Pumba: chaos testing tool for Docker

Pumba is a chaos testing command line tool for Docker containers. Pumba disturbs your containers by crashing containerized application, emulating network failures and stress-testing container resources (cpu, memory, fs, io, and others).

Prerequisites

Important:: Minimal required Docker version v18.06.0.

Demo

Usage

You can download Pumba binary for your OS from release page.

$ pumba help

Pumba version [VERSION](./blob/master/VERSION)
NAME:
   Pumba - Pumba is a resilience testing tool, that helps applications tolerate random Docker container failures: process, network and performance.

USAGE:
   pumba [global options] command [command options] containers (name, list of names, RE2 regex)

VERSION:
   [VERSION](./blob/master/VERSION) - `git rev-parse HEAD --short` and `build time`

COMMANDS:
     kill     kill specified containers
     netem    emulate the properties of wide area networks
     pause    pause all processes
     stop     stop containers
     rm       remove containers
     help, h  Shows a list of commands or help for one command

GLOBAL OPTIONS:
   --host value, -H value      daemon socket to connect to (default: "unix:///var/run/docker.sock") [$DOCKER_HOST]
   --tls                       use TLS; implied by --tlsverify
   --tlsverify                 use TLS and verify the remote [$DOCKER_TLS_VERIFY]
   --tlscacert value           trust certs signed only by this CA (default: "/etc/ssl/docker/ca.pem")
   --tlscert value             client certificate for TLS authentication (default: "/etc/ssl/docker/cert.pem")
   --tlskey value              client key for TLS authentication (default: "/etc/ssl/docker/key.pem")
   --log-level value, -l value  set log level (debug, info, warning(*), error, fatal, panic) (default: "warning") [$LOG_LEVEL]
   --json                      produce log in JSON format: Logstash and Splunk friendly
   --slackhook value           web hook url; send Pumba log events to Slack
   --slackchannel value        Slack channel (default #pumba) (default: "#pumba")
   --interval value, -i value  recurrent interval for chaos command; use with optional unit suffix: 'ms/s/m/h'
   --label value               filter containers by labels, e.g '--label key=value' (multiple labels supported)
   --random, -r                randomly select single matching container from list of target containers
   --dry-run                   does not create chaos, only logs planned chaos commands
   --help, -h                  show help
   --version, -v               print the version

Kill Container command

pumba kill -h

NAME:
   pumba kill - kill specified containers

USAGE:
   pumba [global options] kill [command options] containers (name, list of names, RE2 regex)

DESCRIPTION:
   send termination signal to the main process inside target container(s)

OPTIONS:
   --signal value, -s value  termination signal, that will be sent by Pumba to the main process inside target container(s) (default: "SIGKILL")
   --limit value, -l value   limit to number of container to kill (0: kill all matching) (default: 0)

Pause Container command

pumba pause -h

NAME:
   pumba pause - pause all processes

USAGE:
   pumba pause [command options] containers (name, list of names, RE2 regex)

DESCRIPTION:
   pause all running processes within target containers

OPTIONS:
   --duration value, -d value  pause duration: should be smaller than recurrent interval; use with optional unit suffix: 'ms/s/m/h'

Stop Container command

pumba stop -h
NAME:
   pumba stop - stop containers

USAGE:
   pumba stop [command options] containers (name, list of names, RE2 regex)

DESCRIPTION:
   stop the main process inside target containers, sending  SIGTERM, and then SIGKILL after a grace period

OPTIONS:
   --time value, -t value  seconds to wait for stop before killing container (default 10) (default: 10)

Remove (rm) Container command

pumba rm -h

NAME:
   pumba rm - remove containers

USAGE:
   pumba rm [command options] containers (name, list of names, RE2 regex)

DESCRIPTION:
   remove target containers, with links and voluems

OPTIONS:
   --force, -f    force the removal of a running container (with SIGKILL, default: true)
   --links, -l    remove container links (default: false)
   --volumes, -v  remove volumes associated with the container (default: true)

Network Emulation (netem) command

pumba netem -h

NAME:
   Pumba netem - delay, loss, duplicate and re-order (run 'netem') packets, to emulate different network problems

USAGE:
   Pumba netem command [command options] [arguments...]

COMMANDS:
     delay      delay egress traffic
     loss
     duplicate
     corrupt
     rate       limit egress traffic

OPTIONS:
   --duration value, -d value   network emulation duration; should be smaller than recurrent interval; use with optional unit suffix: 'ms/s/m/h'
   --interface value, -i value  network interface to apply delay on (default: "eth0")
   --target value, -t value     target IP filter; comma separated. netem will impact only on traffic to target IP(s)
   --tc-image value             Docker image with tc (iproute2 package); try 'gaiadocker/iproute2'
   --help, -h                   show help

Network Emulation Delay sub-command

pumba netem delay -h

NAME:
   Pumba netem delay - delay egress traffic

USAGE:
   Pumba netem delay [command options] containers (name, list of names, RE2 regex)

