The future of the CentOS Project is CentOS Stream, and over the next year we’ll be shifting focus from CentOS Linux, the rebuild of Red Hat Enterprise Linux (RHEL), to CentOS Stream, which tracks just ahead of a current RHEL release. CentOS Linux 8, as a rebuild of RHEL 8, will end at the end of 2021. CentOS Stream continues after that date, serving as the upstream (development) branch of Red Hat Enterprise Linux.
A lot of people are not going to be happy with this announcement, and it seems this is the first clear insight into what IBM is planning to do with the Red Hat acquisition. Expect a new CentOS to rise to the occasion and takes its place.
RH said this was the plan when they released CentOS 8. It’s kind of surprising they didn’t wait until RHEL/CentOS 9 to do this, but CentOS Stream being RHEL v.n+1 as the primary product going forward was well known.
Nothing at all to do with IBM, and everything to do with the Fedora community prioritizing features over bug fixes. Fedora is only tangentially related to RHEL, and that was a problem for RH as RHEL didn’t have an upstream project. There wasn’t a community which gave RH feedback on the direction of RHEL, and there wasn’t a community to build things on RHEL, which is why all of the cloud stuff was built on Ubuntu. 15 years later, RHEL once again has an upstream project, and RH once again has a beta distro.
It’s kind of cool RH is going to produce a rolling distro.
There is also this comment on the CentOS blog: https://blog.centos.org/2020/12/future-is-centos-stream/#comment-183642
Flatland_Spider,
Personally I’m always very cautious when it comes to corporate buyouts like this, IMHO it can go badly for the community. Although I don’t follow red hat very closely so I don’t really have enough information to form an opinion about it one way or another in this case.
I think there’s plenty of merit in a rolling release too. All else being equal, sometimes I prefer rolling releases. A lot of users expressed skepticism in those comments. Many are already happy with centos as it was, and I understand their point of view. That’s interesting they’re discussing a new fork…honestly I’m not sure another fork is needed but who knows, if there’s enough demand for it maybe it will happen.
So am I.
I spend my days in the RH ecosystem, so I need to keep up on what’s going on with them in order to plan.
I would prefer a rolling release. I think Canonicial missed an opportunity when they based Ubuntu on Debian instead of Arch or Gentoo.
Now that I think about it, I am curious about how they are going to pull that off. Fedora has rolling release like capabilities with their system upgrade plugin, and I’m curious if that is how RH is going to do a rolling release. Then again, maybe it’s more like SuSE Tumbleweed which is a true rolling release.
There is a market for a free version of RHEL. HPC runs on CentOS, and many scientific packages support RHEL/CentOS.
An other fork is probably needed to serve the RH ecosystem.
Yes RHEL needed something like CentOS Stream. But the users of CentOS do not, they want RHEL, but free. I’m not sure what users of CentOs are going to do. I think there is going to be a defection, but I don’t know where they will go. maybe Open SUSE, which has 36 months of support? Ubuntu LTS only has two but is used more.
My guess would be OEL. I’ve come across a lot of people moving there from RHEL because you only need to pay for support for the systems that you actually want support on, instead of having to pay for every single system you run the OS on.
Of course, it will be interesting to see what Oracle does as a result of this (OEL is largely just CentOS), as well as what Amazon may do with their Amazon Linux offering.
The bigger killer here though is the blatant destruction of trust by changing the EOL date on a product by 8 years with only 1 year of notice. A much better way to handle this would have been announcing CentOS Stream for GA, and then just not releasing a CentOS 9 when that would come out.
Agreed, OEL makes a lot of sense, my understanding is that their upstream was RHEL itself and not CentOS. So if that’s the case it should be fine.
I agree. Dropping C8 like that was a bad look, and they should have waited to kill it at RHEL9 like everyone was expecting.
Eh? A new LTS comes out every 2 years, but is supported for 5 years. But that being said, my vote also goes to openSUSE Leap, which is what I’ve been running since about 2016 on a server that I don’t want to change much.
ArsTechnica did a round-up of RHEL rebuilds.
