Red Hat defends its CentOS decision, claims Stream version can cover '95% of current user workloads'

Red Hat's Karsten Wade, a Senior Community Architect and member of the CentOS board, has defended the decision to kill off CentOS Linux in favour of CentOS Stream, saying the two projects were "antithetical" and Stream is a satisfactory replacement in most cases. CentOS Linux is downstream of Red Hat Enterprise Linux (RHEL), …

[naive]

The RedHat corporate spinners won't fix this

They should have a law against old dying dinosaurs buying up innovative companies just to take them down in their death struggle.

The news just broke in time for steering away from using CentOS 8 for a major project, using Ubuntu instead.

The world is an ironic place, Open-Source enabled startups like google, facebook and AWS to grow exponentially without going bankrupt on paying MS their CAL's.

Little they return to the Open-Source world, maybe one of them could make good on the false promises of RedHat and maintain a good distro for company use that doesn't cost and arm and a leg for small businesses.

[AMBxx]

Re: The RedHat corporate spinners won't fix this

Couple of general questions:

How easy do you find it to switch between distros?

How different are they in functionality and use?

Genuine question, I work mostly with Windows but have a couple of Ubuntu servers running in Azure.

[Steve Davies 3]

Re: How to switch?

Something like this:-

dnf install centos-release-stream

followed by a

dnf update.

Then your /etc/redhat-release will read: CentOS Stream release 8

Not tried it yet but it isn't ecactly difficult.

[Anonymous Coward]

Re: How to switch?

You missed the "fix anything that broke" step

Soemtimes that takes almost no time, other times you just end up rolling back.

[nematoad]

Re: How to switch?

I think you are misreading AMBxx's question.

By my reading they are asking how to move away from the RH/Centos world and move to a totally new distro.

From my limited experience of running different distros I would say that it can make a lot of difference. The use of .deb or .rpm for a start. Then there are the various inits available. Do you want systemd, SysVinit, OpenRC and so on? Add in a difference in file systems not to mention the Xorg/Wayland choice and you can soon find yourself struggling to make sense of it all.

A lot of due diligence and research would be needed to make sure that the transitions is as painless as possible.

This is not something to be undertaken lightly.

[lybad]

Re: How to switch?

I agree - we're mainly a Windows and RHEL shop - with some CentOS boxes, which are being moved to RHEL. However, we do have some Ubuntu machines used in some labs - trying to learn the subtle differences when setting them up is not simple.

We had the same type of thing when VMWare stopped including SLES with the licence and we started moving to RHEL based setup.

[sansva]

Re: How to switch?

"The subtle differences" LOL A good Linux admin will easily understand the nuances between different distros. Many of the comments to this article really reveal a lot about the "market" for CentOS.

[Anonymous Coward]

Re: How to switch?

Yes, the subtle differences - where I work, IT is hugely under-resourced, and there are only a couple of people with a reasonable amount of linux sysadmin knowledge. Technically, there is nobody in the department who is a sysadmin for any platform as part of their job description.

Then there's the "engineer" responsible for linux, who doesn't really care beyond their high performance cluster, and the people who deploy machines, who are expected to have knowledge of so many different things and so overworked, they can't get in depth knowledge, and get into trouble for going off script.

I agree a linux sysadmin can cope - but not everywhere has a structure which includes a sysadmin.

[NogginTheNog]

Re: How to switch?

Ah sneering at those less experienced than yourself. Classy.

[John Robson]

Re: How to switch?

And for those of us who use the systems daily, but actually only really use the system that we have to support on a daily basis?

How many CentOS systems do I touch on a typical day? Anywhere from a handful to hundreds.

How many Debian derived systems to I touch on a typical day? The two that I run personally, although I don't often touch them - they just work - and one of those is about to migrate to Slackware (well, unRaid at any rate). Next time I need to clear out my VPS and start again I was going to put CentOS on it, just because that's what I currently use professionally, but docker has made that task significantly less frequent than it used to be.

I moved it from Fedora to Debian matching a change in my professional life (I changed jobs). With most of it's actual work (and most of my day to day work) being dockerised... there isn't much complexity that I need to deal with in terms of the distro specific stuff.

[roytrubshaw]

Re: How to switch?

I think there needs to be a "distro sync" in there somewhere.

I was intrigued by the fact that when I did that it downgraded my Apache installation - weird given that stream is suppose to be "upstream".

[LosD]

Re: The RedHat corporate spinners won't fix this

90% of the difference on a server is the package manager (yum vs apt), the rest is rather minor. A few differences in default packages, and a few different configuration helper programs. Maybe also a few differences in how they structure larger service configurations, like the nginx folder.

Switching difficulty is down to how you use it. If you're just using e.g. a web server with PHP, it's pretty simple, especially if you have no special configurations. Just install the new server and copy over your stuff (this is definitely the best way for cloud services, if you're looking to do auto instantiations, where setting up a new server must be scripted).

Perhaps unsurprisingly, the more complicated your setup and customizations, the more trouble changing is, and the more you will realise the differences.

[DrewWyatt]

Re: The RedHat corporate spinners won't fix this

I have changed between distro's more often than I want to count. I have used Ubuntu, SUSE, Redhat, Centos, Arch, Gentoo, Debian, Devuan, YellowDog and more, along with HP-UX, Solaris and AIX. Mainly because the distribution was chosen for the task at hand, or was vendor supplied.

I look at the different distro's as different accents. Cockney and Glaswegian are two very different accents, but they use the same underlying language. If you moved from Glasgow to London, you may have some initial difficulties with unfamiliar expressions, but you will quickly become used to the local lingo. Same when swapping distro. It doesn't take long to find your feet.

[AMBxx]

Re: The RedHat corporate spinners won't fix this

Thanks all for the replies. One of the joys of running Ubuntu in Azure is that it's SSH only. None of the religous wars about which UI to install!

[Marcelo Rodrigues]

Re: The RedHat corporate spinners won't fix this

"Thanks all for the replies. One of the joys of running Ubuntu in Azure is that it's SSH only. None of the religous wars about which UI to install!"

But the UI is completely optional. All Distros (all normal ones?) can be installed headless (without UI), and managed by SSH.

[Dazed and Confused]

Re: The RedHat corporate spinners won't fix this

When I make server VMs I don't even bother giving them a virtual graphics card. When it comes to Linux servers and graphics heads we really are into fish and bicycles territory.

[David 132]

Re: The RedHat corporate spinners won't fix this

One of the joys of running Ubuntu in Azure is that it's SSH only. None of the religous wars about which UI to install!

There were *nix religious wars long before UIs came on the scene.

Just as long as in your SSH sessions you're running the One True Editor, Emacs, and not that benighted tool of the godless heathens, Vi...

(Trollface icon denotes I should not be taken seriously. Merry Christmas, peace and goodwill to all, even Vi users)

[Dazed and Confused]

Re: The RedHat corporate spinners won't fix this

Just as long as in your SSH sessions you're running the One True Editor, Emacs, and not that benighted tool of the godless heathens, Vi...

and don't forget the war over whether sizes (du, ls -s) should be expressed in 1K or 512byte units. Or whether the output of df should give you a percentage or not.

And Merry Christmas and a Happy New Year even to those poor lost souls of the Eight Megabytes and Constantly Swapping camp :-)

(PS, if the vulture can have a Christmas hat can the troll either have one too, or a humbug?)

