Doing some work for my upcoming State of Fedora talk, and as a side-effect have some new EPEL graphs. Rough work in progress so not as pretty as could be. But still interesting for #centos #rhel #almalinux #rockylinux (as well as EPEL itself, of course). a with
s
First — RHEL 9 is new, of course, and by definition EL users are conservative. So it's not surprising that EPEL 9 is only a tiny fraction at this point.
Here's the older, more crude IPs-per-day counting, which has the advantage of showing older versions. You can see that while 8 overshadows 9, even-older EL 7 is still dominant and growing.
But, anyway, looking at EL 9 first — here's the breakdown by OS id. (Which I am calling `variant` here because I'm shoving the EPEL stuff through the same code I use for Fedora Linux where we have Fedora Workstation, Fedora Server, etc., and that's just easier.)
But that's not the whole story — here, I've used a heuristic to estimate the number of systems that don't stay around for more than a week. That is, test systems, short-lived containers and cloud instances, CI builders, etc.
And the opposite: systems that are are long-lived, or for new installs are estimated as likely to be. (See caveat in next tweet.)
For the most recent week (and to a lesser extent the week before that), the heuristic is pretty much just a guess if there isn't much historical data. So it's possible that a month from now in retrospect this will show a little different story.
... and the "ephemeral" view — with lingering classic CentOS Linux removed so it doesn't dominate the chart....
I probably won't have time to show all this in my Nest talk, but I'll definitely show similar for Fedora Linux. So if you have gotten this far in this thread, you won't want to miss that. Register at https://hopin.com/events/nest-with-fedora-2022 !
@mattdm Would be nice if EPEL7 would start going down soon
@kwf yeah, I expect that to happen in 2026 or so. (Not joking really.)