DESCRIPTION:
   delay egress traffic for specified containers; networks show variability so it is possible to add random variation; delay variation isn't purely random, so to emulate that there is a correlation

OPTIONS:
   --time value, -t value          delay time; in milliseconds (default: 100)
   --jitter value, -j value        random delay variation (jitter); in milliseconds; example: 100ms ± 10ms (default: 10)
   --correlation value, -c value   delay correlation; in percentage (default: 20)
   --distribution value, -d value  delay distribution, can be one of {<empty> | uniform | normal | pareto |  paretonormal}

Network Emulation Loss sub-commands

pumba netem loss -h

NAME:
   Pumba netem loss - adds packet losses

USAGE:
   Pumba netem loss [command options] containers (name, list of names, RE2 regex)

DESCRIPTION:
   adds packet losses, based on independent (Bernoulli) probability model
   see:  http://www.voiptroubleshooter.com/indepth/burstloss.html

OPTIONS:
   --percent value, -p value      packet loss percentage (default: 0)
   --correlation value, -c value  loss correlation; in percentage (default: 0)

pumba netem loss-state -h

NAME:
   Pumba netem loss-state - adds packet losses, based on 4-state Markov probability model

USAGE:
   Pumba netem loss-state [command options] containers (name, list of names, RE2 regex)

DESCRIPTION:
   adds a packet losses, based on 4-state Markov probability model
     state (1) – packet received successfully
     state (2) – packet received within a burst
     state (3) – packet lost within a burst
     state (4) – isolated packet lost within a gap
   see: http://www.voiptroubleshooter.com/indepth/burstloss.html

OPTIONS:
   --p13 value  probability to go from state (1) to state (3) (default: 0)
   --p31 value  probability to go from state (3) to state (1) (default: 100)
   --p32 value  probability to go from state (3) to state (2) (default: 0)
   --p23 value  probability to go from state (2) to state (3) (default: 100)
   --p14 value  probability to go from state (1) to state (4) (default: 0)

pumba netem loss-gemodel -h

NAME:
   Pumba netem loss-gemodel - adds packet losses, according to the Gilbert-Elliot loss model

USAGE:
   Pumba netem loss-gemodel [command options] containers (name, list of names, RE2 regex)

DESCRIPTION:
   adds packet losses, according to the Gilbert-Elliot loss model
   see: http://www.voiptroubleshooter.com/indepth/burstloss.html

OPTIONS:
   --pg value, -p value  transition probability into the bad state (default: 0)
   --pb value, -r value  transition probability into the good state (default: 100)
   --one-h value         loss probability in the bad state (default: 100)
   --one-k value         loss probability in the good state (default: 0)

pumba netem rate -h

NAME:
   Pumba netem rate - rate limit egress traffic

USAGE:
   Pumba netem rate [command options] containers (name, list of names, RE2 regex)

DESCRIPTION:
   rate limit egress traffic for specified containers

OPTIONS:
   --rate value, -r value            delay outgoing packets; in common units (default: "100kbit")
   --packetoverhead value, -p value  per packet overhead; in bytes (default: 0)
   --cellsize value, -s value        cell size of the simulated link layer scheme (default: 0)
   --celloverhead value, -c value    per cell overhead; in bytes (default: 0)

Network Emulation Duplicate sub-commands

pumba netem duplicate -h

NAME:
   Pumba netem duplicate - adds duplicate packets

USAGE:
   Pumba netem duplicate [command options] containers (name, list of names, RE2 regex)

DESCRIPTION:
   adds duplicate packets, based on independent (Bernoulli) probability model
   see:  http://www.voiptroubleshooter.com/indepth/burstloss.html

OPTIONS:
   --percent value, -p value      packet duplicate percentage (default: 0)
   --correlation value, -c value  duplicate correlation; in percentage (default: 0)

Network Emulation Corrupt sub-commands

pumba netem corrupt -h

NAME:
   Pumba netem corrupt - adds corrupt packets

USAGE:
   Pumba netem corrupt [command options] containers (name, list of names, RE2 regex)

DESCRIPTION:
   adds corrupt packets, based on independent (Bernoulli) probability model
   see:  http://www.voiptroubleshooter.com/indepth/burstloss.html

OPTIONS:
   --percent value, -p value      packet corrupt percentage (default: 0)
   --correlation value, -c value  corrupt correlation; in percentage (default: 0)

Examples

# add 3 seconds delay for all outgoing packets on device `eth0` (default) of `mydb` Docker container for 5 minutes

pumba netem --duration 5m delay --time 3000 mydb

# add a delay of 3000ms ± 30ms, with the next random element depending 20% on the last one,
# for all outgoing packets on device `eth1` of all Docker container, with name start with `test`
# for 5 minutes

pumba netem --duration 5m --interface eth1 delay \
      --time 3000 \
      --jitter 30 \
      --correlation 20 \
    "re2:^test"

# add a delay of 3000ms ± 40ms, where variation in delay is described by `normal` distribution,
# for all outgoing packets on device `eth0` of randomly chosen Docker container from the list
# for 5 minutes

pumba --random netem --duration 5m \
    delay \
      --time 3000 \
      --jitter 40 \
      --distribution normal \
    container1 container2 container3

