I spent the entire weekend updating MaraDNS’s Docker testing container to run Ubuntu 20.04 instead of CentOS 8, and am in the process of replacing all of my CentOS 8 machines with Ubuntu 20.04 LTS machines.
RedHat flat out made a pinky promise that CentOS 8 would be supported with security updates until 2029: https://archive.is/7Qmtw I installed CentOS 8 on my home network (a non-trivial task) with the understanding things would still be secure in 2029. RedHat broke their pinky promise: CentOS 8 will no longer be supported after 2021. That in mind, while I have been using RedHat since the dot-com days, I am now a Ubuntu user.
Breaking a promise like this is non-trivial, and I have stopped using RedHat as a result. If RedHat had said “We’re not going to release CentOS 8” or if they had said from the beginning “CentOS 8 will only be supported until the end of 2021”, I would be singing a very different tune right now.
In the same boat, but after this lesson I don't think I can even trust Canonical to not do something drastic. Debian is looking to be my CentOS replacement.
To me, Debian sounds like it should be hands down the de facto standard for production environments. It's rock solid, and it sets the bar on reliability.
I agree! i use CentOS streams because i need some sort of redhat environment for my customers, but if i had the choice or if i have to do some opensource work, i would choose Debian over anything else....
Christoph Galuschka maintains that wiki page you reference. Does he represent Red Hat? If not, how does a community member “pinky swear” and Red Hat is now on the hook?
RedHat had a year to remove the statement that CentOS was supported until 2029, and could had protected the page so only admins could edit it if vandalism was, in fact, an issue.
RedHat is on the hook because they did not make it clear to the CentOS community that support would be cut short back when CentOS was released in 2019. This is something RedHat should had decided on and communicated before CentOS 8 was made public.
So am I correct that you’re going from being a non-paying user of one distro to being a non-paying user of another distro? Not sure sure RH will be deeply moved by your bravery.
It sounds like MaraDNS was paying RedHat with their time by testing against Centos.
Like many other pieces of upstream open source software, they're no longer (financially / logistically able to) provide that free testing work to RedHat.
I'm guessing RedHat will be "deeply moved" if their actions create a mass developer exodus.
>Under Red Hat, the project was directed by an independent governing board and many of CentOS's head developers were Red Hat employees, working apart from RHEL developers as part of the company's open source and standards team. As part of the agreement, Red Hat took possession of CentOS's trademarks.
Embrace
>CentOS itself was not actually providing that much usefulness to Red Hat
> Hopefully, the "extend" and "extinguish" are not on the agenda.
There are certain organisations whose prior behaviour precludes "hope" from a reasonable set of expectations. Whether or not Oracle is included is left as an exercise to the reader.
As a former Oracle employee I am assuredly biased. But here's my point of view: if Oracle tried to pull one of those moves, people would just jump ship to something else (just like they are doing here in response to Red Hat/IBM). And likely Oracle management knows this, and so they likely won't.
Oracle Linux is very different from other Oracle products (e.g. database, ERP systems) in that there are quite low levels of lock-in / switching costs. "Embrace and extinguish" works best for products with high switching costs, which isn't Linux.
If you are particularly worried about this, just avoid using any of the features Oracle added to RHEL, like DTrace or Btrfs. Just stick to the RHCK (Red Hat Compatible Kernel) instead of UEK (Unbreakable Enterprise Kernel).
For the reader: Using no-cost Oracle software is playing with fire. Oracle is a machine build to extract license fees and, when necessary, to litigate.
He's not joking either. I did contracting work for hospitals, upgrading their infrastructure. Strictly backend network stuff -- routers, switches -- but got to sit in on all of the IT meetings.
Oracle asked them for license count and a core count of their environment. Newer IT guy sent them all of that info, asked for pricing for upgrades to a few newer Dell systems.
Oracle thought they were only using about 3/4 of the machines / cores / licenses they were billing them for, and promptly demanded the hospital group pay for the rest of their usage, as far back as they could prove. Lawyers got involved fast.
My work finished before it was resolved so I never got an exact answer but IIRC they paid something extra to Oracle.
I tried this back in the RHEL 6 days. Do not recommend.