[David 132]

Re: The RedHat corporate spinners won't fix this

Wandering a little off-topic here, but I too have often thought that the Reg icons should get modified for the Christmas season, much as the VLC icon is.

[Doctor Syntax]

Re: The RedHat corporate spinners won't fix this

"I have changed between distro's more often than I want to count. "

That may be easier if you're predominantly a Unix/Linux user because then the similarities outweigh the differences. For a mostly Windows user it would probably be easier to move from Ubuntu to Debian than to something Red Hat based. Very likely Centos/RH users would take a move to Ubuntu in their stride although if I were in that position I'd probably think more in terms of Debian.

[lybad]

Re: The RedHat corporate spinners won't fix this

It depends - the powers that be moved us to RHEL, as they allegedly wanted a support package in place. But suppliers of support need to be on an approved list before they can used.

[Morten Bjoernsvik]

Re: The RedHat corporate spinners won't fix this

Problem with stream is that it is always upfront. It is like Fedora which I always hated for all the new stuff breaking backward compliance. I just want things to be compliant and work with monitoring software and sysadmin procedures already in place.

[alain williams]

Re: The RedHat corporate spinners won't fix this

Perhaps unsurprisingly, the more complicated your setup and customizations, the more trouble changing is, and the more you will realise the differences.

This is the key and this is why I liked CentOS - I put a lot of work into getting the machine exactly as I want it and, by and large, not have do much sysadmin for 9 years (other than installing completely new stuff).

I have upgraded from CentOS 6 to Debian 10 on my desktop ** most of it was not a problem, some details were hard; over and above running s/ware a decade newer.

** To get a decent desktop manager - which Gnome 3 is not.

[amacater]

Re: The RedHat corporate spinners won't fix this

Do a text only install - deselect / untick Debian default environment. Use tasksel to install whichever environment you want - Debian provides seven or eight. Ubuntu nromally provides one per variant.

If you _really_ want to, it's possible to have multiple desktop environments and switch between them at login - you will have to pick one login manager, however.

CentOS 6 to Debian 10 is quite a jump :)

[RegGuy1]

** To get a decent desktop manager - which Gnome 3 is not.

Gnome 3 is shit.

Hot corners? No thanks

Auto maximize? No thanks

Top bar? No thanks

That stupid thing on the side? No thanks

...

[DCFusor]

Re: ** To get a decent desktop manager - which Gnome 3 is not.

Mint Mate works for me. And the "real" one, not the bastardized one you can put on ubuntu.

Simple, doesn't get in the way...little to no BS (and yes, I put the bar on the bottom).

[JAB van Ree]

Re: The RedHat corporate spinners won't fix this

Linux is Linux ; switching between CentOS and pretty much every other Linux usually only directly affects the deploy / management mechanisms. Most of the major packages are available for all distro's , if not it's usually easy for a competent IT department to fill the void.

CentOS, like RHEL and Fedora use dnf with RPM for package management

Debian with Ubuntu and other clones use apt with dpkg for package management

If you're a big RHEL/CentOS shop you might have a spacewalk / satellite management server, I don't know of a full equivalent for Debian Ubuntu

[Dazed and Confused]

Re: The RedHat corporate spinners won't fix this

witching between CentOS and pretty much every other Linux usually only directly affects the deploy / management mechanisms.

The way they manage the networks is all totally different too.

Even with moving to NetworkDamager (coz even systemd needs something to look down on) they use different file formats.

[Blake St. Claire]

Re: The RedHat corporate spinners won't fix this

Linux is Linux ; switching between CentOS and pretty much every other Linux usually only directly affects the deploy / management mechanisms.

Said no developer, ever.

Packaging on Debian/Ubuntu is a nightmare too.

[chasil]

RHEL switch

I have only done this once, with RHEL v4 onto Oracle Linux. The switch just changed my update provider, and Oracle supported the previous OS packages that had been shipped by Red Hat.

Red Hat now has a script to convert an installed CentOS or Oracle Linux to RHEL; previously a wipe and reinstall was required ("have fun reinstalling your system" is still on Oracle's CentOS site). The description looks much more thorough in replacing all possible packages with Red Hat versions:

https://access.redhat.com/articles/2360841

I don't know of any other current repo conversion options for a CentOS install, but some will likely emerge (Rocky and Scientific Linux come to mind).

Oracle's CentOS notice, with (some) recent updates:

https://linux.oracle.com/switch/centos/

New CentOS conversion blog entry:

https://blogs.oracle.com/linux/reasons-for-switching-centos-to-oracle-linux

New repo conversion script, supposedly works with CentOS 8:

https://github.com/oracle/centos2ol

After conversion from CentOS to Oracle Linux, free updates are available indefinitely. For those who desire paid support, there are several (yearly) price tiers ($119; $499; $1,199; $1,399; $2,299). I will say that I've had more problems with Oracle support this year than ever before.

For those uncomfortable with "free" products from Oracle, bear this in mind: in the last two decades, Red Hat has terminated two Linux distributions, while Oracle has terminated none.

[The Brave Sir Robin]

Re: The RedHat corporate spinners won't fix this

A few subtle differences that any competent Linux admin will be able to sort out fairly easily. I've been using CentOS for many years but we're moving our estate to Ubuntu 20.04 LTS from CentOS 7 rather than moving to CentOS Stream.

We use Ubuntu on the desktop and with WSL anyway so changing Ansible scripts to rock up Ubuntu servers shouldn't be too big a job.

[mistergoodbytes1]

Re: How to Switch

There is no installer which will upgrade CentOS to Ubuntu. That said, it's not so difficult. Back up everything before you begin, take the time to set up a similar configuration regarding partitions, and most of your content can be put into Ubuntu. Some adaptation of scripts, etc. will be needed but not so much as you might think. Also note that there is an Ubuntu flavor MATE-Ubuntu which installs MATE instead of Gnome 3 as a default GUI.

Having said that...

If you want more of a pure Linux installation you might want to try plain vanilla Debian rather than Ubuntu. Ubuntu has name recognition and some end user oriented bells and whistles but all things considered, I lean toward pure Debian the same way that CentOS 5 and 6 were pure RHEL.

Having run CentOS 5 and 6, I was so disappointed that CentOS opted for Gnome 3 in 7. That made me bail, but it was their choice to make. I ran CentOS because is did give so much control to the operator, and Gnome 3 kind of ended that.

I will say that CentOS provided plenty of input for RHEL as far as forward looking improvements to the OS. The CentOS model/community was something RHEL could draw upon to fix problems before the next roll out of RHEL. It constituted tens of thousands of free troubleshooters.

[Anonymous Coward]

"They should have a law against old dying dinosaurs buying up innovative companies"

Why? Nobody forced the sale. That's "freedom" too. You own something, you should be free to sell it to whoever you like.

In old times, when innovative companies generated real profits, innovative ones bought old dinosaurs to get customers, IP or patents, maybe some useful old brands and trademarks.

Then came FOSS, and the problem of generating profits and pay operations.... so old dinosaurs - Oracle, IBM, even MS - became able to buy innovative companies... because "innovation" no longer means "profits".

FOSS still needs to find a working business model. It looks even the "successful" RedHat had and has issue to sustain itself.

[Electronics'R'Us]

Re: "They should have a law against old dying dinosaurs buying up innovative companies"

Red Hat has been doing very well for itself for a long time. If they had not been, then IBM would not have ponied up $34 Billion for them.

