Fedora publishes countme data weekly into a database that can be picked up from here. For the week of 2025-02-02 data it looks like our FCOS node count has increased to over 105k (36.2k transient, 68.9k long running):
NOTE: These are only nodes with countme left enabled and able to reach the internet. There are probably more nodes out there.
Our architecture breakdown is about 47.8%aarch64 and 52.2%x86_64. This is a continuation of strong growth for the aarch64 architecture. In October 2024 it was previously at 41.1%.
Our breakdown based on Fedora Linux release shows the majority of users still on Fedora Linux 40 so that may be indicative of unreported issues where people aren’t able to move to Fedora 41 for some reason. Please do consider re-enabling zincati to get your systems up to date if you’ve disabled it for some reason.
This time I also wanted to see how CoreOS is comparing to other similar variants of Fedora. Here are a few graphs showing static/transient/total node count for Fedora Cloud/CoreOS/Server:
These graphs show a little bit of what I would expect. Over time Server has had more static nodes than transient nodes and Cloud has had more transient nodes than static nodes, which is understandable considering the target use cases of those variants.
Either way it is apparent CoreOS is showing steady growth, and has finally reached a point where it is battling with Server and Cloud over the top node count for any given week.
Thank you to everyone who helped us get to this point. Now let’s hit 200k!
P.S. We do like hearing from you. If you use Fedora CoreOS (or have tried it in the past) let us know what you like and what can be improved!
Thank you for this! I’m dying to switch from Fedora Server to CoreOS on my Hetzner boxes but always get stuck somewhere between Teraform, Butane, Ignition files, and copying them back and forth…
Our breakdown based on Fedora Linux release shows the majority of users still on Fedora Linux 40 so that may be indicative of unreported issues where people aren’t able to move to Fedora 41 for some reason. Please do consider re-enabling zincati to get your systems up to date if you’ve disabled it for some reason.
I’ve actually begun migrating back to Fedora because of it. I realized that although I like the automatic updates, that wasn’t enough for me to keep using FCOS. Fedora can have automatic updates too, via dnf-automatic, and I have a lot more control over it than I do of Zincati. I never did manage to get Fleetlock to work how I wanted, so although my machines only rebooted one-at-a-time, they would do so at random points in the day. The also wouldn’t always finish draining their pods before doing so, causing fairly large downtimes for some applications.
For my non-Kubernetes workloads, I also realized FCOS is quite difficult to manage. No Python means no Ansible, so I was stuck trying to distribute configuration and secrets in various other ways. On a Raspberry Pi, rpm-ostree is so slow that installing Python on a FCOS machine takes longer than installing an entire fresh Fedora OS via PXE. In the end, again, I decided what FCOS offers just doesn’t align with my needs, and have been moving these machines back to Fedora as well.
Nevertheless, I’m happy to see your project is going well, and I wish you all the best.
Ouch. Sorry to hear and thank you for sharing your experience.
Were you by chance running a portion of your fleet on any of our preview streams? Running on testing and next can help you know when breakage is coming and report it prior to it reaching stable for everyone.
I actually don’t have any numbers related to that, but it would be interesting to try to pull some data from GitHub and compare it to known RH employees versus extrernal contributors. That’s a good idea.
