CentOS Stream 9: Understanding the new Red Hat OS release for non-Red-Hat-type people
- Reference: 1639177448
- News link: https://www.theregister.co.uk/2021/12/10/centos_stream_9/
- Source link:
This is the second release of the new CentOS Stream distro, and presumably the IBM subsidiary hopes it will offer a more appealing migration path for CentOS users than for them to jump ship.
Notably, in CentOS Stream 8, RH's [3]Application Streams – analogous to [4]Fedora's "Modularity" – were mandatory, but they're optional in 9.
[5]
This is a [6]big deal in the Red Hat world , but can be mysterious to the millions of non-Red Hat Linux users. Since it seems to please Red Hat to imagine that Red Hat is the entirety of the Linux world, its official materials don't really give you any context, so The Register will try to translate for you.
Executive historical summary
The [7]original Red Hat Linux (RHL) was released in May 1995, making it one of the oldest distros, but the company killed it off after version 9 in 2003. RHL was replaced with Red Hat Enterprise Linux (RHEL), which became the company's only supported distro. The word supported is key here.
The pre-existing Fedora project, formerly a third-party repo for RHL add-ons, got promoted to being the free, unsupported, community distro: no paid support, updates for [8]13 months , a new version roughly [9]twice a year, and upgrades from one version to the next could be tricky, although that's much better now.
[10]
[11]
Fedora tends to be pretty bleeding-edge compared to most regular-release-cycle distros – that's why the codeword "innovative" features heavily in the project's [12]mission statement . The Reg has [13]liked Fedora for [14]a long time . The new shiny that works in Fedora, when it ends up boring and stable, goes into RHEL. As RH puts it, Fedora is "upstream" of RHEL.
RHEL, on the other hand, was and is commercial: you can only get RHEL by buying it, which in reality means buying a support contract. But it's still FOSS, meaning that RH is legally required to make the source code available. So anyone – not just customers – could download Red Hat's source code packages for each and every package in RHEL, free of charge, and recompile them all to build a free clone of RHEL.
[15]
Various third parties started doing this and producing independent distros which were RHEL-compatible, such as Scientific Linux (from [16]FermiLab ) and White Box Linux (from the Beauregard Public Library in Louisiana). The biggest of these, CentOS (Community Enterprise OS), started in 2004.
The idea of these distros is that they are basically identical to RHEL, but with the names changed. Not just the same commands, but the same versions of the same binaries, with the same config files, in the same directories, for perfect compatibility. Whatever fixes RHEL got, soon afterwards the rebuilds got too. You could learn using a freebie prototype and test on it, but deploy on the real thing. Theoretically, if you didn't want to pay for lots of RHEL licences, you could pay for one copy, get official support for it, but run all your other boxes on CentOS, and save a packet. You only get official support for one copy, but all the same methods and tools work on all of them.
Oh, and in 2006 that well-known friend of FOSS, Oracle, got into the game by [17]making another such rebuild . This is often known by the name of its (paid, optional, and significantly cheaper than the official Red Hat Network) support service: Oracle Unbreakable Linux.
[18]
This was all fine, clear, and easy to understand. Sometimes, RH [19]tried to make life difficult for the cloners, but they survived. Despite them, RHEL has been a huge success – by 2011 RH was [20]the first billion-dollar Linux vendor .
The beginning of the end
Then in 2014 RH did something very strange: it [21]brought CentOS in-house . So now the leading freebie was kinda-sorta officially sanctioned. Unsupported, but given the nod. Please pay for RHEL for your production boxes, but if you'd rather not, you could run CentOS Linux for nothing.
This was, obviously, very good for CentOS – but very bad for the other rebuilds, and as a result, most of them (except Oracle) shut down.
This left RH with an odd range of distro offerings. Before 2014, the positioning was clear: freeloaders got Fedora, and needed to upgrade regularly. If you wanted support, you paid for RHEL. After 2014: if you want stability and support, please buy RHEL; if you don't want to pay, well, there's our fast-moving, free community distro… or, you could have this nice stable distro that's identical to RHEL, and it's free, you just get no support. Effectively, the company offered a free product specifically designed to compete with its own commercial flagship.
Compare this to SUSE's easier-to-understand proposition: you can [22]download a complete, fully functional evaluation version of any of its products, but you only get a [23]60-day trial period. After that, no more updates.
Or run the free openSUSE [24]Leap , which has the same release cycle as SUSE Linux Enterprise. If you want support, for a price you can [25]migrate from Leap direct to SUSE Linux Enterprise.
