chaos-mesh 1.2.7

A client to create experiments in ChaosMesh

Chaos Mesh Client

Introduction

Chaos Mesh is an open source cloud-native Chaos Engineering platform that allows you to simulate various faults and orchestrate fault scenarios in your kubernetes cluster. This client is written in Python and provides a single point of entry to create and manage experiments in Chaos Mesh.

Getting Started

To start using Chaos Mesh, please follow the installation steps in the documentation.

To create a Chaos Mesh client, you can use the following code:

from chaosmesh.client import Client, Experiment
from chaosmesh.k8s.selector import Selector

# creating the ChaosMesh client
client = Client(version="v1alpha1")

# target pods selector; by labelSector or by pods in specified namespaces
selector = Selector(labelSelectors={"app": "filebeat"}, pods=None, namespaces=None)

Experiment Types

Chaos Mesh supports various types of experiments, including Pod faults, stress tests, JVM faults, and Host faults.

Pod Faults

• Pod failure
• Pod kill
• Container kill

Stress Tests

• CPU
• Memory

JVM Faults

• GC
• Exception

Host Faults

• CPU
• Memory

Network Attach

• Partition
• Bandwidth

Experiment Examples

Here are some examples of how you can create experiments in Chaos Mesh:

Pod Failure Experiment

# name of the experiment
exp_name = str(uuid.uuid4())

# starting up the pod failure experiment
client.start_experiment(Experiment.POD_FAILURE, namespace="default", name=exp_name, selector=selector)

Pod Kill Experiment

exp_name = str(uuid.uuid4())

# starting up the pod kill experiment
client.start_experiment(Experiment.POD_KILL, namespace="default", name=exp_name, selector=selector)

Container Kill Experiment

exp_name = str(uuid.uuid4())

# starting up the pod kill experiment
client.start_experiment(Experiment.CONTAINER_KILL, namespace="default", name=exp_name, selector=selector, container_names=['main'])

CPU Stress Test Experiment

exp_name = str(uuid.uuid4())

# starting up the pod kill experiment
client.start_experiment(Experiment.POD_STRESS_CPU, namespace="default", name=exp_name, selector=selector, container_names=['main'])

Memory Stress Test Experiment

exp_name = str(uuid.uuid4())

# starting up the pod kill experiment
client.start_experiment(Experiment.POD_STRESS_MEMORY, namespace="default", name=exp_name, selector=selector, container_names=['main'])

GC Experiment

# name of the experiment
exp_name = str(uuid.uuid4())

client.start_experiment(Experiment.GC, namespace="default", name=exp_name, selector=selector, port=8080)

Exception Experiment

exp_name = str(uuid.uuid4())

client.start_experiment(Experiment.RAISE_EXCEPTION, namespace="default",
                        name=exp_name, selector=select

Host CPU stress

exp_name = str(uuid.uuid4())

# starting up the host cpu stress experiment
client.start_experiment(Experiment.HOST_STRESS_CPU, namespace="default", name=exp_name,
                        address=["10.225.66.224", "10.225.67.213", "10.225.66.231", "10.225.66.138", "10.225.66.192", "10.225.67.52", "10.225.67.103"],
                        load=1000)

Host Memory stress

exp_name = str(uuid.uuid4())

# starting up the host memory stress experiment
client.start_experiment(Experiment.HOST_STRESS_MEMORY, namespace="default", name=exp_name,
                        address=["10.225.66.224", "10.225.67.213", "10.225.66.231", "10.225.66.138", "10.225.66.192", "10.225.67.52", "10.225.67.103"],
                        size="30GB")

Network Partition

exp_name = "network-partition-" + random.randint(0, 1000000).__str__()

# starting up the network partition experiment
client.start_experiment(Experiment.NETWORK_PARTITION, namespace="default", name=exp_name, selector=selector, external_targets=["target"], direction="both")

Network Bandwidth

exp_name = "network-bandwidth-" + random.randint(0, 1000000).__str__()

# starting up the network bandwidth experiment
client.start_experiment(Experiment.NETWORK_BANDWIDTH, namespace="default", name=exp_name, selector=selector, rate="1bps", buffer=1, limit=1, direction="to",
                        external_targets=["target"])

Pause an experiment

In order to pause an experiment you can use the following command

# pausing the experiment
client.pause_experiment(Experiment.POD_STRESS_MEMORY, namespace="default", name=exp_name)

Delete the experiment

The experiment can be removed from the k8s cluster using the following command

client.delete_experiment(Experiment.POD_STRESS_MEMORY, namespace="default", name=exp_name)

Logging

Initializing the ChaosMesh logger

import logging, sys

logging.getLogger("chaosmesh")
logging.basicConfig(stream=sys.stdout, level=logging.DEBUG)