Cabot is a application monitoring service. Basically a system that you deploy to your own infrastructure and configure it to check the URLs of your applications at a given interval, and possibly enable some kind of alert when things go down.

It looks like this:

screen shot 2016-11-19 at 11 23 48 pm

There's a lot of services that offer similar functionality (Pingdom, Uptime Robot), but Cabot is self-hosted, open-source and free, so you don't have to worry about constraints like a minimum check interval or a maximum number of applications that you can monitor.

This is a pretty basic "get started" that shows how to quickly get Cabot working. I will show only one way of doing each thing.

Installation

Install with docker, using cabot-docker:

  1. Install docker + docker-compose.

  2. git clone https://github.com/shoonoise/cabot-docker.git

  3. cd cabot-docker and take a look at cabot_env file. There's a lot of variables you can define there to configure your alerts and other integrations, just comment all of them so we don't have invalid values later.

  4. docker-compose up -d

It should go up on port 8080 (change it on docker-compose.yml if you want). The default credentials are docker/docker.

Concepts

3 major concepts in Cabot: check, instance and service.

Check is some particular task you want to run to check something. Checks can be of some predefined types, like:

  • ping: a ping to a host
  • HTTP: call an URL and check the HTTP status.

Instance is an actual instance of a machine that will have some service running. It will have a IP/hostname.

Service is the macro stuff you want to monitor.

Cabot's documentation is pretty obscure about those concepts, so here is my take on how to use them.

Configuring

There's an N-N relationship between those 3 entities, so it can be pretty confusing how to organize and relate everything. This post is pretty opinionated about how to do it.

You can have a service with multiple instances, and multiple checks pointing to each instance.

I'm using Cabot to monitor both my front-end servers and my API, of multiple applications, running each on multiple servers.

I start by creating my instances. Every newly created instance will have an IP/hostname, and will automatically create a ping check to it. For each instance, I create an HTTP check that checks for HTTP status and some text match on the return. Then I create a service and group on it all of its instances.

Automating

Clicking is tedious, error prone and slow. Mostly, it is the wrong way.

Unfortunately, Cabot doesn't have a REST API that we can use to configure our services automatically.

The way I see it, we have 2 options:

  • Edit 2016-12-23: deprecated. See cabot-db-config bellow I've created cabot-zombie, that spins up a headless browser and configures Cabot based on a Javascript object. Check out the project for more details.

  • Access the Postgres instance that Cabot uses and make some inserts directly. The tables involved are:

    • cabotapp_instance_status_checks
    • cabotapp_instance
    • cabotapp_instancestatussnapshot
    • cabotapp_service_instances
    • cabotapp_service_status_checks
    • cabotapp_service
    • cabotapp_servicestatussnapshot
    • cabotapp_statuscheck
    • cabotapp_statuscheckresult

  • I've created cabot-db-config, which uses a configuration similar to cabot-zombie, but is way faster and more reliable, because it connects directly to your Cabot database to insert the configuration.

Actually what I use currently is cabot-db-config to create stuff, and if I want to change something, I just delete everything from the tables above and rerun cabot-zombie with my changes.