The command line for your datacenter!
Python HTML Shell Makefile
Pull request Compare This branch is 310 commits behind dcos:master.
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Failed to load latest commit information.
bin Updated typo in lowercasing 'exe' variables instead of 'EXE'. Jul 12, 2017
LICENSE Update description and classifier, add version specification and long… May 6, 2015


DC/OS Command Line Interface

The DC/OS Command Line Interface (CLI) is a cross-platform command line utility that provides a user-friendly yet powerful way to manage DC/OS clusters.

Installation and Usage

If you're a user of DC/OS, please follow the installation instructions. Otherwise, follow the instructions below to set up your development environment.

Detailed help and usage information is available through the dcos help command and for specific subcommands through dcos <subcommand> --help.

Additional documentation for the CLI and for the DC/OS in general is available in the DCOS docs.

Parsing CLI Output

The CLI outputs either whitespace delimited tables which can be processed by all of your favourite Unix/Linux tools like sed, awk and grep, or text formatted as JSON when using the --json flag.

If using JSON, you can combine it with the powerful jq utility. The example below installs every package available in the DC/OS repository:

dcos package search --json | jq '.[0].packages[].name' | xargs -L 1 dcos package install --yes

Note: The CLI output supports support UTF-8 encoding for stdout and stderr. Please follow your platform's instructions on how to do that.

Development Dependencies

  1. git must be installed to download the source code for the DC/OS CLI.
  2. python version 3.5.x must be installed.
  3. If make env fails you may be missing required dependencies for cryptography. See here for more information or use our dockerfile that builds with all necessary dependencies.
  4. virtualenv must be installed and on the system path in order to install legacy subcommands. New subcommands are packaged as platform specific executable or platform specific Zip archives.
  5. win_bash must be installed if you are running this in Windows in order to run setup scripts from the Makefiles.


  1. Make sure you meet requirements for installing packages

  2. Clone git repo for the dcos cli:

    git clone
  3. Change directory to the repo directory:

    cd dcos-cli
  4. Create a python virtual env for the dcos project:

    make env
  5. Create a virtualenv for the dcoscli project:

    cd cli
    make env

Configure Environment and Run

  1. From the cli directory, source the virtualenv activation script to add the dcos command line interface to your PATH:

    source env/bin/activate
  2. Configure the CLI, changing the values below as appropriate for your local installation of DC/OS:

    dcos cluster setup
  3. Get started by calling the DC/OS CLI help:

    dcos help

Running Tests


Tox, our test runner, tests against Python 3.5. We have a set of tests in the dcos package (root directory) and in the dcoscli package (cli directory). When running the tests described below, change directory to one of those two and follow the instructions.


Before you can run the DC/OS CLI integration tests, you need to modify your environment appropriately.

  1. Create a temporary directory to hold your DC/OS configuration files for the duration of the tests:

    $ export DCOS_DIR=$(mktemp -d)

    NOTE: You must set the environment variable for DCOS_DIR when creating this directory. The CLI relies on this variable to know where to look store its config data.

    NOTE: You don't have to create a new directory every time you run the tests, but make sure you clear out DCOS_DIR each time you run the tests to avoid conflicts with previous runs.

  2. Copy a static dcos.toml configuration file from the source repo into this folder:

    $ cp cli/tests/data/dcos.toml ${DCOS_DIR}
  3. Set the proper permissions on this file so that it can be used by the CLI:

    $ chmod 600 ${DCOS_DIR}/dcos.toml
  4. Export the DCOS_CONFIG environment variable so that the CLI knows to use this file for its default config:

    $ export DCOS_CONFIG=<dcos-config-folder>/dcos.toml
  5. Set the CLI_TEST_SSH_KEY_PATH to point at appropriate ssh credentials to your cluster. This is used by the node integration tests:

    $ export CLI_TEST_SSH_KEY_PATH=<path-to-ssh-key>
  6. Add the following resolution to your /etc/hosts file. The ssl integration tests resolve to test SSL certs:

    $ echo "<cluster-ip-or-url>" >> /etc/hosts
  7. Finally, once all of this is set up, you need to launch a DC/OS cluster with the appropriate capabilities (see below in the section on Running) and manually log into it:

    $ dcos cluster setup <cluster-ip-or-url>


There are two ways to run tests, you can either use the virtualenv created by make env above:

make test

Or, assuming you have tox installed (via sudo pip install tox):


Either way, tox will run unit and integration tests in Python 3.5 using a temporarily created virtualenv.

NOTE: In order for all the integration tests to pass, your DC/OS cluster must have the experimental packaging features enabled. In order to enable these features the staged_package_storage_uri and package_storage_uri configuration paramenters must be set at cluster setup. See dcos configuration parameters for more information.

The easiest way to launch a cluster with these capabilities is to use dcos-launch with the configuration listed below:

launch_config_version: 1
deployment_name: ${DEPLOYMENT_NAME}
template_url: ${TEMPLATE_URL}
provider: aws
aws_region: us-west-2
    KeyName: default
    PublicSlaveInstanceCount: 1
    SlaveInstanceCount: 1

Where DEPLOYMENT_NAME is a custom name set by the user, and TEMPLATE_URL is the URL of an appropriate EC2 cloud formation template for running the integration tests. Unfortunately, the full integration test suite can only be run against an Enterprise DC/OS cluster (which you need special permissions to launch).

For Mesosphere employees the URL of this cloud formation template can be found here:

For everyone else, you can still run the integration test suite against a non EE cluster, but please be aware that some of the tests may fail.

Assuming you have tox installed, you can avoid running the full test suite by running a specific test (or any tests matching a specific pattern) by executing:

tox -e py35-integration /<test-file>.py -- -k <test-pattern>

Other Useful Commands

  1. List all of the supported test environments:

    tox --listenvs
  2. Run a specific set of tests:

    tox -e <testenv>
  3. Run a specific integration test module:

    tox -e py35-integration /


Releasing a new version of the DC/OS CLI is only possible through an automated TeamCity build which is triggered automatically when a new tag is added.

The tag is used as the version number and must adhere to the conventional PEP-440 version scheme.

The automated build starts up three jobs to build the platform dependent executables (in Windows, OS X, and Linux).

The executables are pushed to s3 and available at<platform>/x86-64/<tag>/dcos. The links to each of the platform executables and the release notes are published at:<tag>

The automated build also publishes two packages to PyPI using the script:

  1. dcos
  2. dcoscli

These packages are available to be installed by the DC/OS CLI installation script in the mesosphere/install-scripts repository.