grafana-alert4ml-datasource

Anomaly detection based on machine learning.

Grafana supports a wide range of data sources, including Prometheus, MySQL, and even Datadog. There’s a good chance you can already visualize metrics from the systems you have set up. Alert4ML serves as the observability layer for SRE teams, automatically detects anomalies in time-series data based on collected data, thereby reducing the effort required by fixed-threshold methods and manual identification of abnormal system behavior.

Architecture

Technology Stack

• Frontend: TS / React
• Backend: Go / CGO / Rust
• Data Transform: Arrow Dataframe

Algorithm

1. extended-isolation-forest
2. STL
3. perpetual

Getting started

Searching in the Grafana plugin catalog is not supported due to the following limitations:

• Grafana plugins in general should not access any file system. github issues
• CGO and glibc dependencies.

You can download and install this grafana plugin using various options

• From Grafana plugin catalog

• From Github release page

• Using grafana cli

  grafana-cli plugins install ibumblebee-alert4ml-datasource
  

Installation

The Grafana plugin directory must be /var/lib/grafana/plugins because CGO specifies this path for searching. Reference.

1. Download from release

2. Unzip file. unzip ibumblebee-alert4ml-datasource-v0.1.21-linux-amd64.zip

3. Move to /var/lib/grafana/plugins/ibumblebee-alert4ml-datasource

Configuration

1. Request a service account token. Home --> Administration --> Users and access --> Service accounts
2. Enter the address and token to access the Grafana API. Home --> Connections --> Data sources --> ibumblebee-alert4ml-datasource

This plugin relies on Grafana's Mixed data source mode.

Development building and running

git clone https://github.com/IBUMBLEBEE/grafana-alert4ml-datasource.git
cd grafana-alert4ml-datasource
make

Demo

Reference

• https://grafana.com/docs/grafana-cloud/machine-learning/dynamic-alerting/forecasting/
• https://docs.victoriametrics.com/anomaly-detection/