
|---|---|
| Commands | Look up a command, its flags, its --json fields, its exit codes, and what each check state means |
| Configuration files | Look up a setting, its default, a built-in limit and its reason, or a validation error |
| Alert format | See the title and body of each kind of alert, the webhook's JSON, and the fields of an event |
| Writing modules | Look up the module description, the primitives, findings, state, and the protocol of an executable module |
| Installed files | See every file, account, and unit that an install puts on a node, and what each is for |
| Module pages | Look up a module: what it checks or where it delivers, its inputs, examples, and errors |
| Limits on each distribution | See what MMWatchdog cannot do on each tested distribution, and why |
Understand and fix
| Page | Read it when you want to |
|---|---|
| How MMWatchdog works | Understand why MMWatchdog behaves as it does: checks, alerts, delivery, the fleet, and the security model |
| When something fails | Start from what you see, such as an alert that did not arrive, and find the cause and the fix |
Change MMWatchdog itself
docs/changing-mmwatchdog/ holds the release plan, how to release MMWatchdog, the design, the testing guide, the validation record, the code map, the implementation rules, the acceptance contracts, and the documentation map.
This page is docs/README.md in the MMWatchdog repository.