[26]Alma and Rocky Linux release 8.5 builds, Rocky catches up with secure boot
[27]Red Hat 8.5 released with SQL Server and .NET 6 ... this is Linux, right?
[28]CentOS Stream^W^W Red Hat Enterprise Linux 9 emerges in beta form
[29]Fedora 35 is out: GNOME 41 desktop, polished UI, easier-to-install closed-source apps
It took RH about five years, but eventually it appeared to realise this. In the company's terms, CentOS Linux was "downstream" of RHEL; although the word "community" is prominent on the CentOS [30]site , there wasn't any way to contribute to it: it was a straightforward rebuild of RHEL.
The first visible result was a new direction for CentOS, with CentOS Linux 9 being cancelled and CentOS Linux itself being replaced by [31]CentOS Stream in 2019 .
Rather than duplicating RHEL's release cycle, the new Stream edition gets a continuous stream of updates. Critically, this means it's no longer a one-to-one exact duplicate of RHEL; from the sales point of view, it's no longer competing directly with its parent product.
For the company, this is all good. Presumably, it didn't tempt many CentOS Linux users to cross the Stream, because [32]just five months later , the company announced that the end of life for CentOS Linux 8 was dropping from 2029 to 2021 – 36 months [33]sooner than the version 7.
As is its wont, RH prefers to phrase this in terms of upstreams, downstreams, and communities. Before, RHEL was downstream of Fedora, and CentOS Linux was downstream of RHEL. Now, CentOS Stream is downstream of Fedora, and RHEL is downstream of CentOS Stream.
The company's Rich Bowen told us: "From my perspective as a community manager, CentOS Linux was not open source, was not a collaborative project, and was not really a community in any meaningful way. CentOS Stream gives us the opportunity to make CentOS into an actual community project, with an actual contributor/collaborator path."
Which is laudable and entirely understandable. In principle, migrating from CentOS Linux 8 to CentOS Stream 8 is as simple as two commands:
dnf swap centos-linux-repos centos-stream-repos
dnf distro-sync
RH also offers a [34]tool to migrate from CentOS Linux direct to RHEL – which looks considerably simpler than the [35]documented upgrade path from RHEL 7 to 8.
Interestingly, Bowen told us: "If people want or need (or think they do) a 1:1 RHEL rebuild… we are working very closely with [36]Alma to make sure that they have what they need."
This even extends to major version upgrades: "At this time, we don't have a Stream 8 to 9 migration tool. However, the folks at Alma are working on [37]ELevate which will have that as a feature."
Red Hat made a mistake when it adopted CentOS. CentOS Linux fans may disagree, but for all the grief killing it off has caused, the current steps are fixing the mess the company made for itself. RH is doing all it can to give CentOS Linux users a choice of migration paths, including smoothing the way for people to jump to Stream.
Stream itself should offer smaller, less dramatic updates and be more competitive with the other server distros out there, for which Fedora was frankly never an ideal choice. It also gives the denizens of the Hativerse a way to gain some input into the direction of RHEL. And it's especially good to see [38]Big Purple cooperating with other distros. ®
Get our [39]Tech Resources
[1] https://blog.centos.org/2021/12/introducing-centos-stream-9/
[2] https://www.theregister.com/2020/12/09/centos_red_hat/
[3] https://www.redhat.com/en/blog/introduction-appstreams-and-modules-red-hat-enterprise-linux
[4] https://www.theregister.com/2016/10/19/fedora_facelift/
[5] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/oses&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=2&c=2YbQwigiqtT6zYRDPqmSk-gAAABE&t=ct%3Dns%26unitnum%3D2%26raptor%3Dcondor%26pos%3Dtop%26test%3D0
[6] https://www.theregister.com/2020/07/29/red_hat_enterprise_linux_83/
[7] https://www.theregister.com/2021/08/25/30_years_of_linux_red_hat/
[8] https://docs.fedoraproject.org/en-US/releases/lifecycle/
[9] https://docs.fedoraproject.org/en-US/releases/
[10] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/oses&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44YbQwigiqtT6zYRDPqmSk-gAAABE&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0
[11] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/oses&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=3&c=33YbQwigiqtT6zYRDPqmSk-gAAABE&t=ct%3Dns%26unitnum%3D3%26raptor%3Deagle%26pos%3Dmid%26test%3D0
[12] https://docs.fedoraproject.org/en-US/project/
[13] https://www.theregister.com/2017/07/24/fedora_26_review/
[14] https://www.theregister.com/2008/11/25/fedora_10_review/
