I'm following the CoreOS Docker Documentation and it mentions starting containers with commands like:
docker run someImageName /bin/somebinary
Where someImageName
is an image. When /bin/somebinary exits, the image will no longer be running.
I would simply like to run an image, without specifying any binaries to run. Instead, I simply want to run the services (eg, systemd / sysvinit) that are normally run inside the images OS.
This seems like the most common thing anyone would ever want to do with Docker, but trying to run an image without a command returns:
2014/02/05 14:49:19 Error: create: No command specified
How can I start a Docker container and run a full OS, rather than specifying a command?
As documented here, you simply run
/sbin/init
as the command just like any other unix booting from single user to multi-user mode.https://stackoverflow.com/questions/19332662/start-full-container-in-docker
Containers can be full blown OS's, they just don't have to be (neither do VMs for that matter, it's just more complicated to configure and manage).
I would say the whole point of Docker is to make application containers easy, so that you only have to configure an app, not the whole OS.
Docker is a system for management and deployment of application containers, not operating system containers. It seems as if you're conflating running a docker container with booting an operating system.
Your Docker containers should be single-purpose, very narrowly-scoped applications that can be started with a single command. If you're looking for something more complex than that, then Docker is not the solution you're looking for. In that case, check out KVM, ESXi, OpenVZ, LXD etc.
If you're just looking for how you can specify a default
CMD
andENTRYPOINT
for your containers, you can do that at build-time using a Dockerfile.To run a full operating system in a container create the following Dockerfile:
Then build and start the container and enter a shell inside it to explore the services running inside it:
Full systemd services inside the container. Beautiful.
Just run from the same image as many times as needed. New containers will be created and they can then be started and stoped each one saving its own configuration. For your convenience would be better to give each of your containers a name with "--name".
F.i:
That's it.
After that you have your containers created forever and you can start and stop them like VMs.
To get in the container and do what you wanna do: