ROS Deployment Meeting (#1?)

@ciandonovan and @all

can you share the slide deck for today’s meeting? if possible, that would be really appreciated!

and thank you very much for joining and sharing the information.

I believe that we can have a next meeting when we have any updates or topics that community wants to share.
if you have ideas, problems and experience, please reach out to me, so that we can have discussion for future of ROS deployment and fleet control.



Thanks for the opportunity Tomoya, much appreciated.

Here’s a reproducible source of the slides along with the PDF, in the spirit of the talk, might put it in a Podman container eventually :wink:

thanks again,


I remember there was talk of recording the talk. Was anyone able to record it? and if so could you share it?

@ARK3r sorry there is no recording available on this time, we had some issue on zoom setting so we ended up having the meeting with GoogleMeet that does not allow me to record. my apology.

@ciandonovan Thanks for the interesting presentation!
I was also looking into the best way to setup a development and deploying environment for my projects.

I took, more or less, the same path and was now arriving at docker containers / kubernetes with an ubuntu server minimal installation kind of setup so it was interesting to see your trajectory.

Did you notice perfomance gains when moving over to a Debian based setup?
I’m also running a control loop with EtherCAT slaves so good real-time performance and deterministic access to the NIC is important.

Are there any follow-up actions to be taken after this talk? Like is there a sort of working group that will take this up and try to make a template / best practice out of this?

1 Like

I personally do not have plan for any further concrete actions at this moment, I was thinking that we can have meetup need-to-do basis when we have topics. if anyone is willing to take a lead for community meeting, i am happy to help and join.


what do you think? i would like to see opinions and what community feels about this? such as regular meeting and working group? please feel free to share your thoughts!

Happy Holidays :christmas_tree:


@PieterjanCottignies the kernel with PREEMPT_RT is what will influence your real-time performance, distro is largely irrelevant. However Debian does make installing the RT image as easy as an apt install from the default repos, whereas on Ubuntu you must sign up for Ubuntu Pro and get it through that channel, at least officially.

I’ve found Debian easier to automate its install with netboot/preseed, which AFAIK is no longer an officially supported option with Ubuntu. The absence of NetPlan is also a plus in my book.

However we will be moving to a Yocto/Poky base shortly, and of course since everything important is containerized already it will hopefully be a relatively seamless transition. Some of the major benefits here for us are the ease of deploying custom kernels with patches for better RealSense camera support, building the Google Coral accelerator drivers along with PREEMPT_RT, etc.

1 Like

I’m curious about this. Don’t custom kernels work fine with Ubuntu, too? What makes this easier in Yocto?

The BitBake tooling is really nice, can take a standard Yocto distro kernel, then layer on BSP additions, PREEMPT_RT, patches etc.,


I agree to @ciandonovan . If you are a company that is deploying embedded linux as part of your product, you might as well use one of the embedded linux “meta-distros” out there (at the end of the day, it boils down to buildroot and its flavors (e.g. openwrt) or Openembedded (e.g. yocto)). It gives you versatility you will not find anywhere while being prepared that you are actually building a proper product, that has to comply with OSS licenses (can you extract the source code of all packages including patches with one line of configuration) and quality like atomic updates. Also, you get a system that is configured like you want. It saves the hassle of “yeah, install this distro, then run this huge bunch of scripts and hope for the best that the last upstream update did not break all your custom scripts”.

I know, just my 2 cents.

1 Like

@RethinkMatze I agree, my setup was also first to set it up using Ubuntu to familiarize with ROS 2 and setup a prototype / test. After this the goal was to make a yocto based OS.

I see the embedded working group already has a meta layer for ROS 2 but we probably want to handle this inside docker / podman with the described setup.

@ciandonovan Yes performance is definetly related to the RT-Patch and is looking good. As I had some issue with slave state changes on EtherCAT I wondered if the NIC was more stable / performant with the Debian distro or was it only on WiFi that you had the described issues kernel issues on Ubuntu?

@everyone Maybe an idea to make documentation / describe best practices for deployment based on the talk and this discussion maybe with a basic example setup for future references?

Am willing to help with this but don’t have time enough to take lead or coordinate this.

1 Like

This sounds interesting, I am very new to ROS but I am pretty experienced with K8s, this is pretty cool

This topic was automatically closed 30 days after the last reply. New replies are no longer allowed.