They changed things just enough that you can't just drop in RHEL and expect drivers, performance, and other kernel-level stuff to be the same. Now I understand why, it was to better support their database. But it was frustrating to deal with unless you were going all in with them and planned to pay them for systems that needed support.
That takes me back. Yes, we could not use the UEK so we tried using the RHCK. But we were still running into issues trying to debug kmods that we weren't running into on CentOS. I guess we weren't in the target market there for Oracle. It left me with a negative impression that maybe is unwarranted, but I've had plenty of other bones to pick with Oracle so I would be very cautious in depending on them for a platform.
See Java as an example. OpenJDK (and IBM/RHEL, ironically) is the only thing keeping that from turning into a bigger support trash fire.
Second, it would likely be possible to run the CentOS kernel on Oracle Linux. My research for Btrfs ran the other way, but I was able to get both the latest Fedora and Oracle UEK kernels working on CentOS with various levels of Btrfs support.
So Red Hat is promising a possible migration path for CentOS 8 users to some limited, restricted form of RHEL next year, but not announcing it yet? So they deprecate a product, and expect people to wait for an announcement that may or may not work for them? Who is making these decisions???
Why on earth not announce the free RHEL offering now. Or wait until those details are ready to deprecate CentOS 8.
It sounds like some CentOS users might be able to migrate to some free version of RHEL, but it's unclear how restricted that free version will be. Probably good for home users, but how about medium or large businesses? And then, why should anyone even use a restricted RHEL when Rocky will be full RHEL for free.
I imagine it's IBM leadership that decided this one.
"It's unclear how restricted that free version will be"
Probably not directly restricted. Just restricted in the sense that it won't be an LTS type distribution that matches RHEL's LTS product, package versions, etc. That was the value prop for CentOS.
This is IBM signaling that the good-ole' days are in the past. If you want to run RH stuff, they will get their pound of flesh. If you're not paying, you are not something they're concerned with.
When the corporate tone projected externally changes, that usually signals more than just a product change.
"When CentOS 8 reaches its end-of-life, Red Hat will become something of an outlier among commercial Linux server developers, as the only one not making free version of their product available to users. Ubuntu Server is free for users to download and use, and a drop-in replacement for SUSE is available in a downstream community distribution, openSUSE Leap."
Sounds like a great opportunity for Ubuntu and SuSE to pick up market share as the CentOS replacement.
To me, OpenSuSE is a perfect balance between the fad-chasing Ubuntu and the glacial CentOS (well, the glacial RHEL now). It’s stable, does not break nearly as easily as Ubuntu, but is still more current than RH. I hope they manage to get some users from this fiasco.
I've been an OpenSuSE/Leap user for many years now and agree about the balance and stability. However there's still a pretty big gap between Leap and SLES (their enterprise distro) plus each Leap point release is only supported for 2 years (something like that).
Maybe it's better now, but there were a lot of packages available in Leap that weren't part of SLES. It's been a little while since I've last tried SLES - still super happy with Leap 15.2.
Aside from the pace of updates, are there any meaningful differences that new OpenSUSE users should be aware of? (I figured I'd switch from CentOS to Ubuntu, but I hadn't considered OpenSUSE).
Compared to CentOS, it is really easy to use common desktop environments. I can’t say about very niche ones, but I use XFCE, which works well out of the box. YaST is the convenient GUI where settings and some common tools are located. It is more than adequate: not perfect but much better than a myriad of small setup programs or the abomination that is whatever Windows does. The package manager (zypper in the command-line, or a YaST module for the GUI) is a bit different but not revolutionary and still RPM-based.
Compared to Ubuntu, it tends to be more conservative and better tested on server or workstation hardware. What finally drove me away from Ubuntu was a very painful episode when the SAS driver would freeze randomly about 4 times out of 10 when loading. It is also far less opinionated than Canonical regarding which desktop you should use. The main drawback is that OpenSUSE Leap is supported for 2 years, so no equivalent to Ubuntu’s LTS releases.
OpenSUSE also uses systemd, but this should not be a shock for someone coming from Red Hat. It works decently with nVidia’s proprietary driver. Well, as decently as this garbage can run, anyway. I haven’t tested Wayland, as AFAIK XFCE still requires X.
