Gansoi Infrastructure Monitoring

Gansoi is a modern monitoring solution inspired by classical network monitoring software (Nagios and friends) updated to modern and best practices.

Disclaimer and Current State

At the moment this software is unusable for anyone but early adopters and developers. It's a just-working proof-of-concept.

Current Limitations

• No useful agents implemented yet.
• The web interface is rudimentary at best. Useless at worst.

What We Currently Have

• A distributed core implemented using Raft and BoltDB.
• Full encryption of internal cluster communication using an internal CA.
• The project requires nothing but a Go compiler when building. There is no runtime dependencies.
• A simple Slack notifier.
• A few example agents for testing.
• LetsEncrypt support for the web interface.
• It's quite performant. The smallest cloud server we could rent can easily handle hundreds of checks per second. With no optimizations yet.

Mission Statement

Gansoi's mission is to radically improve self-hosted health monitoring and alerting for ops and devops.

Design goals

• Fault tolerant and distributed - should tolerate multiple node failures with no single point of failure.
• Performant - should scale to thousands of checks per second on basic hardware.
• Zero dependencies - should install and operate without any other software.
• Geodistributed - should be deployable across the globe for full redundancy and geographically distributed checks.
• Multidimensional severity grouping - an alert should be categorized in both severity and urgency.
• No false positives - the system should never cry wolf.
• Easy encryption - should support Let’s Encrypt out of the box.
• Future-proof - Everything must support IPv6 out of the box.
• Binary outcome - either a human should do something or not.

Alert integrations

• Slack - it's what we all use and love.
• Pagerduty - oldtimers swear by this. Let's support it.
• Email - CEO's want their alerts too.
• Twilio voice calls - for waking up your ops team at 3 AM.

Transports

Checks could be self-contained and run on a Gansoi node - or they can run on a third party host. This will require a transport.

• ssh - Gansoi should support SSH as a transport. It is universally supported as a remote access protocol. Gansoi must support some form of keep-alive to avoid the constant reconnecting and handshake.
• NRPE - This is the industry standard and we should support it.

Building and development

Go 1.9 or newer is required for building Gansoi.

go get ./... should get all dependencies.

go build . should be enough to build Gansoi.

You can run a small local Gansoi cluster for testing and development:

$ ./test3.sh

A three-node local cluster will be started shortly. You can visit the web interface at https://cluster.gansoi-dev.com:9002/.

Alternatively, you can launch a single node:

$ ./gansoi demo

The node will be available at https://gansoi-dev.com:9002/.

Docker

We provide a Docker image for your convenience.

Plain Docker

You can start a Gansoi node using regular Docker with a simple command: docker run --rm -p 80:80 -p 443:443 gansoi/gansoi.

Compose and Stack

Included in this project is a docker-compose.yml for running Gansoi with Docker Compose or Docker Stack.

If you're using Docker Compose you can try Gansoi using the command docker-compose up and visiting https://gansoi-dev.com/.

The docker-compose.yml file can also be used for a Docker Stack deploy on a Swarm: docker stack deploy -c docker-compose.yml gansoi

Slack Channel

If you would like you can join other developers in the #gansoi channel on the Gophers Slack (invites to Gophers Slack are available here).