cEOS hanging on "A start job is running for EOS Warmup Service" on ARM with qemu #1637
Replies: 1 comment 4 replies
-
|
I think @ankudinov is the best one to provide his experience on that one |
Beta Was this translation helpful? Give feedback.
-
|
Hi @neilschark
|
Beta Was this translation helpful? Give feedback.
-
|
Thanks for your reply. I am aware of the possibilities to have containerlab running with docker-in-docker. I actually contributed parts of the current documentation on this set-up. I am aware, that cEOS is not available on ARM, the question is why it won't work with qemu directly via docker. qemu is capable to run this cEOS image on ARM, but only when a whole x86-VM is run on it, not when I simply run the cEOS docker container via qemu x86. |
Beta Was this translation helpful? Give feedback.
-
|
The main reason is documented here (Apple Silicon section): https://docs.docker.com/desktop/troubleshoot/known-issues/
Lack of |
Beta Was this translation helpful? Give feedback.
-
|
Thanks for your answer. I am currently running a remote dev container on an x86 server when I need cEOS, so that works fine. While I understand why the lack of inotify can be a problem I am still irritated why qemu can start cEOS in an UTM-x86 VM (even if it is really really slow), but simply can't in a qemu x86 docker container. But as the remote dev container setup is a good solution for us for now, we will simply wait for the future and hopefully arm compatible containers. And in the mean time asre using less and less cEOS and more and more linux containers in our labs, as they support ARM natively. For our use-cases cEOS is not really needed. |
Beta Was this translation helpful? Give feedback.
-
Hello,
I am still trying to get our containerlab setup running on my ARM-Mac. I know it's not supported, but I won't give up easy 😉 .
If somebody has some input, I would be gratefull. Furthermore, maybe my finding can help somebody in the future.
I use devcontainer setup based on the setup in the containerlab documentation. It uses docker in docker:
Configuration:
Its working fine on a x86 Linux machine!
On my M1 MacBook it also works, but containerlab fails to start the cEOS image. The lab I am trying to start is totally basic:
It gets stuck on the postdeploy actions.
If I look into the docker logs, I see, that it is stuck on
[ *** ] A start job is running for EOS Warmup Service.Logs:
One thing to note is, that the "loading animation" with the stars is really, really slow. Like one character per minute slow.
The system itself is at high load at the time, the load comes from these three processes inside of the container:
These processes sometimes run in parallel, sometimes just one. It switches back and forth.
I waited a long time and you can see, that the container is leaking memory. It crashed when it reached the limit of the docker VM.
I also tested a different approach. I used a ARM-VM via UTM and used docker with qemu inside of it (so instead of qemu x86 docker in docker in the docker vm it is only qemu x86 docker in an ARM VM) and I get the exact same error. So it seams that something in the x86 qemu translation in the arista commands is having problem.
If I use the provided UTM-x86 VM from the containerlab team, then I can deploy the lab successfully. The problem is, that the x86 qemu VM is so incredible slow, that it is unusable.
What I don't understand is why the problem only occurs, if qemu is used on the docker layer and not if qemu is emulating the whole VM...
Does somebody have an idea?
Beta Was this translation helpful? Give feedback.
All reactions