-
Notifications
You must be signed in to change notification settings - Fork 4.9k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Make the provisioner interface support more container runtimes #10883
Comments
It should be possible to start with a pretty "blank" machine. And the Provisioner installs the container runtime. Currently there are some "grayzones" like CRI* and CNI*. * mostly because Docker didn't use these by default |
This is a prerequisite for removing the runtimes: #9989 It's still possible to provide images with pre-installed runtimes. |
thanks for noting this @afbjorklund I think this is important, since we making containerd as Default as possible. I |
The main downside of this at the moment, is that you have to install and configure your own container runtime. |
this would allow us to install specific version of container-runtime, is there a reason you prioritize this as soon ? how about backlog ? |
one challenge would be how we learn to install Containerd or Docker or Crio on different OS (for example in case of SSH driver) |
Currently Kubernetes defaults to using Docker to install containerd (?), rather than using the CNCF project https://containerd.io/ https://kubernetes.io/docs/setup/production-environment/container-runtimes/#containerd
I started on making an install script for CRI-O, but I'm not sure if it is available from the home page yet ? https://cri-o.io/ https://kubernetes.io/docs/setup/production-environment/container-runtimes/#cri-o
|
We would also need some way to install CRI for Docker. https://github.com/Mirantis/cri-dockerd And to install |
@medyagh : actually I did put this in the backlog The workaround is that the none or generic user would have to their own provisioning of the container runtime packages. This issue is part of the "removing libmachine legacy", where Docker was the only provisioner that could ever be imagined |
This comment was marked as outdated.
This comment was marked as outdated.
This comment was marked as outdated.
This comment was marked as outdated.
Freezing this because it's still relevant |
https://minikube.sigs.k8s.io/docs/drivers/ssh/ Might want to update the docs to say that only Docker container runtime is supported at the moment, and/or throw an error when trying to pass --container-runtime='containerd' with --driver=ssh |
even if the box you are sshing to already has containerd on it, it seems like minikube doesn't support this because it's looking for a docker libmachine profile to provision, and Flatcar Linux is not supported, and the |
It is supposed to work with
the repo didn't get moved anywhere, but the previous Docker Toolbox product got replaced by the Docker Desktop product... |
As far as I know, Flatcar Linux (née CoreOS) should come with the container runtime pre-installed ? So provisioning should be quite fast, basically |
Currently the libmachine interface is quite targetted towards Docker:
It needs to at least know which runtime to install, during provisioning ?
And the swarmOptions and engineOptions need a generic replacement...
Currently the provisioner in minikube is broken, in that it doesn't provision.
It just assumes that all container runtimes are always present on the node.
The original provisioner would install Docker, if it was not already available:
The text was updated successfully, but these errors were encountered: