These labs run a set of Chaos Engineering experiments against a very simple web application.
A front application is listening for incoming requests on its root path.
When it gets one, it calls the middle service over HTTP. The middle service
in turn calls the back service. The response of the back service is
forwarded back to the middle then front services, finally to the client.
The front and middle services have a timeout of 1.0 second for their
respective calls. When such timeout, or any transport error, occurs, they
still return a 200 response but with the last seen value known to them and
an error message. The client can therefore inspect the payload to determine
it receives an actual stalled/errored response.
All the applications expose a /metrics endpoint monitored by Prometheus. In
particular they expose the number of errors seen so far.
All applications run in the default namespace.