I think I’m neutral on this, some will claim it’s IBM using the community as a guinea pig, but the truth is most of the people using CentOS or Fedora are there because they want the latest and greatest not the LTS version.
You certainly can’t have it both ways!
Ironically everyone I know using CentOS was using it for the LTS aspect of it. When I switched over to using Debian/Ubuntu server the “shorter LTS cycle” was one of the things my CentOS friends would hold up as a reason not to do it.
I think that was true in the old days, maybe when we first started using CentOS, but I think that changed in the last few years as distro priorities wax and wane.
For Fedora maybe the reverse applies, it was clearly more bleeding edge when it was launched and has now morphed into something more stable, but it’s still relatively fresh compared to it’s siblings.
Nope.
CentOS is a bug for bug rebuild of RHEL, and that means a glacial pace of change exactly as it’s always been. RHEL has worked to address the critique about glacial change with AppStreams, and the community has setup various repos to add updated packages to RHEL/CentOS. However, stability is priority number 1 for CentOS proper.
Fedora is just as bleeding edge as it’s always been. RH is just incredible at release engineering. There is a lot of package churn, and packages will get new versions as soon as the package maintainer gets around to it.
Statement about Fedora, likely true. Statement about CentOs likely false.
Wasn’t “It’s RHEL but free” the main reason to use CentOS?
Bye Bye CentOS. Hello Scientific Linux. If that doesn’t work out… Springdale(PUIAS), or Oracle Linux. I sure hope there is a legit successor to CentOS old school. Wait.. there’s Rocky Linux!
Well not official yet, but let’s see what happens: https://github.com/hpcng/rocky
Scientific has been dead for some time. They switched to CentOS for CERN a while ago.
Thanks Bill,
I believe there is a need for a free EL solution. Maybe this is the death knell for Redhat sadly as with the demise of SUN (x86 and Opensolaris), I believe the community needs an ecosystem that mimics the enterprises that they work for. Rolling releases dont fit the bill.
This is the death of Red Hat Linux and the birth of Fedora Core all over again. No more free enterprise version, instead a rolling distro that most old users of the discontinued product probably don’t want.
Bill Shooter of Bul,
Did you mean the death of centos? It seems to me that centos is mainly used by people who wanted/need red hat but don’t want to pay for it, and for this group a centos clone is a compelling option. Aside from the financial barrier I think it’s safe to conclude that 100% of centos users could be served by red hat.
I read this from CentOS FAQ.
https://centos.org/distro-faq/
I find this argument really odd and non-nonsensical. Obviously it’s far less work to use redhat’s packages as they’ve been doing all these years. Automation (with some oversight) is how centos has managed to get such quick releases of redhat updates for all these years. If anything leading redhat is going to be more work. I do wonder the motivation behind this change. Most centos users probably don’t want this. Users who want to lead redhat already have fedora. So I have a hard time figuring out who is asking for this and I think it may be redhat (aka ibm) influencing the project to suit it’s own interests.
https://itsfoss.com/centos-stream-fiasco/
“Rocky Linux” could gain steam very quickly among disenfranchised CentOS users much like how Linux Mint grew very quickly from disenfranchised Ubuntu users.
https://news.itsfoss.com/rocky-linux-announcement/
No, i was referring to the history of Red hat. In the beginning there was Red hat Linux and it was good. Then they introduced Red Hat Enterprise Linux, that was more focused on long term support and stability. Then they killed off Red Hat Linux and transformed it into Fedora Core, which later dropped the “core” off the name.
But really upon reflection, its a revert of that history. Centos Stream IS the spiritual successor to the old Red hat Linux. At the time of the killing of Red Hat Linux, I swore off all Red hat products I thought Fedora would be too immature for my uses and would always prioritize instability with half baked projects. So this is me confronting my younger self and asking if I’m happy now. Honestly, I’m not sure. Maybe? I like the idea of CentOS stream, but I’m not happy about the loss of the previous mission of CentOS. They server two different purposes, IMHO.
Bill Shooter of Bul,
Ok, I got it. I’m not really sure that red hat cares any more what centos users think. They’re probably interested in bumping users off of centos or at least putting them to better use on a testing track. I can’t think why else centos would be motivated to take this path independently.