People I'm dealing with seem to be concerned that Canonical/SUSE might pull the same stunt. Personally I think the backlash RH have created is enough to discourage either company from following this path.
I actually don't understand the purpose of CentOS if Red Hat owns it. My understanding is that CentOS is mostly RHEL with the all trademarks replaced.
So if that's the case, what's the difference between a free RHEL without a commercial support contract/license and CentOS other than Red Hat not having to maintain those IP-related changes for CentOS?
The alternative scenario that most people seem to be assuming, where there will be no free RHEL or RHEL-like OS, seems impossible to me as somebody somewhere will just recompile RHEL source and release it under a new name just like CentOS did before. And of course Red Hat knows this.
People keep saying this all over these threads, and I can't grasp why.
The difference is that a free RHEL without support doesn't exist, while CentOS does.
You can get RHEL for free through various developer focused programs, but that is most certainly not the same as having an unencumbered ISO available for unlimited downloads with no EULA or sign-in requirement.
I think you're missing the point. Nobody would argue that CentOS doesn't make sense - the argument is that it doesn't make sense if RHEL owns it, as there is a clear conflict of interest.
I think you might be replying to the wrong comment, as this doesn't really relate to my point at all.
I don't disagree with the sentiment that RHEL also controlling the free alternative doesn't make sense. I do disagree with repeated assertions that people can just use RHEL for free - for the most part, they can't.
There is no evidence that RHEL will be releasing a free version to replace CentOS, and the fact that others will take over a replacement OS development is already known.
My comment was focused on the confusion/exasperation at why people keep implying this is no problem, because we can just use RHEL for free, as that is not true.
"Look, just because the terms under which we got the bulk of our product stated that 'You can have this, and build upon it, and even sell it, and all you have to do is make it available for the next guy same as you got it yourself...' and we agreed to those terms, that doesn't mean we can actually tolerate a bunch of parasites mooching off us."
Maybe not the most popular opinion but one company maintaining two stable server focused Linux distributions which mostly have the same focus seems to be quite a wast of limited resources (dev time for maintaining both).
So it seems to be quite a reasonable decision.
Through I'm also not a user of either of them so I might have missed some important differences between them.
I agree, maintaining two popular distro's is pretty distracting and a potential waste of resources. But CentOS was just RHEL with the branding removed, I'm not sure it was that much of a headache to maintain.
IMO CentOS was kind of a bait to get people into the tent, and then if they needed help once using it across all of their infra they would eventually have to migrate to RHEL w/ enterprise support.
Also what confuses me here, is that they claim CentOS Stream is going to be the cutting-edge distro for RHEL... however Fedora already serves that purpose.
The explanation I've read for the difference between Fedora and CentOS Stream is that RHEL branches from Fedora at major releases, but those are so far apart that there could be years and massive divergences between today's Fedora and the RHEL 8.x release that's about to come out. So CentOS Stream is the dynamically updating "thing that will become the next RHEL 8.x" and Fedora is "things that will go into RHEL 9 one day". Basically CentOS changes from "whatever was in the previous RHEL point release" to "whatever will be in the next RHEL point release".
I've been watching the way OpenSuse is matching binaries with SLES. This is something CentOS never had. The binaries were generated from the same source, but inevitably had some differences.
It would be much simpler for them to provide a version of Redhat that could be run without subscriptions and some way for it to convert to Subscriptions.
Not really. CentOS was the “gateway drug” to get everyone hooked on the RHEL ecosystem, which they could then convert to paying customers. It also allowed many vendors free and fully compatible access to make their products for the RHEL platform. Now that that’s gone, the converting users will dry up, and the vendor software will dry up.
By saying it was “not useful” to them, they really have an extremely limited and short-sighted definition of “useful”.
It really should not have been such a huge extra effort to keep CentOS around in its previous form. RH is already building the packages for RHEL, and CentOS could almost be a side-effect of that process.
> Maybe not the most popular opinion but one company maintaining two stable server focused Linux distributions which mostly have the same focus seems to be quite a wast of limited resources (dev time for maintaining both).
This is kind of obvious. When RH bought CentOS, everybody was saying "They will kill it". But they didn't and said, "See? We are nice folks!" The problem is they got bought themselves and can't be nice folks anymore.
