EPEL Distribution Statistics

In 2022, the EPEL team published statistics detailing Enterprise Linux distribution usage based on what they were seeing in their countme data. The intent was to provide some picture regarding EL Distribution usage (aka uptake in the RHEL clones). I thought it would be interesting to look at this a year later, to see if any trends or stats have changed.

At the time, Rocky and Alma were still building their infrastructure to be able to provide builds of RHEL 9. Today, there is a lot of focus on what the future holds for Alma and Rocky with Red Hat changing the way they distribute SRPMs. The intent of this post is not to dive into the reasons of that decision, as discussion around that and Red Hat's subsequent response, can be found elsewhere. Nor is it the intent of this post to dig into the merits of the Rocky Linux response to continue with 1:1 RHEL compatibility or the AlmaLinux decision to no longer be 1:1 RHEL bug compatible.

Disclaimer

There is a disclaimer around this data: it is only using data from the Fedora Project countme database. There are likely large populations of installations that are not being counted in this which could heavily change how the data looks. Systems that are not using the Fedora EPEL DNF metalink, a system that is using a local mirror, or a system that has disabled countme will not be counted in this data. Additionally, if a system is not configured to use EPEL, then it will obviously not be counted here.

Data as of 2023-07-25

Link

Takeaways

Rocky Linux is still on top.
CentOS (non-Stream) usage has fallen. This is likely due to the retirement of CentOS 8, and the upcoming EOL for CentOS 7.
RHEL usage is continuing to grow at a steady pace, keeping up with the use of EL in general.
AlmaLinux has overtaken CentOS Stream. CentOS Stream is now the the second least used (by a small margin) EL distribution, with Oracle Linux on the bottom.

Looking Ahead

I am genuinely curious how these charts will look 3-6 months from now, with the source code changes announced by Red Hat. My intuition tells me it will not change trends much, as most folks that are vocal about the Red Hat changes are running a personal system, a homelab, or a small deployment ( <100 servers). Most businesses that have large fleets of servers will likely not move or be concerned by the source code changes. It is also pretty much a certainty that the large server fleets are underrepresented in this data, because they are the most likely to be using a local mirror, or not using EPEL at all.