iter8-analytics
iter8 enables statistically robust continuous experimentation of microservices in your CI/CD pipelines.
For in-depth information about how to use iter8, visit iter8.tools.
In this README:
Introduction
Use an iter8 experiment to safely expose competing versions of a service to application traffic, gather in-depth insights about key metrics for your microservice versions, and intelligently rollout the best version of your service.
Iter8’s expressive model of cloud experimentation supports a variety of CI/CD scenarios. Using an iter8 experiment, you can:
- Run a conformance test with a single version of a microservice.
- Perform a canary release with two versions, a baseline and a candidate. Iter8 will shift application traffic safely and gradually to the candidate, if it meets the criteria you specify in the experiment.
- Perform an A/B test with two versions – a baseline and a candidate. Iter8 will identify and shift application traffic safely and gradually to the winner, where the winning version is defined by the criteria you specify in the experiment.
- Perform an A/B/n test with multiple versions – a baseline and multiple candidates. Iter8 will identify and shift application traffic safely and gradually to the winner.
Under the hood, iter8 uses advanced Bayesian learning techniques coupled with multi-armed bandit approaches to compute a variety of statistical assessments for your microservice versions, and uses them to make robust traffic control and rollout decisions.
Repositories
The components of iter8 are divided across a few github repositories.
- iter8 The main iter8 repository containing the kubernetes controller that orchestrates iter8's experiments.
- iter8-analytics This repository containing the iter8-analytics component.
- iter8-trend The repository contains the iter8-trend component.
In addition,
- iter8's extensions to Kiali is contained in kiali, kiali-ui, and k-charted.
- iter8's extensions to Kui is contained in kui.
Developers
This section is for iter8 developers and contains documentation on running and testing iter8-analytics locally.
Running iter8-analytics v1.0.0 locally
The following instructions have been tested in a Python 3.7.4 virtual environment.
1. git clone [email protected]:iter8-tools/iter8-analytics.git
2. cd iter8-analytics
3. pip install -r requirements.txt
4. pip install -e .
5. export ITER8_ANALYTICS_METRICS_BACKEND_URL=<URL of your prometheus service>
6. cd iter8_analytics
7. python fastapi_app.py
Navigate to http://localhost:5555/docs on your browser. You can interact with the iter8-analytics service and read its API documentation here. When you POST a request to iter8-analytics, it interacts with Prometheus -- make sure your Prometheus URL in step 5 is accessible if you want the POST to work.
Running unit tests for iter8-analytics v1.0.0 locally
The following instructions have been tested in a Python 3.7.4 virtual environment.
1. git clone [email protected]:iter8-tools/iter8-analytics.git
2. cd iter8-analytics
3. pip install -r requirements.txt
4. pip install -r test-requirements.txt
5. pip install -e .
6. export ITER8_ANALYTICS_METRICS_BACKEND_URL=<URL of your prometheus service>
7. make test
You can see the coverage report by opening htmlcov/index.html on your browser. The prometheus URL in step 6 is a dummy URL since all Prometheus calls are mocked in unit tests.
472 Jan 02, 2023
1 Aug 26, 2021
162 Dec 28, 2022
31 Dec 09, 2022
132 Dec 28, 2022
251 Jan 09, 2023
6 Nov 29, 2021
64 Nov 25, 2022
2 Apr 19, 2022
7 Jan 06, 2022
1 Oct 31, 2021
75 Dec 08, 2022
188 Jan 09, 2023
60 Dec 17, 2022
239 Dec 28, 2022
3.7k Jan 05, 2023
18 Oct 31, 2022
1.6k Dec 31, 2022
2 Apr 28, 2021
274 Jan 08, 2023