-
Notifications
You must be signed in to change notification settings - Fork 59
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
cannot install "nfs-utils" any longer - conflict with nfs-utils-coreos #572
Comments
We already include |
But I need an NFS server, not only client. |
The error message is not very helpful here I know, we probably want |
You are certainly right. The install of nfs-utils does not even work any longer on instances which I installed two weeks ago at a time where it definitely worked. I have, for example, two instances installed two weeks ago. On one I installed nfs-utils at this time. No OS-updates have been applied since then (I disabled zincati). But I cannot install nfs-utils NOW on the other one of that age where I did not install it yet then. So it must be a problem in the package in the repository online. |
Package layering is subject to these sorts of problems. Layering of arbitrary packages is not recommended. Can you run your NFS server in a container? |
I can try this ... |
Agreed, using containers for servers would be recommended. That said |
@kai-uwe-rommel, for context, Fedora CoreOS is primarily an image-based distribution. We recommend that any software not provided with the OS should be run in containers. #401 discusses providing some mechanism for allowing users to install certain curated packages, but in general, we reserve the right to make changes which might break layered packages. |
Great! Thanks. I'm not that long into FCOS so that I was not aware of that I can "remove" a base package this way. |
@bgilbert You are right. But I think that NFS server should be possible to run natively, not as a container, for performance reasons? |
Linux containers are just Linux processes. If you use a container with the host network namespace, there is (basically) zero overhead to networking for example. |
One of these days I will try to build a suitable slim container. BTW, my use case is to provide a lightweight NFS server for OCP/OKD clusters. (And another such VM with a HAproxy load balancer.) |
See #121 (comment) for an example config to create a containerized NFS server. |
Creating just the systemd unit is trival. But I would not rely on someone else's NFS server container from Docker hub. So the actual work needed (for me) is to create a container that is generic and simple enough. Not rocket science either, agreed. But it is work that could be saved. Just because we can run containers does not mean that it makes sense to run everything in containers. "If all you have is a hammer, everything looks like a nail" ... not. |
related: https://src.fedoraproject.org/rpms/nfs-utils/pull-request/9 -- I think @stevedd's plan is to eventually split nfs into a client and a server package. |
nfs-utils-coreos conflicts with nfs-utils, dependency of targetd. We will run iscsi as part of a container so no need to install it here. Refers coreos/fedora-coreos-tracker#572
nfs-utils-coreos conflicts with nfs-utils, dependency of several packages. Although some packages can be removed in favor of running their services on top of containers (see coreos/fedora-coreos-tracker#572), nfs-utils is still a dependency for other packages we need to install in the host.
The base packaging issue here reported in the original post is detailed in https://src.fedoraproject.org/rpms/nfs-utils/pull-request/9#comment-59201. Nodes shouldn't be stuck on this forever. Practically, this is a perennial issue that'll hit users layering |
This still happen if you install cockpit-fileshare too. |
Can't auto-update FCOS on my raspberry-pi 4B
|
This is not the newest release of FCOS, but it fails with the newest release as well.
I explicitly went back one release because I DEFINITELY was able to do this successfully only a couple of days ago.
Please help, I do need this rather urgently as I we use this as part of a larger cluster deployment and I am basically stalled ... :-/
I don't know a workaround either.
Does this still work for someone else?
(This is in instances deployed with the VMware OVA.)
The text was updated successfully, but these errors were encountered: