Open source · Apache-2.0 Shipped
rotelle breaks your cluster. On purpose.
A small, sharp Kubernetes failure simulator: five reproducible failure scenarios you can trigger on cue, in your own cluster. rotelle is the public part of Infraware: no engagement needed, no account, no us. Run it with us, without us, against us.
The scenarios
Five ways to have a bad night, on demand.
Each one is a real failure mode, reproduced honestly, the pod actually crashes, the memory actually runs out.
-
crash-loopA pod that exits at startup and backs off, again and again, the classic page. -
oom-killA workload that eats memory until the kernel ends it. -
missing-env-varA deploy that ships without its configuration, the classic first run. -
service-unreachableA service whose endpoints stop answering. -
ingress-conflictTwo routes fighting over the same path.
Why it exists
One tool, three jobs.
The demo ritual
Every live demo starts with rotelle breaking a real cluster in front of you, not a recording, not a hosted sandbox on our infrastructure. You pick the failure.
The acceptance test
Pilots are accepted when a rotelle scenario runs end to end in your cluster, alert, investigation, brief, signature, audit line, with your security stakeholder watching.
The evaluation harness
Reproducible failures are how we test Virgil against known ground truth, and how you can, before believing a word of this site.
The install, the scenario controls, and the docs live in the repository README, the project stands on its own, with or without Infraware attached.