Years ago, when I asked a senior admin what I should install to learn enterprise Linux, he handed me a CD and walked away. I booted it, configured a few services, and built my whole mental model around the assumption that this distro, CentOS, was permanent. It was not. In late 2020 Red Hat killed CentOS as a downstream rebuild of its commercial product, and a generation of small shops and self-learners lost both their practice lab and their production platform overnight. Six years later, the “free enterprise Linux” space has split wide open, and picking a replacement is more work than the original CentOS choice ever was.
This is the story of what CentOS actually was, why Red Hat pulled the rug, and what replaced it. If you study for the RHCSA, run a homelab, or want to know which distro the next interview will assume you know, you are living inside this mess.
The CentOS that almost everyone met first
The name CentOS meant “Community Enterprise Operating System.” Behind the marketing it was a recompile of Red Hat Enterprise Linux with the trademarks stripped. The names and logos stayed Red Hat’s, but the source packages were open, which meant a community rebuild was legal and faithful. A sysadmin running CentOS got the same binaries as someone paying Red Hat, only a few days later.
That parity mattered because RHEL was the working definition of serious Linux in the data center. Banks, hospitals, governments, and large research institutions all paid annual subscriptions for the right to run it, because they wanted a vendor legally on the hook for security backports and multi-year stability. The sticker price was painful, often a four-figure line item per server per year.
CentOS turned that line item into zero. Three groups depended on it heavily:
- Small and mid-sized businesses running their own server rooms, who wanted Red Hat behavior without the invoice.
- Aspiring admins building hands-on RHEL fluency because the job market rewarded it.
- Scientific outfits like CERN and Fermilab, whose workloads needed RHEL-class stability but whose budgets could not survive a recurring license. The Scientific Linux collaboration, in fact, was itself a CentOS spin-off.
Red Hat tolerated the free rebuild for about ten years. In 2014 they quietly absorbed CentOS into the company, taking direct control of the rebuild rather than letting it stay a pure community project. At the time I read that as goodwill. Looking back, it was a strategic move that paid off six years later.
The 2020 reshuffle that broke everyone’s mental model
In December 2020, Red Hat told the community that CentOS 8 would stop getting updates early. CentOS proper would no longer mirror RHEL point for point. In its place would sit CentOS Stream, a rolling preview branch that lives upstream of RHEL instead of below it. The terminology sounds like a small reorg. The reality was not.
The old pipeline looked like this:
- Red Hat built RHEL from internal sources.
- CentOS rebuilt RHEL a few days later from the same sources.
- You ran CentOS and saw what paid customers saw, just slightly behind.
The new pipeline inverts the relationship:
- CentOS Stream becomes the public staging branch where Red Hat parks candidate packages for the next RHEL release.
- CentOS proper stops being a free rebuild of a stable product. It becomes a moving branch ahead of RHEL.
- Production environments that relied on perfect parity lose their floor.
Anyone with automation pinned to CentOS package versions had to audit every assumption. Anyone running CentOS in production got a forced migration window. Anyone using CentOS as a study platform had to accept the textbook changing underneath them.
IBM had bought Red Hat in 2019, and the calculus for a publicly traded parent company was obvious. Free rebuilds of a flagship product are hard to defend when the board wants recurring revenue. Whether that was the whole motivation or just part of it, the community response was fast and loud.
What filled the gap
Within weeks of the announcement, two replacement rebuilds surfaced. Both promised drop-in RHEL replacements, both took their names from people who built CentOS, and both are still maintained today.
- AlmaLinux came out of CloudLinux, a hosting-focused company with operational experience at scale. The pitch was binary parity with RHEL, predictable releases, and a commercial sponsor with skin in the game.
- Rocky Linux came from Gregory Kurtzer, one of CentOS’s original creators. The pitch was the same: bug-for-bug parity with RHEL, community-governed, free, a continuation of what CentOS used to be.
A third option had been quietly available the whole time: Oracle Linux. Oracle had been rebuilding RHEL with its own kernel patches and shipping it for free for years. For some shops that made migration easy, since the option already existed. For others, betting infrastructure on Oracle meant betting on Oracle the vendor, which carried its own weight.
Beyond those three, the long tail is real. SUSE’s free developer program and openSUSE Leap fit people who did not need Red Hat specifically. Debian and Ubuntu Server LTS, with their decade-long support windows and trademark-free governance, became a much more obvious answer for shops whose workloads never actually needed RHEL in the first place.
The net result is fragmentation. The one question CentOS used to settle, “where do I get RHEL-class behavior without paying Red Hat,” now has at least four real answers in play. Each comes with a slightly different release cadence, governance model, and vendor relationship, and none of them are bad enough to disqualify on their own.
How I would actually pick one today
I run a homelab, do occasional contract sysadmin work, and I have moved a couple of small production clusters off CentOS 8 since the change. This is the lens I actually use.
For learning and homelab use, the natural starting place is either Rocky Linux or AlmaLinux. Either one will give you hands-on time with the RHEL family and prepare you for the RHCSA exam. Pick the one whose documentation reads better on day one. They are close enough that the switching cost later is annoying but not catastrophic.
If you are running CentOS in production at a small business, kick the tires on both rebuilds on a non-critical workload for a full release cycle before committing. Both projects ship convert-from-CentOS scripts that work well. Run those scripts on a clone of a real server, never on the production box itself, and validate every service afterwards.
When a workload genuinely needs Oracle, whether through compliance or through existing tooling, Oracle Linux works. The free tier is genuinely usable. Just understand the support implications before you stake a production workload on a free Oracle Linux tier.
For everyone else, the honest question is whether you actually needed RHEL in the first place. Debian and Ubuntu Server LTS will give you a decade of updates with less drama and no trademark concerns. If your workload does not care about RHEL specifically, the free enterprise Linux question is the wrong question.
Trade-offs to name out loud
A few things deserve a clear sentence. Both rebuilds depend on Red Hat continuing to publish source under permissive terms. If Red Hat tightens that further, both projects carry the same upstream risk. Oracle Linux is free at the download, but Oracle as a vendor brings its own kind of operational baggage, and lock-in cuts both directions. CentOS Stream is fine as a development branch, but I would not stake a production workload on a rolling-upstream release. And Debian and Ubuntu are excellent servers, but they do not teach you RHEL, so if your next job expects RHCSA fluency, study on Alma or Rocky.
- Both rebuilds share an upstream dependency on Red Hat’s source publication policy.
- Oracle Linux is free, but the vendor carries its own complications.
- CentOS Stream is a rolling-upstream release, not a stable production target.
- Debian and Ubuntu give long support windows but do not teach RHEL.
- The community is now split across at least four serious options with no clear winner.
Bottom line
The fragmentation is annoying, but it is also a sign that the community reacted fast and built real replacements. You do not have to love this state of the world. You just have to make a decision and stick to it.
If you are starting fresh this week, my advice is to skip the rest of the comparison articles and stand up a fresh VM. Whichever of AlmaLinux or Rocky Linux has the friendlier docs at the moment you need them is the right first pick. Install it, run a small service on it for a month, and treat the hands-on month as the actual test. Once you have committed, run the platform the way you would run any production box: backups that you have restored from, monitoring that pages you, patch cadence that does not surprise you. The free enterprise Linux story is not over. It is just harder to spell, and that is fine.