Indeed. I’ve mostly remained on the deb-buntu side of things. I’m ok with rolling releases, but I’d probably be frustrated by this too – the whole point was to follow red hat.
While red hat has control over the centos board, It’s not hard to imagine the industry moving very quickly to a new clone that once again provides binary compatibility with RHEL. I’m getting ahead of myself here but it makes me wonder if red hat would try to sabotage other efforts to clone red hat.
Yeah I mean, I swore off Redhat after they killed Red hat linux and went to fedora/RHEL. But I came back, because it made sense and Fedora turned out to be stable enough and fast moving enough to make sense for workstation use, while RHEL/CentOS had verified non upstream drivers for odd ball hardware I needed on servers and I was sick of porting driver code between kernel versions.
RH doesn’t care about community efforts to rebuild RHEL from source, but they get a little miffed when other corps create RHEL clones.
RH started obfuscating patches with RHEL7 in response to Oracle creating their RHEL clone. This caused the infamous months long delay in C7’s release while they reverse engineered RH’s efforts.
It’s perfectly inline with CentOS’s new mission. Every RH project has an upstream FOSS project. For instance, JBoss is the downstream of WildFly. RH devs work on WildFly, and JBoss is the stable version people pay for. There is a direct correlation between the two.
Until CentOS Stream, RHEL didn’t have an upstream. Sort of in Fedora, but the gap between the two is enormous. Packages in Fedora do not trickle down to RHEL, and the Fedora community is more focused on pushing out new versions to get tested for the upstream projects. This doesn’t work for RH because they need a place to stabilize packages for RHEL and shake out bugs, and RH needs a project they can work on in the open which correlates directly to RHEL. Thus, CentOS Stream is where all the development work for RHEL is going to take place.
It’s also very savvy of RH to separate the upstream project name from the paid project. This is a problem Canonicial has. Ubuntu is synonymous with free, and people don’t pay for it. RHEL is synonymous with paid support, and CentOS is synonymous with the free less stable version.
Flatland_Spider,
Well yes it obviously makes sense for red hat, but I’m talking about the motivation for centos itself. From the looks of it, they did not consult with users or consider their needs regarding this new mission and discontinuing the previous mission.
https://wiki.centos.org/FAQ/General
So I’m not trying to argue that there isn’t merit in having the latest features for some kinds of users, but I am saying that there’s a huge base of centos users in enterprise for which this new mission is both a non-goal and discontinuing traditional centos is a legitimate con. Remember centos probably has millions of deployments in an enterprise setting. Of course I could be wrong, but It’s just really hard to understand the motives here unless it’s red hat’s (and IBM’s) pulling the strings – that make clear sense.
I agree, from red hat’s perspective it is a problem to have a free clone that’s virtually identical to your product (albeit minus “support”). They obviously would rather enterprise customers pay for RHEL licenses than use centos for free. On the one hand there’s an awful lot of entitlement on the part of freeloading users/companies that expect it all for free, but on the other hand the FOSS licenses that covers all of this software explicitly gives users this right. Over the years I’ve found it extremely difficult to reconcile these views.
I don’t think they did. Everyone expected this to happen with RHEL9, but no one saw then shuttering CentOS 8 so soon.
It makes sense since C8 is redundant, and it was probably going to be neglected. Pulling the plug was probably the correct thing to do.
CentOS Stream is like enabling the RHEL testing repos. That’s all. People running CentOS Stream are going to be running one minor version ahead of RHEL. RHEL v.n+1. The current version of RHEL is 8.3, so CentOS Stream is equivalent to RHEL 8.4.
CentOS isn’t going to become Arch. Being one step away from upstream is the job of Fedora.
There is a subgroup within Fedora dedicated to prepping packages for inclusion in RHEL.
The setup looks like this:
Fedora –> Fedora ELN –> CentOS Stream. EPEL –> RHEL
Fedora ELN: https://docs.fedoraproject.org/en-US/eln/
That’s the thing. A**, cash, or grass. No one rides for free.
