This is a Radial Wheel repository for running dnsmasq as PXE+DHCP server serving Coreos images.
YOUR_SSH_PUBLIC_KEY_HERE
/usr/share/oem
for example), you can create
a tarball (with the root of the tarball the same as '/'), upload it into the
hub and specify it's location with $AMEND_IMAGE
. The entrypoint script will
automatically amend it before serving it out.A very important note: in order for DHCP to work, this container must run
using the docker run --net host
option. This option uses the hosts network
stack as the containers own. That means you must choose ports in your dnsmasq
configuration that will not conflict with any ports already used on the host.
The --net host
option is not as secure as other options, so make sure to use
appropriately.
Tunable environment variables; modify at runtime. Italics are defaults.
Radial is a Docker container topology strategy that seeks to put the canon of Docker best-practices into simple, re-usable, and scalable images, dockerfiles, and repositories. Radial categorizes containers into 3 types: Axles, Hubs, and Spokes. A Wheel is a repository used to recreate an application stack consisting of any combination of all three types of containers. Check out the Radial documentation for more.
One of the main design goals of Radial containers is simple and painless modularity. All Spoke (application/binary) containers are designed to be run by themselves as a service (a Wheel consisting of a Hub container for configuration and a Spoke container for the running binary) or as part of a larger stack as a Wheel of many Spokes all joined by the Hub container (database, application code, web server, backend services etc.). Check out the Wheel tutorial for some more details on how this works.
Note also that for now, Radial makes use of Fig for all orchestration, demonstration, and testing. Radial is just a collection of images and strategies, so technically, any orchestration tool can work. But Fig was the leanest and most logical to use for now.
In case you need to modify the entrypoint script, the Dockerfile itself, create your "config" branch for dynamic building, or just prefer to build your own from scratch, then you can do the following:
fig up
A standard feature of all Radial images is their ability to be used dynamically. This means that since great care is made to separate the application code from it's configuration, as long as you make your application configuration available as a git repository, and in it's own "config" branch as per the guidelines in the Wheel template, no building of any images will be necessary at deploy time. This has many benefits as it allows rapid deployment and configuration without any wait time in the building process. However:
Dynamic builds will not commit your configuration files into any resulting images like static builds.
Static builds do a "COPY" of files into the image before exposing the
directories as volumes. Dynamic builds do a git fetch
at run time and the
resulting data is downloaded to an already existing volume location, which is
now free from Docker versioning. Both methods have their advantages and
disadvantages. Deploying the same exact configuration might benefit from a
single image built statically whereas deploying many different disposable
configurations rapidly are best done dynamically with no building.
To run dynamically:
fig-dynamic.yml
file to point at your own Wheel repository
location by setting the $WHEEL_REPO
variable. When run, the Hub container
will pull the "config" branch of that repository and use it to run the Spoke
container with your own configuration.fig -f fig-dynamic.yml up
MIT
Much thanks to Jérôme Petazzoni for PXE, which this container is mainly based off of.