I don't think Red Hat needed a buyer, but an offer was made that its shareholders and board believed to be of great value to them.

Such is the way of corporations.

FOSS has been very successful for Red Hat pre-IBM buyout - this looks like IBM pushing to get more paid licences in contradiction to the IBM statement that "...vowed to keep its new unit independent...".

This simply smacks of IBM being In Beancounting Mode.

[fnusnu]

Re: "They should have a law against old dying dinosaurs buying up innovative companies"

"I don't think Red Hat needed a buyer, but an offer was made that its shareholders and board believed to be of great value to them."

It's got nothing to do with anyone else...

[Anonymous Coward]

Re: It's got nothing to do with anyone else...

Exactly. Stupid plebs think companies should care about their "customers" or "users". They're all plebs, good only for generating revenue for the sharedholders and board.

Thank fuck for Brexit!!!!!

[Doctor Syntax]

Re: It's got nothing to do with anyone else...

If companies don't think about their users they're either a monopoly suppliers or on their way to circling the drain.

[AdamWill]

Re: It's got nothing to do with anyone else...

s/users/customers/

[Anonymous Coward]

"don't think Red Hat needed a buyer, but an offer was made that its shareholders "

Shareholder who believed evidently they could get more money from a sale than from future profits. You don't kill (or sell) a golden eggs goose.

Like many companies today, it looks the value is in their sales (or IPOs), not in their capability to generate profits for a long time. Now evidently after spending so much IBM needs to generate profits from RedHat, and the actual business model doesn't work.

Maybe in a few years it will set it to Oracle for a fraction of the price it paid...

[Doctor Syntax]

Re: "don't think Red Hat needed a buyer, but an offer was made that its shareholders "

"the actual business model doesn't work."

From the Wikipedia article: !In 2012, Red Hat became the first one-billion dollar open-source company, reaching US$1.13 billion in annual revenue during its fiscal year. Red Hat passed the $2 billion benchmark in 2015. As of February 2018 the company's annual revenue was nearly $3 billion."

I suppose the takeaway from that is that if you have a business model that doesn't work then fail big.

[Anonymous Coward]

Re: "They should have a law against old dying dinosaurs buying up innovative companies"

The price someone (remember, it need only be one suitor!) is willing to pay for a business is at most weakly connected with its profitability or prospects. In your introductory finance course, you were taught that the present value of a business is the sum of its discounted future cash flows; the discounting rate takes into consideration both future business prospects (risk) and expected future riskless interest rates.

That's theory. In practice, CEOs get paid obscene sums to occupy a chair. They are permitted to occupy that chair only so long as they are seen to be "doing something". One of the best ways to achieve that is to make big, splashy acquisitions that will at worst muddy the waters around the company's probably dismal earnings and at best drum up a lot of interest in the acquiring company's own stock (because that company became "more innovative" or whatever bosh sounds good). Because the acquirer's shareholders are also speculators who care only about the stock price and ignore earnings and dividends, this makes them more willing to keep the CEO around. The CEO, of course, does not spend her own money but the money that should have been paid out to shareholders as dividends, so the incentives don't match up; you were taught that this is called the principal-agent problem.

The last fundamental piece here is that even in theory, when interest rates are expected to be zero forever, a component of the proper valuation includes a division by zero. Any future cash flow at all becomes infinitely valuable *provided you believe you can sell the asset at a future time for the same or higher price* (the greater fool exists). In practice this isn't really true, because in acquiring equity risk becomes the dominant factor, but even so it becomes possible with some handwaving to justify truly absurd valuations. I don't for an instant believe that IBM will ever get $34 billion in nominal profits out of RH, but if you squint real hard and take a 100-year view, it's possible to imagine you'll end up with a 3% nominal ROI, which is less than it costs IBM to borrow. Of course, no sensible person expects RH's current business to be relevant in 100 years, but that's part of the folly of interest rates so low that you have to look that far out for any kind of return: investors are accustomed to thinking about getting most of their money back on an equity deal within 10 years, a horizon on which the future probably does look a lot like the present. All of today's deals are done with a horizon in mind on which the future will probably look very little like the present.

And the very last piece is this: every deal has two sides. IBM may have decided RH is worth $34 billion to them, on whatever reasoning. RH's shareholders decided it wasn't worth $34 billion to them. Exactly one of these points of view will turn out to be correct. As someone who didn't own RH at any of its much lower valuations prior to the sale, I'm obviously in the "not worth it" camp. I also happen to think RH is probably worth less as part of IBM than it was on its own, if in fact it was ever worth anything at all.

So yeah, one shouldn't conclude anything about a business from what is or was paid to acquire it.

[vtcodger]

Re: "They should have a law against old dying dinosaurs buying up innovative companies"

"...vowed to keep its new unit independent...".

Well ... possibly true for some definitions of independent

[AdamWill]

Re: "They should have a law against old dying dinosaurs buying up innovative companies"

IBM had nothing at all to do with this. Rich Bowen (who was there) has said publicly that no non-RH IBM people were in any of the meetings that led to this move. Like it or not, it's an RH decision. Not an IBM one. (I work for RH, but not on centos and wasn't involved in this at all).

Even if you think the motive is to make more money - we liked money before IBM bought us too. IBM bought RH at a value that implies a lot of pressure to continue revenue growth. This is true. But before the buyout, RH shares were priced at a level which implied shareholders had high expectations of continuing revenue growth. This is also true. Whether we were owned by multiple shareholders or one big company, RH has been under pressure to keep growing revenue for like a decade or more this point.

[Anonymous Coward]

@LDS - Re: "They should have a law against old dying dinosaurs buying up innovative companies"

I don't see greed anywhere in your post.

[iron]

Re: The RedHat corporate spinners won't fix this

> They should have a law against old dying dinosaurs buying up innovative companies just to take them down in their death struggle.

Red Hat was founded on 26 March 1993, I'd say that makes them an old dinosaur themselves. (like me)

[Dave559]

Re: The RedHat corporate spinners won't fix this

"Little they return to the Open-Source world, maybe one of them could make good on the false promises of RedHat and maintain a good distro for company use that doesn't cost and arm and a leg for small businesses."

That's essentially how Ubuntu came about: Mark Shuttleworth made lots of money from Thawte, and then used some of that to develop a polished distro based on Debian. It's still free to anyone who wants it, and therefore useful for startups or non-profits, although, like any distro, if you are a medium/large company using it at scale, it makes sense to buy some support.

[Jakester]

Re: The RedHat corporate spinners won't fix this

And I am going to switch from Ubuntu because of their phasing in "Snap" applications which made the boot and login time go from 34 seconds to 121 seconds after switching from release 18.04 to 20.04. Snap applications also take more time to install and to open. I don't know what release will work for me now.

[pompurin]

Re: The RedHat corporate spinners won't fix this

Linux Mint or Debian straight appear to be the best options. I tried installing Net-SNMP on Ubuntu 20.04 and I couldn't understand why it wasn't available with apt. You are forced to install it with snap. Why after all this time force a user onto snap when we've installed it with apt for a decade?

Linux Mint have made it clear on their website that they are against the Snap ecosystem and will exclude it from the builds.

[Daniel von Asmuth]

Re: The RedHat corporate spinners won't fix this

In cleartext: IBM Linux Love is for enterprises, not freeloaders.

[Anonymous Coward]

So much spin I'm getting dizzy

And, of course, this has nothing to do with trying to force more people to switch to paid offerings.

