Utility to launch and manage containers using YAML based configuration data
Switch branches/tags
Nothing to show
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Failed to load latest commit information.
doc/source
examples
paunch
releasenotes
.coveragerc
.gitignore
.gitreview
.mailmap
.testr.conf
CONTRIBUTING.rst
HACKING.rst
LICENSE
MANIFEST.in
README.rst
babel.cfg
requirements.txt
setup.cfg
setup.py
test-requirements.txt
tox.ini

README.rst

paunch

Utility to launch and manage containers using YAML based configuration data

Features

  • Single host only, operations are performed via the docker client on the currently configured docker service.
  • Zero external state, only labels on running containers are used when determining which containers an operation will perform on.
  • Single threaded and blocking, containers which are not configured to detach will halt further configuration until they exit.
  • Co-exists with other container configuration tools. Only containers created by paunch will be modified by paunch. Unique container names are assigned if the desired name is taken, and containers are renamed when the desired name becomes available.
  • Accessable via the paunch command line utility, or by importing python package paunch.

Running Paunch Commands

The only state that paunch is aware of is the labels that it sets on running containers, so it is up to the user to keep track of what paunch configs should be running so that others can be deleted on cleanup. For these examples we're going to store that state in a simple text file:

$ touch paunch-state.txt

We'll start of by deleting any containers that were started by previous calls to paunch apply:

$ paunch --verbose cleanup $(cat paunch-state.txt)

Next we'll apply a simple hello-world config found in examples/hello-world.yml which contains the following:

hello:
  image: hello-world
  detach: false

Applied by running:

$ paunch --verbose apply --file examples/hello-world.yml --config-id hi
$ echo hi >> paunch-state.txt

A container called hello will be created, print a Hello World message, then exit. You can confirm that it still exists by running docker ps -a.

Now lets try running the exact same paunch apply command:

$ paunch --verbose apply --file examples/hello-world.yml --config-id hi

This will fail with an error because there already exists a container labeled with "config_id": "hi". WARNING TODO NOT IMPLEMENTED YET

Lets try again with a unique --config-id:

$ paunch --verbose apply --file examples/hello-world.yml --config-id hi-again
$ echo hi-again >> paunch-state.txt

Doing a docker ps -a now will show that there are now 2 containers, one called hello and the other called hello-(random suffix). Lets delete the one associated with the hi config-id:

$ cat paunch-state.txt
$ echo hi-again > paunch-state.txt
$ cat paunch-state.txt
$ paunch --verbose cleanup $(cat paunch-state.txt)

Doing a docker ps -a will show that the original hello container has been deleted and hello-(random suffix) has been renamed to hello

Generally paunch cleanup will be run first to delete containers for configs that are no longer apply. Then a series of paunch apply commands can be run. If these apply calls are part of a live upgrade where a mixture of old and new containers are left running, the upgrade can be completed in the next run to paunch cleanup with the updated list of config-id state.

Paunch can also be used as a library by other tools. By default running the paunch command won't affect these other containers due to the different managed_by label being set on those containers. For example if you wanted to run paunch commands masquerading as the heat-agents docker-cmd hook then you can run:

paunch --verbose apply --file examples/hello-world.yml --config-id hi --managed-by docker-cmd

This will result in a hello container being run, which will be deleted the next time the docker-cmd hook does its own cleanup run since it won't be aware of a config_id called hi.

Configuration Format

The current format is loosely based on a subset of the docker-compose v1 format with modifications. The intention is for the format to evolve to faithfully implement existing formats such as the Kubernetes Pod format.

The top-level of the YAML format is a dict where the keys (generally) correspond to the name of the container to be created. The following config creates 2 containers called hello1 and hello2:

hello1:
  image: hello-world
hello2:
  image: hello-world

The values are a dict which specifies the arguments that are used when the container is launched. Supported keys which comply with the docker-compose v1 format are as follows:

command:
String or list. Overrides the default command.
detach:
Boolean, defaults to true. If true the container is run in the background. If false then paunch will block until the container has exited.
environment:
List of the format ['KEY1=value1', 'KEY2=value2']. Sets environment variables that are available to the process launched in the container.
env_file:
List of file paths containing line delimited environment variables.
image:
String, mandatory. Specify the image to start the container from. Can either be a repository/tag or a partial image ID.
net:
String. Set the network mode for the container.
pid:
String. Set the PID mode for the container.
privileged:
Boolean, defaults to false. If true, give extended privileges to this container.
restart:
String. Restart policy to apply when a container exits.
user:
String. Sets the username or UID used and optionally the groupname or GID for the specified command.
volumes:
List of strings. Specify the bind mount for this container.
volumes_from:
List of strings. Mount volumes from the specified container(s).