API reference¶
Everything here is generated from the source docstrings, so it matches
what help(...) prints, organized by what you are trying to do.
To run your system as a fleet of nodes, you need
Cluster (Docker, on your laptop) or
KubeCluster (the class Kubernetes cluster) — both
expose the same nodes, the same process faults, and the same PEERS contract.
To break the network, use the Net object the net fixture
hands you (partitions, and delay/loss via netem — Linux's
network-emulation tool — all verified before your test continues), or
chaos.apply for Chaos Mesh experiments of
your own design on Kubernetes. To put load on the system and record what
happened, run_workload drives your ops generator through
your adapter and writes a run directory; RunHandle
reads it back. To ask questions of the run, the check
functions turn a run's files into three-way verdicts —
violated, clean, or insufficient-evidence, never a silent pass. And the
pytest fixtures and the hokea CLI (command-line
tool) connect these together: the fixtures wire all of the above into a
test function's arguments, and the CLI covers running the suite on the
class cluster (hokea test), exports, and deploying your system and
leaving it running so you can look at it by hand.