[Anonymous Coward]

Sooooo sour.

I think this is just one huge _final_ spin, which is RHEL is now using the "FREE" OSS model with CentOS but its commercial variant, RHEL, is NOT FREE. Basically, RedHat just joined Amazon/Google/Microsoft etc.

On a side note, this move finally rounds out the commercial & "free" players to purely commercially driven by free OSS (ie. unpaid labor). I think you can finally argue that free OSS is dead for large projects and any future it has left will be undoubtedly in the financial interest to someone other than the free OSS contributor. Of course, maybe it will take one more project to turn sour for people to see the truth (then one more, then one more, ...).

20 years ago Google brought the wolves in like no other before, but what to do now since all the sheep have changed clothes? Free OSS is not what many think it is... not anymore. What's kind of retrospectively funny (or naive) is that everyone thought it was going to be 1 company, Microsoft, and 1 piece of software, Java. However, nobody thought it was going to be EVERYONE!!

[Anonymous Coward]

Re: Sooooo sour.

It's not free labour if you're one of the biggest contributors to many of the projects in question...

[Mark 110]

Re: So much spin I'm getting dizzy

Of course its about getting people to pay for stuff.

I guess they would tell you to fork it and maintain it yourself for free for everyone if you don't like it. We are are a very young industry in real terms. We need to knock the kinks out of this stuff and if a FOSS commnity wants to keep stuff FOSS they may need to continue to fork stuff to keep it in FOSS space whenever a capitalist tries to take ownership of it.

Damn - I sound like Cory Doctorow - read one of his novels once . . hmmm

[Howard Sway]

a way of making it easier for the community to contribute to RHEL

Oh thanks IBM, you're just doing this out of the kindness of your blessed hearts to enable us to fulfill our lifelong dreams of making your sold-for-money product better. How generous of you.

[Anonymous Coward]

you could just

Perhaps just stop panicking, donate some time/money/resoruces to one of the replacement builds (like rocky) and carry on.....

[keithpeter]

Re: you could just

...or* possibly, depending on the versions of packages needed, downgrade to CentOS 7 that has security only support until 2024 and use the time to plan future roadmap?

Icon: I'm a clueless end user

* inclusive

[FatGerman]

The whole point of continuous delivery is to make each release as stable as the one before

Whereas the actual effect is that something different breaks every week See Windows 10. When you're running a mission critical system you want something where the components you rely on rarely change, and when they do it is in small, thoroughly tested increments. You want that for 5 years minimum. CentOS was good at that. Stream is the opposite of that.

Yeah I could switch to one of the inevitable new forks that will step in to fill the gap, but it's another chunk of work I haven't planned or budgeted for. This decision makes no sense technically, so it must be financially driven. Fucking bean counters.

[Len]

Re: The whole point of continuous delivery is to make each release as stable as the one before

I was just going to ask if this could have repercussions on stability but you have already more or less answered that question.

I am not that familiar with CentOS but if the order was CentOS Stream -> RHEL -> CentOS Linux doesn't that mean that CentOS users would be the last with new stuff? I'm not thinking of 'teh shiny' but support for most recent hardware, new protocols, new APIs etc.

In that sense moving to Stream would provide more access to new stuff, or am I mistaken.

[FlaSheridn]

Re: The whole point of continuous delivery is to deliver continuously

> In that sense moving to Stream would provide more access to new stuff

Yes, which is bad. The point to CentOS, and the reason it’s used by a lot of financial companies (which was why I tested our stuff on it) is that it’s stable, while still getting security updates. This has a downside—in my case having to tweak our build system to use vintage versions of CMake and LLVM, but that’s part of the tradeoff.

> "The whole point of continuous delivery is to make each release as stable as the one before."

No, the only point to continuous delivery is to deliver continuously. It may be possible to make each release as stable as the one before—the easiest way is if the previous release wasn’t well tested either. But if you properly tested the last real release, which generally involves some manual high-level testing that only humans can do (such as ensuring that your claims to customers still match the deliverable), then your continuous deliverable will have gone through fewer tests.

[James Anderson]

Re: The whole point of continuous delivery is to deliver continuously

Live production systems do not need "new stuff" they need a stable tested environment.

So stream would be great if you were developing a system to go live next year and wanted to use the latest technology. On the expectation that by the time you went live it was in RHEL/Centos thoroughly tested and stable.

[Wellyboot]

Re: The whole point of continuous delivery is to make each release as stable as the one before

Broadly speaking with numerous caveats..

Stream is suited to development work where you need to try out new features before they hit production.

Redhat is suited to mission critical highly customized production setups where heavy duty OS vendor support is worth the money.

CentOS is suited to plug/play production where minimal system tweaking & stability are prime.

I'm at a loss as to how Stream & CentOS could (from a technical pov) be described as incompatible to the point of ditching one entirely when in reality there could be anything up to two years lag between features appearing in stream and hitting the production server running CentOS.

For a company running CentOS only, swapping to another distro is relatively straight forward*. If a company is using both RH & Centos then the choice becomes a decision based on a lot of variables and I wouldn’t bet much on ‘buy more R-H’ as a regular outcome. This isn’t Microsoft, there are alternatives for users to step sideways onto.

* and they’re almost never going to be in the position of thinking about buying R-H support again.

[Marcelo Rodrigues]

Re: The whole point of continuous delivery is to make each release as stable as the one before

"I am not that familiar with CentOS but if the order was CentOS Stream -> RHEL -> CentOS Linux doesn't that mean that CentOS users would be the last with new stuff? I'm not thinking of 'teh shiny' but support for most recent hardware, new protocols, new APIs etc.

In that sense moving to Stream would provide more access to new stuff, or am I mistaken."

Yes, it would. And this is THE problem. We want boring, tested and debugged stuff. We want the system to just go on, without changes, without surprises, without new things to test and validating.

Boring is good.

More boring is "more better".

As a user I want the new shiny shiny - but as an admin I want my servers as boring as possible.

[Doctor Syntax]

Re: The whole point of continuous delivery is to make each release as stable as the one before

"May you live in interesting times" does not constitute good wishes.

[katrinab]

Re: The whole point of continuous delivery is to make each release as stable as the one before

People who want "access to new stuff" would chose Fedora or some other distro that does that.

[daflibble]

Re: The whole point of continuous delivery is to make each release as stable as the one before

I seem to recall a time that was exactly what they said. I still see no point in CentOS stream other than trying to shut down a lot of CentOS users.

[Mark 65]

Re: The whole point of continuous delivery is to make each release as stable as the one before

It's all about the money money money...

Plenty of places I know of used to run a single production RHEL system surrounded by a myriad of compatible CentOS ones. Not sure if that ruse still works these days, but it would be a good reason to get rid of CentOS - you want stable? $$$$$

[Len]

FreeBSD

I have to use the odd Linux instance (and use it quite a bit at home) but I am glad all our core business servers run FreeBSD. I don't think FreeBSD has ever surprised me, which for servers is a good thing. There is just one FreeBSD project, one code base, one set of conventions, one package manager, one repository and one group of developers responsible for both kernel and all base packages *. Sure, there are countless maintainers of external packages ("ports") but these are not a required part of the OS or they wouldn't be external.

Sure, it has downsides too. FreeBSD's business model is more a charity than a corporate and is largely reliant on corporates donating to the FreeBSD foundation as a thank you for providing the OS they have built their business around. That means that they are some times cash starved. Also, practically every VPS provider offers ready-made CentOS instances whereas with some you'll have to install FreeBSD from scratch yourself (I'm looking at you, Hetzner).

