1. Setting up steadybit
First of all you need to decide whether to use our platform as Software as a Service (SaaS) or run it locally in your own infrastructure (on-prem). In order to gather enough information about your landscape you need to install an agent on each of your hosts.
2. Auto Discovery & Updates
Today's systems are anything but rigid. They change state at any time. steadybit and its agents provide you the necessary information about your system and identify relevant targets for chaos experiments by leveraging different technologies from host, container and your cloud.
3. Security first
Fine grained security down to single targets: Security, trust and confidence are extremely important when carrying out chaos experiments. For this reason steadybit offers you on many levels possibilities to reduce the effects of experiments. Model your team structure and define who can attack which targets and which attacks they can use.
4. Plan your Attack
You can choose between application, network, resource & state attacks. You can exploit resources, affect the application state or mess with the network. Regardless of the environment you operate in, steadybit provides the appropriate attacks to help you discover weaknesses.
5. Run your Experiment
As every good researcher knows, the first thing you need is a hypothesis. Craft your experiment by selecting the relevant team, choose the category( Resource, State or Network)and specify the target type (Process, Container or Host). Next up you can set the blast radius and add the steady state check you want to monitor during the experiment. If the steady state check fails at any time, the experiment is immediately stopped.
6. Check the Results
Each experiment is recorded and the results can be viewed in the Reports section. In the detailed view for the execution you can document any findings you made by the experiment. You can also indicate if any additional action is required to mitigate the incident in the future.
7. Plan and Schedule Follow-ups
Before you schedule experiments on a regular basis it is advised to successfully execute an experiment at least once. As your system state alters, regular execution of experiments is your best choice to keep track and uncover weaknesses.