Beta testing RH’s products is the price of admission. Either put dollars down for the supported, stable version, or pay via sweat equity.
I reconcile my use of FOSS by being a beta tester and doing questionable things. I send telemetry back, and I file reports as things come up.
It’s the least I can do, but it’s more then some.
I’m of many minds of the subject, but I think the answer is in Cloud offerings. Everyone using an Amazon linux AMI in the cloud, is paying for it hopefully Amazon is investing some of that back into the FOSS ecosystem. I pay money for a VM, how much you want to use that money to pay for the OS vs the hardware is out of my hands.
Indeed. Let’s see if the community takes over and CentOS goes off in a wildly divergent direction like Fedora did. LOL
Check the FAQ:
https://centos.org/distro-faq/#q6-will-there-be-separateparallelsimultaneous-streams-for-8-9-10-etc
CentOS Streams is rolling but versioned per major version, what will happen is not that different:
you’ll have an enterprise distro for free, done mostly by engs paid by RH, supported for 5y.
What changes is that you’ll get first features and changes and the lifecycle that is moving from 10y to 5y.
But really, in a world where containers dominate, new releases get out every day, do you really want to stay on the same bits for that long? also 5y is the standard for the community stuff: check debian/ubuntu lts.
With this move CentOS gains dignity like a full original OSS distro, rather than passive rip off of someone else job. Others than that, with rh investment in streams, that will deiver innovation + stability.
Masses don’t read and freak out for a perceived right (have enterprise distro for free? ah lazyboy) rather than understanding that this will provide to centos community more power to influence, while now were the rh customers who were driving the agenda of RHEL and of CentOS as consequence.
Plus, if you really want 10y support, you have to be honest and recognize you need to pay the Vendor that provides you more value, Red Hat or other vendors of your choice.
andreamtp,
Except this has nothing to do with not reading, to the contrary I think almost everyone understands what this means: centos streams is replacing the distro they want to use with another they didn’t ask for. They want CentOS to follow RHEL and I suspect most CentOS users care much more about stability than “influence”, that was a big reason for choosing centos!
I understand your opinion completely, yet at the same time RH’s open source licenses explicitly suggest the opposite is true: you can use the software standalone without paying vendors for any unwanted support. Many users will probably seek another free clone, would you have a problem with that? Is it sustainable? Is it fair? Is it profitable? It’s hard to say because there are rational arguments to be made for everyone’s opinions.
This is true. It’s also true, people using RHEL without support have to manually download the rpms or rebuild them from source in order to get updates.
For most, they don’t want to run RHEL that badly.
I don’t care. Maybe they’ll have some sort of imagination about it this time.
There are a several good candidates.
* Manjaro (Arch based)
* Alpine (Gentoo + musl based)
* NixOS
* Guix
* Slackware
* Deriving something entirely new from Gentoo/Funtoo
Honestly, NixOS or Guix would be a good. I’d like to see them get more traction. NixOS is doing pretty well in that department, and it seems sustainable right now.
Ultimately when people need to get paid for their time, it’s always going to come back to a model where the community version is the development version, and everyone in the community is a beta tester. The “Pro” version is the paid, stable version with long term support.
Some communities handle this better then other, and the CentOS community is mad their free lunch isn’t free anymore.
Flatland_Spider,
Well, that’s where CentOS and possibly “Rocky Linux” come in. Obviously plenty of people and companies are content running alternatives and I don’t think this will change even if CentOS withdraws from it’s current position.
I’m not trying to say that red hat is in the wrong for doing any of this. Realistically though anybody can come and clone the so called “pro” version. That’s not something red hat has much control over. Red hat has the CentOS board at arm’s reach, but if they pull CentOS from it’s core demographic market, they’ll likely end up creating demand for a new distro that red hat has less control over. I’m not saying it’s good or bad, but just pointing out this will probably happen and if it does a lot of current centos users are going to flock to it.
Flatland_Spider,
Actually I think my point may have been misunderstood. I was expressing concern for red hat when people choose to use free clones. Is it sustainable for RH? Is it fair to RH? Is it profitable for RH? And with this clarification it seems like you do care. I was trying to highlight the double edged nature of FOSS.