It says further down the article that the restrictions include providing names and email addresses to Red Hat and that the software can't be redistributed. Not really in the spirit of CentOS and open source.
They have hinted at what it will be, which is essentially an expanded developer license that’s allows use only in limited ways. There’s no indication they will be releasing anything that allows wholesale replacement of production CentOS infrastructures with something officially RHEL.
I am thinking this might end up hurting Red Hat more then many people think. I worked at Red Hat in global support in the RHEL 5/6 era and the developers were very opinionated and the dedication to Open Source amongst developers especially was almost cult like.
Many developers were brilliant, and it wasn't uncommon to hear that they took the role with Red Hat earning substantially less then their previous roles so they could have 20-50% work time dedicated to their open source project. Also Red Hat would be fairly flexible and many worked offsite, so were able to have a cheaper cost of living.
Many times these open source project wasn't even part of the RHEL / paid offerings, however it was obvious that management liked having a team of brilliant engineers on the book to throw at troubleshooting serious issues. When the largest customers would have a serious ongoing problem, they would often demand these engineers, who would often be thrown in a room with engineers from all the other vendors to "Fix It".
I wonder if these people are going to start to leave and need to be retained at higher salaries. If this would then cause a loss of talent that will affect the ability to support those highest value customers. If this ends up costing Red Hat more in the long run.
I mean, it was confusing to see it owned by Red Hat, and these changes to effectively remove a RHEL substitute are surprising only in that it took so long.
I never quite understood this either. The original model where CentOS was not affiliated with red hat did though.
Could an independent group do the same with the latest RHEL sources?
> Could an independent group do the same with the latest RHEL sources?
Yes. But there would be an up-front investment in effort and a slow start, as there was with CentOS itself. The people who started up CentOS and maintained it pre-RedHat will mostly be off doing other things (possibly within RedHat) so that particular source of experience is not likely to be available at all. You'd need to find enough people who are the right combination of interested enough, capable enough on the technical side, have access to enough hosting resource for it to be practical as a public project, and willing to do it for free (unless you can attract some sponsorship or other funding).
Edit: as others said in replies while I was typing the above: it looks like a group has already started working on this, so you might be in luck.
Honest question, because I’d love to know: are you actually a paying Redhat customer and a CentOS user, and is CentOS the reason you’re a Redhat customer?
I’m asking because I get the feeling that there isn’t much overlap between CentOS users and paying Redhat customers.
Where I work there is a bit of overlap because some projects want the dev environment to match the final environment, and Red Hat is basically the only player in the game for prod in government. I can see this pushing some projects to pay for Red Hat for dev machines versus take the risk of having a separate dev environment.
Where that isn't a concern though, it will push devs over to Ubuntu (the only other Linux flavor I've seen around here) probably, and quickly the friction in government against non-Red Hat Linux flavors will drop. As an example of that friction, last I looked Red Hat published themselves guides and scripts to get CentOS stig'ed to the level we need here, while neither Ubuntu nor anyone else had automated it.
I, personally, am not a paying customer but I work for people that are using RHEL for _everything_. I would spin-up CentOS VMs for other things just because I am familiar with RHEL.
Now that that way of working is essentially dead, I have no reason to recommend or use RHEL in any capacity.
Also, the most dependable docker images are Debian based, Google distroless image is Debian based, most single board computer's OS are Debian based, Tails, Kali, Ubuntu, all Debian based and the list goes on and on.
It's the most mature and stable disto out there and is reference in the Linux world for more than 27 years.
EDIT: ...and is more up-to-date than CentOS, I just checked and CentOS seems to be using an old/non-LTS kernel :/
I wish there was something like Ubuntu’s LTS, paid support, and release cycle, but with plain old Debian as the OS. Ubuntu’s “innovation” has been negative value since the beginning of time. Oh snap!
I have been vaugely following this and I am likely going to get it in the neck here, but this seems to be people are upset that they can no longer run their commerical business on something for free, as the other business providing the free thing were getting nothing in return, so they want to change it so that at least get something back in return (which is still not monetary, but help fixng bugs).
Well, RHEL consists of open source software. While not all of that software is copyleft, it has AFAIK always been part of the reality and the very premise of Red Hat that someone else can take that, too.