^{* There is also zero systemd but let's not get into that discussion.}

[Anonymous Coward]

Re: FreeBSD

I was going to say the same thing, I used to use CentOS to get customers to buy RHEL but now I'm likely going to go all in on FreeBSD even more than I used to, which was quite a bit. It's never surprised me either, which is exactly what I'm after for a server. I don't want surprises, at all. Neither do my customers. I want stability and not having to worry about stupid business decisions affecting a product I'm having someone use.

I just wish a desktop oriented downstream version of FreeBSD (like how PC-BSD and FuryBSD were) would survive more than a year or two as FreeBSD's stability is second to none but it's a pretty big pain in the ass to get a DE running with it, it's definitely doable but it was nice to have most of the configuration done already.

[Pascal Monett]

"The whole point of continuous delivery is to make each release as stable as the one before."

Someone should tell that to Microsoft.

Oh no, wait, Borkzilla already is as stable as the one before. Never mind.

[Michael Hoffmann]

Should it really be called Redhat anymore?

"IBM-Redhat"? "IBMHat"? "Big Red"? Ooops, that one's taken.

[Asylum_visitor]

Re: Should it really be called Redhat anymore?

IBMBlueWaffle would be most appropriate at this point.

[Anonymous Coward]

Re: Should it really be called Redhat anymore?

I'll Get My Hat (and go)?

[Anonymous Coward]

Re: Should it really be called Redhat anymore?

Red B.M. is fitting...

[Anonymous Coward]

Re: Should it really be called Redhat anymore?

Well, if you mix blue and red you get purple. So perhaps it should be Big Purple Hat, and maybe a hardened version could substitute in Helmet instead.

[Aqua Marina]

I thought CentOS was one of these bedroom projects, where someone decided to de-Red-Hat a Red-Hat distro and released it. Couldn't someone just do that again?

Genuine question.

[PM.]

They will do it ..

.. but it takes time . Cloud linux, Rocky linux , etc

[R3sistance]

The original developer of CentOS announced Rocky Linux the same day that red hat announced this change to CentOS.

[razorfishsl]

no... they announced a "let's talk ourselves to death" session

quote:

"There is no ETA for a release. "

[Ian Reissmann]

CentOS was exactly that IIUC - a one man band cloning RHEL.

I was surprised after reading this announcement how far CentOS has been assimiliated into the RH/IBM Borg. This will obviously kill CentOS in fairly short time. And that is obviously the intention. Sure - people can set up businesses and sell themselves. But the system can work badly against customers and users when competition is reduced and the market rigged.

I hope someone does come up with a new clone. We're watching the space.

(We used to be a vendor on Solaris, added in Windows. We adopted RHEL/Centos some time ago. Solaris we've now dropped - thanks to Oracle. Even more sadly the majority of our customers buy our products to run on Windows).

[Anonymous Coward]

The real issue with the death of CentOS

How many companies are using CentOS as their dev/test platform with only production systems running on RedHat? Patrticularly after RHEL

Sure you can use other distros but the number of entrerprise applications that are supported on CentOS/RedHat is much greater than Ubuntu/Debian and until alternative distros are fully up and running, paying IBM looks like a safer bet than spending a lot of money and moving to Oracle Linux only to discover you have to not only pay for the migration but also the cost of licenses in 3-5 years that are 10x the price of RedHat. Yes there are other options, but they don't have widespread application vendor support so pushing towards Debian seems to be just as wise as the alternatives.

While I acknowledge there's an element of fear and uncertainty in this based on IBM and Oracle's past behaviour, if you have an estate with $20k of RHEL licenses in it that looks like it is about to turn into $40k-60k, its not a pleasant experience and will be worse for even large companies.

[yoganmahew]

Re: The real issue with the death of CentOS

Add in continuous PCI compliance on a bleeding edge system and components, where a security patch may come with dependencies (since who's going to bother isolating security patches for what's essentially a development OS).

The timing is cynically fabulous too...

[JakeMS]

Wow

Wow... As a Fedora user for over 15 years, generally a supporter of Redhat too, and having all my servers based on CentOS..

This shocks me. So much so that I'm now going to have to look into alternative long term distros for servers.

Any Suggestions?

Top of my may switch to list is Debian, and absolutely never in gonna happen in a million years is uBuntu.

Other than that, I'm all for any suggestions :).

[PM.]

Re: Wow

Cloud Linux will supply free CentOS equivalent , soon .. ? Few months ?

[Steven Raith]

Re: Wow

Ubuntu server is a pretty good intro to the Debian landscape as it does a reasonable amount of handholding (pre-spun distros, lots of meta packages (that's what it's called, innit, when you get one package and it gets lots of 'sideways dependencies' as well as the downstream dependences?), generally more modern base packages etc and a huge, huge amount of guides and documentation for it.

Ubuntu is fine as a server, but if you've been running stripped back, well optimised servers for a while, i expect some of Ubuntu will be frustrating, but there's genuinely nothing particularly wrong with it.

If you were to bung services on an Ubuntu LTS instance, and test with Debian in the background, by the time the LTS ends, you could probably comfortably migrate to straight Debian at that point.

Or you could go the whole hog and microservice the crap out of everything and abstract the base OS into relative obscurity.

Horses for courses, innit.

Steven R

[Doctor Syntax]

Re: Wow

Ubuntu might be a good introduction for someone coming from a non-Unix/Linux background but personally as a long-time Unixer I found Debian - and subsequently, of course, Devuan - closer to past experience.

[Steven Raith]

Re: Wow

That's a fair summary - I was thinking more of my own experience coming from

Windows at home/Windows at work

Linux at home for a few years/Windows at work

Linux at home/Linux at work.

I found Ubuntu quite useful as a 'soft-change', although I had the benefit of being used to server-sensibilities - like not filling the system with unnecessary crap.

....which is why I eventually moved off to more focussed platforms, like Debian (I've not taken the Devuan plunge yet, but it'll happen one day I'm sure).

Steven R

[Anonymous Coward]

Re: Wow

Yep, Debian. Or either wait for Rocky Linux or try FreeBSD.

[Anonymous Coward]

Re: Wow

FreeBSD - you may have slightly bigger initial difference to work with, but you'll thank yourself in the long run.

[JakeMS]

Re: Wow

Thanks for the replies :)

I think we'll go Debian, I'm somewhat familiar with it since our EPOS system runs Debian already.

I threw our development server from C8 to Deb10 this morning, just to see how it worked. I LOVE the old config way of networking. Up to C8 I was still clinging to the old sysconfig network files.

Current Fedora outright drops that ability it seems in favour of nmcli (or gui).. Which I really dislike. Call me old fashioned but I prefer to just edit a bunch of text config files, I find it faster and I'm used to it. I've been doing it that way since before networkmanager even existed.

So I've already fallen in love with Debians network config method, it was super easy to setup multiple IPs on the interface and the vpn.

Firewall was pretty much exactly the same config (shorewall).

Again, ssh and keys easy as pie (as always, same old thing)

PHP is a little different, but I had to use surb repo, who claims to be the official Debian packager.. I've yet to verify that claim.

That's as far as I got before I headed out to work lol.

But.. so far so good.

