← Back

Molina documentation

Fault injection without touching production

The honest places to find out how your software behaves when a disk lies or a network splits are expensive to visit. Running the experiment in production is one of them. Building a hypervisor team is another.

Molina runs your software in small virtual computers on SageTech hardware. The disk, network and clock belong to Molina, so it can cut the power a moment after a database said saved, or split a cluster without touching yours. Nothing is installed in your environment and nothing is added to your tests.

What you give it

Your software as a Linux container image or a docker-compose file, and one promise it has to keep. Something like no saved write is ever lost, or only one leader at a time. The first call is thirty minutes and free.

What you get back

When a promise breaks: a replay code, a recording you can read and check, and, where the failure can be reduced, a smaller set of failures that still breaks it. When nothing breaks, you get a recording of what ran and under what conditions. A clean report means what was explored held up, and nothing more.

Where it stops

Software that only runs on Windows, needs a GPU or has no container image cannot run yet. I cannot promise to find a bug. The replay evidence comes from one machine and a handful of seeds.

What Molina cannot do How it works