Yes, and maybe the new community will have some sort of imagination and build something really cool and innovative with a very healthy ecosystem where everyone chips in.
Probably not though. Companies will probably be happy with the free labor, and people will get burnt out then leave the distro to die like countless others.
Or maybe MS will fork it and everyone will get super excited about MS Linux because MS are on the side of good.
I got tired of talking about RH and went off on a tangent dreaming about the CentOS community doing something other then building another RHEL clone. You know, using their energy for good.
I don’t think RH makes as much money off of RHEL as we think it does. Regardless, I want RH to succeed. I like their distros, they pump a lot of money into FOSS, and I am absolutely in awe of their release engineering prowess.
Do I care that much about another RHEL clone? Not that much. I don’t need strict RHEL compliance, and other distros/OSes would suit what I’m doing right now better.
CentOS Stream doesn’t change that much for me, and I’ve been wondering about moving to another distro/OS for a while. There is one piece of software which I need to navigate (FreeIPA) though.
Arch isn’t stable enough, and I’m not sure where Manjaro is going.
Gentoo/Funtoo is probably the easiest choice, but that’s some work to get that going.
FreeBSD needs FreeIPA integration.
This is good information. Thanks for posting!
Supporting CentOS Stream for 5 years lines up RH’s full support lifecycle for RHEL. The second 5 years in the 10 years is maintenance support. 🙂
https://access.redhat.com/product-life-cycles?product=Red%20Hat%20Enterprise%20Linux
https://access.redhat.com/support/policy/updates/errata
Truth!
I’m pretty happy about this. I’ve got to QA stuff a little bit harder, but I was already QAing stuff before releasing to production anyway. That’s kind of what you do when you run without support. XD
My understanding was RH customers were awful about providing feedback. The customers who bought RHEL licenses weren’t the type to run the latest and greatest.
It’s the thing where they wanted a faster horse.
RH: Customers! What do you want?
Customers: RHEL5 supported indefinitely.
RH: o__o
RH: >__>
I was going to give Oracle Linux a try last night, but couldn’t find a way to download it without creating an account. I mean who wants to be contacted by Oracle just to try out their Linux distro…
Oracle’s entire mission is to get people into their sales/lawyer funnel.
See my other comment, which was meant as a reply to yours :), about the download.
Cheers!
You don’t actually need an Oracle account to download OL, I just downloaded it today from here (haven’t tested it yet):
http://yum.oracle.com/oracle-linux-downloads.html
Cheers,
The branding sucks, but the development process makes more sense now. And it’s virtually identical to how FreeBSD does things. 🙂
Fedora (rawhide?) is the development tree. This is never guaranteed to be useful at any specific point in time.
Fedora is branched to create CentOS Stream X. Development continues on Fedora.
CentOS Stream X is stabilised and released as RHEL X.0.
Development continues on CentOS Stream X, is stabilised, and released as RHEL X.1. Then X.2, X.3, and so on.
Fedora is branched to create CentOS Stream Y. Development continues on Fedora.
CentOS Stream Y is stabilised and released as RHEL Y.0.
Development continues on CentOS Stream Y, is stabilised, and released as RHEL Y.1. Then Y.2, Y.3, and so on.
Not all that different from how -CURRENT, X-STABLE, and X-RELEASE works in FreeBSD.
Will be interesting to see how they manage the development split between Fedora (the desktop OS) and Fedora (the RHEL server OS) if they’re going to be using the same upstream repo for development.
Would make things a lot simpler if they hadn’t repurposed the CentOS brand like this. It’s changing it from “free intro to RHEL” to “name for some development repo”. Going with some other kind of hat-related name (Fedora –> –> RedHat) would have been better, IMO. If you’re going to retire the CentOS distro, then retire the brand completely.
Something getting lost in all of this is RHEL majors will be on a 3 year release cadence starting with RHEL8.
RHEL9 is due in 2022.
Linux Unplugged has a good information about what RH was thinking. Starting at ~3:45 they starting talking about CentOS Stream.
https://linuxunplugged.com/383