[sansva]

Re: Wow

Cringe. It's naive to evaluate an entire OS based on one particular network tool or other tools that might be provided as defaults. Many tools exist and a good sysadmin will know then all and how to choose the most appropriate ones.

[JakeMS]

Re: Wow

Yeah.. I didn't evaluate the entire OS based on one tool.. I said "so far so good" and that I liked/appreciated the better network management (so far, it's better than having to kill off NetworkManager and deal with the mostly borked network daemon in modern Fedora.).

On systems where the network configuration is never expected to change (aka, IPs never changing, gateways not moving, no wifi etc) NetworkManager is simply not necessary, a simple network config file will suffice. In fact for us, it's preferable the network dies outright instead of changing config to reconnect automatically - because there is no valid reason the network should change, and if it does, I need to know, killing the connection is acceptable in that case.

As for SSH, used it for years - will continue to do so. Shorewall, ditto - the previous systems were already using those.

PHP again, required due to existing projects.

The tools I use, with the exception of the network management, were evaluated with the previous OS's and are being used appropriately for the systems at hand.

It's entirely possible that something else might pop up and change my mind. But until then "so far so good".

I've also researched the lifespans of the distro, and the upgrade paths - they seem acceptable for our usage.

The real test will be the KVM host that's already running C8.. the guests are easy to switch OS.. the host itself not so much.

[Smartypantz]

Re: Wow

Been running KVM in production on debian 7,8 and 9 for many years now, so no worries there.

Migrating existing CentOS KVM setups to debian will ofcourse be another matter. maybe just move the guests as raw images if your setup isn't to large and you can find a window for the downtime?

Good luck!

Been a happy user of debian on production servers since "sarge" (3.1). Have one server, distro upgraded (with a lot of applications) all the way from 3,1 to 9. System was first installed in 2006, and still going strong! :-)

[Smirnov]

Re: Wow

openSUSE Leap? It's pretty much the CentOS equivalent to SUSE Linux Enterprise.

Also SUSE is, after all, the #2 enterprise Linux vendor after Red Hat and is used in a lot of businesses in Europe and other countries (it's also the default platform for SAP).

SUSE has been keeping openSUSE around for much longer than CentOS existed, and have built a great community and additional goodies like the OBS for packages which aren't in the standard repos. Nvidia also maintains a separate repository for SUSE/openSUSE so installing their binary drivers is a lot less painful than on other distros. And then there is YAST...

[JakeMS]

Re: Wow

Ah, SUSE. I always forget that one.

I just did a preliminary search (at 3am), according to their documentation, OpenSUSE has a 36 month lifespan between major upgrades. That's a decent amount of time - but I'm worried how well their upgrades would work? Often distro's require complete reinstallation when a new version is released.

Graphics drivers etc are no issue as we're talking about servers here, so the no gui, setup once and maintain type) - That and our EPOS/Desktop systems recently got upgraded to ryzen with amd graphics - I don't miss nvidia :P

Now, Debian on the other hand has upto 5 years with their LTS:

https://wiki.debian.org/LTS

That's still no where near the old, now discontinued CentOS 10 year life span, but it's still a fair amount of time with enough time to plan upgrades as usually, our upgrade is renting a newer replacement dedicated server, setting that one up with the new OS and testing etc, then switching the traffic to it and dropping the old one (Mostly this was done because upgrading between CentOS major versions always required a reinstall).

With that said, 36 months is still 3 years. So it's not an overly short time either and could be do-able. But it does feel short, our servers are usually rented on a 2-year minimum term so we get discounts :P

Thus our absolute minimum expected OS installed time is 2.5 years (allowing time for switching), but we tend to try to keep the servers and OS running longer if we can.

From a Desktop point of view, I'd be interested in OpenSUSE. Sadly not so much for a sever at this time.

But, while we're on SUSE, let's not forget SLES. Out of the SUSE's, SLES would probably be the best candidate next to RHEL or CentOS. But we're a small family run retail business and with everything that happened this year, it's probably out of our budget (last time I checked, it was subscription like RHEL?) - Do Novell still own SUSE?

For the most part I'm just dumping my thoughts here - I'm not an expert in SUSE (any of them) so my information could be incorrect or just straight up wrong.

If I'm wrong, please feel free to correct me with the right information and I'll be happy to take another look into SUSE :-).

[Smirnov]

Re: Wow

"OpenSUSE has a 36 month lifespan between major upgrades."

So has RHEL/CentOS, which also has 3 to 4 years between major versions. But a new major version did not stop the previous major version from being supported.

Similar to RHEL/CentOS, openSUSE Leap 15 now follows the SUSE Linux Enterprise life cycle which means a new major release every 36-48 months and one minor release per year. Which is not much different than for for CentOS/RHEL. SUSE Linux Enterprise 15 is supported for 10 years (2018-2028) and with openSUSE following the same release cycle it will offer a similar support life (i.e. openSUSE Leap 15 support won't stop when openSUSE Leap 16 aka SEL 16 comes out).

Also, unlike with RHEL/CentOS, openSUSE/SEL generally support upgrades from one major release to the next one.

All this is pretty new, though, as in the past openSUSE and SEL had quite a number of differences, with each one being treated independently (which also meant openSUSE suffered from bugs that didn't exist in SEL). But while RH was working on killing off CentOS, SUSE has actually increased its commitment to openSUSE by moving it over to the SEL branch, which brought the stability and reliability of SEL into openSUSE.

"Now, Debian on the other hand has upto 5 years with their LTS"

Yes, 5 years for a community distro which has a very long track record of questionable code quality and a number of pretty bad fuck-ups, some of them pretty bad (like the Debian openSSL fiasco), and where the pseudo-LTS support even comes from a separate (smaller) community. The worst thing is that a lot of the contributions to Debian come from people who have no clue how even to write a proper bug report.

Don't get me wrong, Debian is great if you want to roll your own stuff and have the manpower and the budget to do that (which is why it's the basis for a lot of other distros like Ubuntu), but Debian itself with its strange understanding of security is as far away from real enterprise-class Linux such as RHEL/CentOS and SEL/openSUSE as it gets.

For server use, I tend to stick with the big boys. That both RHEL and SUSE are also some of the biggest contributors to Linux and other FOSS projects who pay professional developer to work on open source stuff so is just the icing on the cake.

[JakeMS]

Re: Wow

So, if I'm understanding you correctly, OpenSUSE is now supported for 10 years the same as SLES, and it's basically the same distro now?

This sounds very similar to what happened with CentOS - CentOS was community only, Redhat came along and said "Hey we just want to fund you for the good of the community, don't worry we won't try to change CentOS" (essentially, not word for word).

Well a couple of years later here we are. I remember reading a few months (or years?) ago OpenSUSE split from SUSE and went community only - has this changed now?

My last question, how well is SELinux supported on SUSE? I've found with Debian SELinux support is patchy at best and for the most part to actually make it work I'm writing my own policies for stuff which should be done out of the box (Like it is in RHEL/Fedora) - I had to fix a couple of policies just to get Debian w/ SELinux to boot in enforcing mode. (I was testing/working on that yesterday, on christmas day, it's more fun than family and gifts anyway lol)

Although, most people will tell you "just disable SELinux you don't need it!" - I've always disagreed with that sentiment, I'd rather stuff gets blocked and breaks until I fix it (or distro devs) than just runs and causes havoc if it's bad.

Debian's works - but it doesn't have setroubleshoot, so you're relying on classic tools such as ausearch and audit2why etc.