# Corrupt 10% of the packets from the `mydb` Docker container for 5 minutes

pumba netem --duration 5m corrupt --percent 10 mydb

tc tool

Pumba uses tc Linux tool for network emulation. You have two options:

1. Make sure that container, you want to disturb, has tc tool available and properly installed (install iproute2 package)
2. Use --tc-image option, with any netem command, to specify external Docker image with tc tool available. Pumba will create a new container from this image, adding NET_ADMIN capability to it and reusing target container network stack. You can try to use gaiadocker/iproute2 image (it's just Alpine Linux 3.3 with iproute2 package installed)

Note: For Alpine Linux based image, you need to install iproute2 package and also to create a symlink pointing to distribution files ln -s /usr/lib/tc /lib/tc.

Stress testing Docker containers

Pumba can inject stress-ng testing tool into a target container(s) cgroup and control stress test run.

NAME:
   pumba stress - stress test a specified containers

USAGE:
   pumba stress [command options] containers (name, list of names, or RE2 regex if prefixed with "re2:")

DESCRIPTION:
   stress test target container(s)

OPTIONS:
   --duration value, -d value  stress duration: must be shorter than recurrent interval; use with optional unit suffix: 'ms/s/m/h'
   --stress-image value        Docker image with stress-ng tool, cgroup-bin and docker packages, and dockhack script (default: "alexeiled/stress-ng:latest-ubuntu")
   --pull-image                pull stress-image form Docker registry
   --stressors value           stress-ng stressors; see https://kernel.ubuntu.com/~cking/stress-ng/ (default: "--cpu 4 --timeout 60s")

stress-ng image requirements

Pumba uses alexeiled/stress-ng:latest-ubuntu stress-ng Ubuntu-based Docker image with statically linked stress-ng tool.

You can provide your own image, but it must include the following tools:

1. stress-ng tool (in $PATH)
2. Bash shell
3. dockhack helper Bash script (in $PATH)
4. docker client CLI tool (runnable without sudo)
5. cgexec tool, available from cgroups-tools or/and cgroup-bin packages

Running inside Docker container

If you choose to use Pumba Docker image on Linux, use the following command:

# run 10 Docker containers named test_(index)
for i in `seq 1 10`; do docker run -d --name test_$i --rm alpine tail -f /dev/null; done

# once in a 10 seconds, try to kill (with `SIGKILL` signal) all containers named **test(something)**
# on same Docker host, where Pumba container is running

$ docker run -it --rm  -v /var/run/docker.sock:/var/run/docker.sock gaiaadm/pumba --interval=10s --random --log-level=info kill --signal=SIGKILL "re2:^test"

Note: from version 0.6 Pumba Docker image is a scratch Docker image, that contains only single pumba binary file and ENTRYPOINT set to the pumba command.

Note: For Windows and OS X you will need to use --host argument, since there is no unix socket /var/run/docker.sock to mount.

Running Pumba on Kubernetes cluster

If you are running Kubernetes, you can take advantage of DaemonSets to automatically deploy the Pumba on selected K8s nodes, using nodeSelector or nodeAffinity, see Assigning Pods to Nodes.

You'll then be able to deploy the DaemonSet with the command:

kubectl create -f deploy/pumba_kube.yml

K8s automatically assigns labels to Docker container, and you can use Pumba --label filter to create chaos for specific Pods and Namespaces.

K8s auto-assigned container labels, than can be used by Pumba:

"io.kubernetes.container.name": "test-container"
"io.kubernetes.pod.name": "test-pod"
"io.kubernetes.pod.namespace": "test-namespace"

It's possible to run multiple Pumba commands in the same DaemonSet using multiple Pumba containers, see deploy/pumba_kube.yml example.

If you are not running Kubernetes >= 1.1.0 or do not want to use DaemonSets, you can also run the Pumba as a regular docker container on each node you want to make chaos (see above)

Note: running pumba netem commands on minikube clusters will not work, because the sch_netem kernel module is missing in the minikube VM!

Build instructions

You can build Pumba with or without Go installed on your machine.

Build using local Go environment

In order to build Pumba, you need to have Go 1.6+ setup on your machine.

Here is the approximate list of commands you will need to run:

# create required folder
cd $GOPATH
mkdir github.com/alexei-led && cd github.com/alexei-led

# clone pumba
git clone git@github.com:alexei-led/pumba.git
cd pumba

# build pumba binary
make

# run tests and create HTML coverage report
make test-coverage

# create pumba binaries for multiple platforms
make release

Build using Docker

You do not have to install and configure Go in order to build and test Pumba project. Pumba uses Docker multistage build to create final tiny Docker image.

First of all clone Pumba git repository:

git clone git@github.com:alexei-led/pumba.git
cd pumba

Now create a new Pumba Docker image.

DOCKER_BUILDKIT=1 docker build -t pumba -f docker/Dockerfile .

License

Code is under the Apache License v2.