Uptime Monitor

A simple downtime monitoring and alert service for all your processes!

Great for processes and microservices that do not expose any ports or endpoints for monitoring via fabulous services such as Uptime Robot (free version). Think push instead of poll.

How it works

It sends a Pushover notification whenever your service stops sending a pulse (see endpoint below).

This app runs a CRON like process every 10 second to check which services are down.

This is a single endpoint application that needs no configuration or set up for whatever new service or process you want to monitor. Just send a HTTP request (see endpoint below) and it will do it's magic.

Dependencies

❤️ Built with NestJS

Node.js (tested with v14.17.0 - left .tool-versions in the root of this directory)
Sqlite3
Pushover account

Environment variables

PUSHOVER_TOKEN="" # https://pushover.net/apps/build
PUSHOVER_GROUP="" # Create subscription and a new custom group after creating app

To run

npm install
npm run start

Endpoint

Send a HTTP POST request to /records

curl -X POST -H "Content-Type: application/json" \
  -d '{"serviceName": "<YOUR_SERVICE_NAME>", "secondsBetweenHeartbeat": 60, "secondsBetweenAlerts": 600, "maxAlertsPerDownTime": 10}' \
  http://localhost:3000/records

Endpoint Variables Defined

serviceName -> A unique name to identify your process
secondsBetweenHeartbeat -> How many seconds before you receive the first alert when your process goes down. Generally, this should also be the same as how often your process sends a ping
secondsBetweenAlerts -> When the second alert gets out if process still not sending a ping after first alert. This to avoid it getting spammy and annoying.
maxAlertsPerDownTime -> Number of alerts per down time session before it stops sending you an alert. This will help prevent annoying you for forgetting to delete the service from the database.

Security

Feel free to implement your own basic auth or whatever. I operate this in a controlled private network environment.

Reliability

Won't it be bad if your uptime monitor goes down instead of your app?

Typical production deployment practices will help prevent that from happening. Or you could deploy a new instance of this app for every instance of this app you deploy. ;p

Tips

Use PM2
Use something like systemd (systemctl) if on Ubuntu and I think other Debian flavored distros
And of course your actual server uptime matters.

Also, may consider changing the DB used. Currently using SQLite. TypeORM and majority of this code will work with whatever Database TypeORM supports out-of-the-box. You will just need to modify ormconfig.ts.

Improvements Needed

Send back friendly errors if the single endpoint request failed
Use upsert instead of select + insert / update whenever a new request arrives
Testing

Beware

I left autoLoadEntities as true in app.module.ts. This is a TypeORM configuration which auto syncs your DB schema to your entities. This can result in deletion of columns if you change the entities. So beware when using in production!

Disclaimer

As per usual, use at your own discretion especially when using to monitor billion dollar processes.

I've no plans to maintain this publicly for now. Feel free to clone, contribute or whatever tickles your fancy.

joshteng/uptime-monitor