Skip to content

anomaly detection for time series databases like prometheus

License

Notifications You must be signed in to change notification settings

ChillkroeteTTS/fischer

Repository files navigation

Fischer Build Status

Named after Gerhard Fischer, one of the inventors of the hand held metal detector. Also the german term for a fisherman.

Proper application monitoring and especially alarming is crucial to quickly identify failures in your production applications. However in an environment with many different metrics and applications maintenance and setup becomes a tedious task.

Fischer aims to keep you up to date about any unusual behaviour of your metrics while keeping the configuration effort minimal. It thus enables you to focus on developing your application instead of watching all day long on your metrics.

Fischer uses an anomaly detection algorithm based on a normal distribution fitted to historical data of your metrics. Currently only prometheus is configurable as data source, but more such as graphite will follow. Please mind that fischer is still under heavy development, if you want to contribute feel free to tackle on of the issues.

Usage

The entire configuration is done from a file called config.edn located in the working directory. See EXAMPLE_config.edn for documentation.

Lein

Do lein run in the repository folder and of you go.

Docker

Just mount your config.edn into the home directory of /fischer/.

Example:

docker run -v [PATH_TO_YOUR_CONFIG]:/fischer/config.edn fischer

Model Training Data

Please note that fischer only uses metrics which have been measured for the complete time range specified in the config.edn. If a metric provides less datapoints than the other metrics, the corresponding metric will be withhold from the anomaly detection algorithm. To get detailed information about which data was used to fit the anomaly detection model, Fischer provides a JSON under http://[HOST]:[PORT]/models containing detailed information about the training data.

License

Copyright © 2017 Tjark Smalla

Distributed under the Eclipse Public License either version 1.0 or (at your option) any later version.

About

anomaly detection for time series databases like prometheus

Resources

License

Stars

Watchers

Forks

Packages

No packages published

Languages