Thankfully, I'm well versed in SELinux due to my CentOS/Fedora experience.

Yes, I'm well aware Debian and many other distro's would rather use apparmor - but that's only effective if you have policies on an application by application basis, and allows all except the applications with a policy. - I prefer the SELinux method of block all and fix later.

[demon driver]

"Antithetical"

The "antithetical" claim is of course wrong and purposefully misleading. CentOS and CentOS Stream are no more antithetical than any two of Debian stable, Debian testing and Debian unstable would be antithetical. Or any other Linux's LTS and non-LTS distro. If we'd call any such pair antithetical, it would just show that being antithetical does not exclude fruitious coexistence. No, it's not CentOS vs. CentOS Stream, it's CentOS vs. RHEL, because profit and non-profit, trying to sell something while giving the same thing away for free are of course the really antithetical concepts here.

And there's no easy way out, either. IBM isn't healthy, they're massively lowering costs, and according to recent news it seems they're about to fire 10,000 people in Europe in preparations for separating their "managed infrastructure services" unit from their "Global Technology Services" division. Accordingly, there probably will be many executives thinking they cannot continue to watch Red Hat giving the same thing away for free that its primary business is built on.

Of course, whoever made the decision to kill off CentOS might not have thought about the possibility that people might actually not go to RHEL (or CentOS Stream) when they cannot have CentOS anymore, but rather to another distro altogether. I guess especially curtailing CentOS 8 support will be seen by many as the breaking of a written promise, which will not improve trust in Red Hat as such and in its commercial products and services, either.

[J27]

Retire the Red Hat name already. Just call them IBM.

[Blackjack]

Red Hat speech in two seconds

Give me money!

[amacater]

Not ideal

This is being covered significantly in the tech press and over at LWN.net. This is a PR disaster for Red Hat. Rightly or wrongly, the CentOS ecosystem gave Red Hat a boost - people deployed CentOS to learn Red Hat ways and then go and take expensive Red Hat certifications, could take a CentOS and trial build a server or two, could gain expertise. That goodwill is gone in a flash. If Red Hat made the decision a while ago, it still wasn't communicated well or at all so that the message didn't get through - the announcement on December 10th and the change in lifespan for CentOS 8 (and Red Hat 8 coming down from 10 -> 5 years support) came as a bombshell to many. [And yes, folk moving _lots_ of machines from CentOS to <anything else> in a relative hurry means that Red Hat will never get those subscriptions. ]

"Gratis/Low cost Red Hat Enterprise Linux for developers" promised for soon to replace CentOS - they really need to sort out their distribution channel first: a single threaded, single download from a distant server of an 8.8G DVD image (I know everybody's got big USB drives, but it could have been made to fit an 8G stick) and, for me, a problematic install on KVM (mouse just not playing properly, meant that I couldn't register the machine properly). There's a lot riding on making Red Hat Enterprise really something that people want to use again.

[amacater]

Re: Not ideal

From a comment on CentOS-devel where people were talking about what it takes to rebuild Red Hat from source:

Red Hat Enterprise Linux 8 is not self hosting

Run, don't walk away from the whole Red Hat ecosystem: bad enough when the installer doesn't work but that's obviously only a minor glitch.

[keithpeter]

The decision was reluctant

The quote below is from Johnny Hughes, one of the CentOS maintainers who has been working on CentOS for 17 years...

"So, no one 'FORCED' the [ CentOS project's] board to do anything. Red Hat told us what they were going to do (what you quoted). The board then made many recommendations in a back and forth negotiation. The board then made a decision. The decision was reluctant .. but it was unanimous."

https://lists.centos.org/pipermail/centos-devel/2020-December/075528.html

That must have been a fun board meeting (not)

Icon: to all those upstream and downstream

[amacater]

Re: The decision was reluctant

One of the comments attached to that, I think, on LWN was along the lines of - if someone is holding a gun to you and says "Your head or your legs" - what are you going to say?

[Pirate Dave]

Shame

CentOS was a great server option for those of us who weren't keen on the wooly bits swirling around the new-ish RHEL 10-15 years ago, but were used to where Red Hat put their config files and just needed a simple No-BS distro for simple webservers, DNS/DHCP, and firewalls. I still run a couple of websites on 6.x ( fuck systemd, but I digress ), guess at some point I'll have to move them to something newer, probably when the Let's Encrypt renewal script decides my system is too old to bother with. Ah well, nothing stays the same like change.

[amacater]

Re: Shame

CentOS 6 - move now - you have NO security support and no fixes. Please don't leave something old and unpatchable anywhere on the 'Net. Regular checks and keeping on top of patching are good and useful no matter which distribution you run.

[TrevorH]

> "the CentOS Board signed on to it."

In the same way that a hijacked plane pilot signs on to going to Cuba...

[Gnomalarta]

Good Choice

Seen from the crapfest formerly known as the United Kingdom Cuba looks like an excellent choice!

[Anonymous Coward]

IBM to PHB:

We want you to use RedHat because - well, you'll have to pay us money but we we can support it for you. The service will be <snigger/chortle/guffaw> great!

Oh, and here is a nice new shiny cloud too! It runs RedHat. Pay us more and move it there. It costs more, but its Opex and not Capex so you can hide the true cost.

PHB:

Ohh. Its shiny. And I get a bonus because it looks like I'll get a great service and reduce Capex. Where do I sign.....

[Anonymous Coward]

It's funny, I used to use CentOS to sell RHEL to customers who were wary of Red Hat as a business. This whole ordeal with IBMhat acting like this means that I likely won't be doing that, it'll likely be SLES instead if someone demands Linux, and overall a lot more FreeBSD, since I don't generally recommend Debian derived distros.

[Doctor Syntax]

"since I don't generally recommend Debian derived distros."

Debian is, of course, a distro in its own right. You can recommend that rather than a derivative.

[Matthew "The Worst Writer on the Internet" Saroff]

Shorter Version of What Red Hat is Doing Comes from Good Fellas

F%$# You, Pay Me.

[cjcox]

Dead broke Red Hat? Really?

Red Hat can continue to provide CentOs (remember, was once an independent non-Red Hat entity, that Red Hat took over in order to "control") as well as having a "test bed" for down the road.

In the beginning, you have to remember that Red Hat believed that Fedora was their sufficient "test bed", but due to community "politics" they had to take their hand off of that one, and it quickly started going in directions away from Red Hat.

And this has been true for many years.

Not sure why Red Hat is making a change now. But they've decided to nuke CentOs (something that would not have been nuked if Red Hat hadn't insisted on "controlling" it). And since Fedora has "gone wild", they want a "test bed" (where you and I are the testers) for their enterprise ($$$$) distribution.

All this says is that Red Hat can't afford to do both a "test bed" and maintain control of something they never should have had control over to begin with.

It takes money and effort... Red Hat is trying to escape both. Real reason? Unknown.

[AdamWill]

Re: Dead broke Red Hat? Really?

This is... utterly and wildly wrong. Fedora and CentOS Stream both fit into the RHEL development process in perfectly clear and simple ways. New RHEL major releases fork from Fedora; after that happens, development within that them RHEL major release branch happens in a "centos stream" repo. See the image from Stef's post, https://blog.centos.org/wp-content/uploads/2020/12/fedora-centos-stream-rhel-high-level.v4.png . it's really not complicated at all.