If a community project itself stopped providing what it has been providing for free, people would probably be disappointed if they had started relying on what was provided. They might also become wary of relying on community-supported projects. But they would probably not be upset in the sense of being angry. Most people wouldn't feel entitled enough for that.
I think what makes people upset in this case is that Red Hat sort of adopted the project, giving it a home, and then stopping to provide what that project had been providing for people outside of Red Hat.
I have no problem with companies saying “if you want this, you have to pay”. I have a problem with a company saying “It will be free, with full security updates until 2029” ( https://archive.is/7Qmtw ) then turning around and saying “Just kidding, it’s only going to be supported until the end of 2021”.
Pretty much. I see CentOS on tons of commercial virtual and network appliances in enterprise environments that bring in a lot of revenue, but Red Hat doesn't see any of it. I'm not shocked if Red Hat sees little value in powering major commercial distributions for free.
After a bit of "anger" I finally switched over to CentOS streams and can't be more happy with it. I guess it's wait and see how things are going to pan out, i have the luxury to be able to afford some downtime if things are not working well...
I was suggested to try Oracle Linux, but I will always trust more Redhat than Oracle..
Not defending RedHat here, but in practical terms doesn't this just mean switching to Rocky linux when it's time to upgrade? Presumably Rocky will have a year to prepare a migration path.
There are famous short-seller investors circling IBM for exaggerated earnings projections. Redhat could very well end up in someone else's hands sooner than later.
Not too surprising, considering how container-based hosting (Kubernetes and friends) are taking off. The market for old-school all-inclusive Linux server distributions is bound to shrink, and it makes little sense for Red Hat/IBM to have two of them.
Because they previously needed it in order to do open-source work that was based on RHEL instead of Fedora, and because as a volunteer-run project, it was not being super responsively maintained. Here is a comment from a long-time Red Hat engineer explaining it: https://news.ycombinator.com/item?id=25351295
Short term: Amazon pays licensing fees to RedHat to keep the product going, if necessary. I don't know if they're at that stage yet, and big companies can have whatever arrangements they would like between them behind the curtains.
Long term: Amazon develops a new Distro unbeholden to RedHat/IBM. Only a matter of time.
CentOS was acquired by RedHat because it was eating into their market.
So, I'm sure that many customers who have been using CentOS just because it existed have no problems upgrading to RHEL because it's such a small part of their budget that it doesn't matter.
Meanwhile, CentOS did serve a purpose and those who absolutely don't want to upgrade to RHEL or are looking to start out without a subscription will be looking at other providers such as Project Lenix or Rocky Linux.
I think this could have been handled better by IBM. They've made a complete mess out of this by introducing CentOS stream and dropping CentOS. There is value in keeping non-paying users in your own ecosystem. Canonical handles it pretty well.
CentOS was acquired as a marketing tool to sell support contracts.
RedHat was acquired by IBM because selling support contracts no longer works as an open source software business model. It was a rescue operation (on both sides), really.
In a twisted way, it's a success story for OSS. The software is so good and its main users so competent that the support contract has moved from being the value to the value-add. (aside)
Unfortunately, this is the same thing driving the adoption of all of these new split licenses, which I don't think is a positive direction.
(aside): This is actually the second demonstration of this. The first was when Rackspace got rid of its world-class support organization in a race to compete on costs.
potential value in the long-term. but real liability in the near-term.
if the company doesn't want to, or thinks that it can't tolerate that risk over whatever the difference between near & long is, they'll cut the liability loose.
to me, this hints at instability in the market and IT sector, which, c'mon, we already know about. so it's likely more of an admission that times are tough, and RHEL/IBM is looking out for themselves.
RedHat flat out made a pinky promise that CentOS 8 would be supported with security updates until 2029: https://archive.is/7Qmtw I installed CentOS 8 on my home network (a non-trivial task) with the understanding things would still be secure in 2029. RedHat broke their pinky promise: CentOS 8 will no longer be supported after 2021. That in mind, while I have been using RedHat since the dot-com days, I am now a Ubuntu user.
Breaking a promise like this is non-trivial, and I have stopped using RedHat as a result. If RedHat had said “We’re not going to release CentOS 8” or if they had said from the beginning “CentOS 8 will only be supported until the end of 2021”, I would be singing a very different tune right now.