[Anonymous Coward]

CentOS → Ubuntu? I find that hard to believe

My feeling is that that commenter was bullshitting. Why would you migrate between the two possibly most incompatible distributions that you can find?

[bombastic bob]

Re: CentOS → Ubuntu? I find that hard to believe

maybe it's time for an "Enterprise Linux" version of DEVUAN ???

I mean, if you're gonna migrate, you might as well REALLY migrate!!!

[amacater]

Re: CentOS → Ubuntu? I find that hard to believe

Sarcasm aside - I'll assume this is a serious point and deserves to be answered seriously: the one problem that all distributions have is a shortage of developers. Devuan is a fork from Debian and has fewer developers for a similar size distro. If you want to make Devuan into something well supported, it requires more developers.

If moving to Devuan because you don't like systemd - fine, but you may need to find extra folks to help with maintaining sysvinit and other init systems for a larger user base and a longer term.

[Doctor Syntax]

Stream can cover "95 per cent (or so) of current user workloads"

It's the users who'll decide that.

[keithpeter]

I'm wondering* how they counted 'use cases' in order to arrive at the 95% quoted. Some possibilities spring to mind...

By organisation...

If CERN counts as one 'use case' for CentOS Linux because it is one organisation, and each of 30 small startups somewhere in California who are quite happy to run CentOS Stream on their VMs count as another 30 distinct 'use cases' because they are all different organisations, I can see how the figure might be arrived at.

By category of 'use case' irrespective of frequency in the wild...

Easy to arrive at 95%. Invent 19 'use case' categories that fit into CentOS Stream's profile. Then the 20th category is CentOS Linux based rock solid server.

I'm guessing that a frequency based census of CentOS consumption at machine/VM level would produce a noticeably different percentage figure.

Icon: counting is hard.

* Does not really matter in the grand scheme of things. RedHat employees are to be congratulated on the range and ingenuity of post-decision rationalisations being communicated. However the decision has been made and is unlikely to change.

[The Pi Man]

Give it back!

I don’t mind that IBM / Red Hat don’t want CentOS anymore, but they could at least offer it back to the community to pick it back up. I know the history of how Red Hat ended up with it. Maybe it’s time to pass it on, not kill it.

[ecofeco]

IBM gonna..

... IBM

[SecretSonOfHG]

The IBMfication of RedHat

Has started even sooner than I expected

[Roland6]

Using Centos but paying RedHat millions for suppport? Doesn't add up

> I now have to migrate 454,000 nodes over to Ubuntu because Redhat just made the dumbest decision short of letting IBM acquire them I’ve seen.

Whitehurst how could you let this happen?

Nothing like millions in lost revenue from a single customer.

Something doesn't ring right here. I presume they are going to pay Ubuntu for support (and help with the migration?), or is this another case of a big non-paying user discovering there is no such thing as a free lunch...

[mark9w]

Re: Using Centos but paying RedHat millions for suppport? Doesn't add up

You can expect to pay up for Ubuntu some time in the future too. Unless a distribution is owned by a foundation and not a private (or even worse publicly traded) company that is what is going to happen in the long term. Especially if they are bought up by IBM...

[sansva]

As a long-time Debian user I stand back and watch and think to myself about all the CentOS fans "You get what you deserve". It's not like you all didn't know what you were dealing with the entire time.

[JakeMS]

Initially, CentOS was a completely independent distro, which is when most of us started using it - it wasn't owned by Redhat or any other company. It was entirely community managed. Only in the last few years did it become company owned.

[razorfishsl]

it's things like this that will kill Linux

[simpfeld]

Such a stupid decision

We are a pure RHEL shop, but RHEL is much more valuable to us with Centos. We need third-party repos and these get built with Centos. Lots of HowTos and documentation exist just because Centos exists. Centos drives ISV's to release RHEL software, as it gives a large market share. Our staff learn with Centos at home. The Centos many users find bugs we do too. Centos streams, by not being the same as RHEL pretty much stops all of these.

This decision massively diminishes RHEL! This could rank up there with Nokia, Ratner etc as bad decisions that hurt companies. RH are basically taking away eye balls from their products.....given they are fighting against Ubuntu in the server space, this is very very unwise. Why would anyone starting out look to RHEL.

[keithpeter]

Re: Such a stupid decision

"Our staff learn with Centos at home."

1) RedHat do offer a 'developer licence' for RHEL for free for use by individuals. Might be worth exploring...

2) Springfield Linux might stand in for CentOS Linux for the 8.x branch

3) CentOS 7 branch supported (no new features, security only) until 2024 as you will be aware

Best of luck with it all

[Herby]

Wasn't Fedora supposed to be that?

As I understand it, Fedora was the upstream of RHEL and then to CentOS. Why change this? Who is in charge for these things? It seems that somebody is using mind altering chemicals to come to this silly conclusion.

Me? My work uses CentOS (Usually 7.x), but at home I'm a Fedora guy. Both have worked for me, and I really like yum/dnf as a package orchestrator. I don't know the next step.

[keithpeter]

Re: Wasn't Fedora supposed to be that?

@Herby

"I don't know the next step."

Standing still for a few months?

The dust will have settled by then and a future direction will be more obvious.

*EL 7.x (CentOS, Scientific Linux, Springdale, Oracle, RHEL) all have non-feature updates until 2024 I gather.

[gatch]

Just be honest

It's disingenuous for Wade to say "the CentOS Board signed on to it" when Red Hat has a controlling interest in the board. People would have more respect if they admitted revenue drove this decision instead of making up excuses. IBM clearly haven't learn a thing from Oracle's handling of the Sun acquisition and are likely to suffer the same outcome.

[rcxb]

Self-inflicted injury

RHEL needs CentOS more than CentOS needs RHEL. If you're paying attention you'll see that EPEL is largely empty for major new RHEL releases until AFTER the first CentOS release of same, even if that's many months later. A large community is contributing to CentOS, not RHEL, but RHEL gets the benefits, too.

I'll admit that CentOS was in need of some changes. Anybody here every reported bugs? You'll NEVER get a response from anyone. Even if that improves, it's still not acceptable to make CentOS the beta (or even RC) testing grounds for RHEL. If there's a bug in a RHEL update, the customers have a RHEL support contract. CentOS users do not.

What does IBM think is even the alternative? That companies are all going to pay for full priced RHEL licenses for every little test system? That companies will be happy to run production systems on bleeding-edge Fedora? IBM better just hope that RockyLinux and CloudLinux work out, because otherwise this move will be the start of a major decline in RHEL's fortunes, as IBM tries and fails to milk the corpse.

SUSE isn't much better, and Ubuntu does some very nasty things. If the new RHEL rebuilds don't pan out, I'll be heading to Debian.

[clickbg]

RH was on a decline stream and their growth has staggered ever since Docker was made popular. I work for a competitor of RH and we stopped looking at them as a viable competition way before the IBM buy. The buy just confirmed that the game is over for them, which was a shame since their products were way better that ours but we have much bigger user base and we actually moved way for selling operating systems to selling solutions. I personally used CentOS for many years and am sad to see it sink with the RH ship but that was also kind of expected when they joined RH. RH never had inceptive to keep CentOS alive. I am moving back to my UNIX roots with a combination of FreeBSD and OpenBSD since Ubuntu (while a great desktop OS) is a terrible server OS. So BSD for personal projects while my employer was smart enough to do its own distro (which is a gigantic turd) but at least we won’t have to migrate it.