[15] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/oses&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44YbQwigiqtT6zYRDPqmSk-gAAABE&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0
[16] https://www.theregister.com/2004/02/02/uk_teen_escapes_jail/
[17] https://www.theregister.com/2006/10/31/oracle_server_linux/
[18] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/oses&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=3&c=33YbQwigiqtT6zYRDPqmSk-gAAABE&t=ct%3Dns%26unitnum%3D3%26raptor%3Deagle%26pos%3Dmid%26test%3D0
[19] https://www.theregister.com/2011/03/04/red_hat_twarts_oracle_and_novell_with_change_to_source_code_packaging/
[20] https://www.theregister.com/2011/03/24/redhat_q4_f2011_numbers/
[21] https://www.theregister.com/2014/01/08/red_hat_to_team_up_with_communitybased_rhel_lookalike_centos/
[22] https://www.suse.com/download/
[23] https://www.suse.com/c/how-to-put-the-suse-linux-enterprise-server-eval-to-the-test-1/
[24] https://get.opensuse.org/leap/
[25] https://www.suse.com/c/closing-the-leap-gap-src/
[26] https://www.theregister.com/2021/11/16/alma_and_rocky_linux_release/
[27] https://www.theregister.com/2021/11/11/red_hat_8_5/
[28] https://www.theregister.com/2021/11/04/rhel_9_beta/
[29] https://www.theregister.com/2021/11/02/fedora_35/
[30] https://www.centos.org/
[31] https://www.theregister.com/2019/09/25/centos_stream_linux_distro_release/
[32] https://www.theregister.com/2021/02/08/centos_board_qa/
[33] https://endoflife.software/operating-systems/linux/centos
[34] https://www.redhat.com/en/blog/introduction-convert2rhel-now-officially-supported-convert-rhel-systems-rhel
[35] https://access.redhat.com/documentation/en-us/red_hat_enterprise_linux/8/html-single/upgrading_from_rhel_7_to_rhel_8/index
[36] https://www.theregister.com/2021/03/30/almalinux_project_issues_first_stable_release/
[37] https://almalinux.org/elevate
[38] https://www.theregister.com/2019/07/09/ibm_red_hat_acquisition/
[39] https://whitepapers.theregister.com/
Re: No thanks.
Both my company and all the sane clients are migrating to other distros asap instead of any RedHat derivatives. Who knows what else these absolute morons invent next? Betting on Titanic not sinking? You lose.
IBM is trying to squeeze money from CentOS-users and the morons in RedHat are saying a lot of horses**it to cover it up.
They've miserably failed: The "gimme money" is so blatant no amount of talk can hide that.
When millions of CentOS users disappear to other distros, RedHat have to write all their own fixes and *that* costs a boatload of money, work they've so far got for free.
That means also that the quality of RHEL and all derivatives will plummet: It's obviously clear that IBM doesn't let them have enough people to keep the quality: IBM wants *profit*.
Profit is the opposite of quality.
Also, RedHat doesn't have anything to take it from. Too bad, that means that the headcount will plummet. A lot. Death spiral is guaranteed and we don't participate in that. It was nice while it lasted.
So far I haven't seen a single fix done by RedHat personnel: It's always a community favor done in CentOS, not RHEL. Obviously RedHat bean counter refuses to see that: He's paid not to acknowledge community in the CentOS.
No-one sane runs Streams in any business use
"including smoothing the way for people to jump to Stream."
No-one sane uses Streams in any kind of actual work. it might be OK at home computer, as Fedora replacement. "Continuous changes" means there's no way to *test* anything against anything else for a significant time. Typically cycle from testing OS-version in development to production takes months. As it should.
Now: have a stream of changes weekly or daily with no version numbers or *actual package of rpms*: How do you know you've same set of packages you tested against? By listing them all, one by one? Utter lunacy.
Just like putting Fedora, with alpha-level of stuff into production. It's just not done. Obviously the ElReg hasn't realized a bit of that by statement above: It really doesn't make any sense: it's like running windows 11 home edition as business critical application platform: No-one sane does that.
Re: No-one sane runs Streams in any business use
"Just like putting Fedora, with alpha-level of stuff into production. It's just not done."
Hah, you might be surprised about that.
Alternate view.
Bringing CentOS in house killed off all of the other RHEL clones because who wouldn't use the (and as near as doesn't matter also guaranteed) official clone.
At some lightbulb moment RH spotted this and realized by moving CentOS from utterly stable to slightly better than Beta would give all the users two options, Pay to stay the same or try a migration with it's unknown level of risk - aka the Microsoft model.
Re: Alternate view.
"Pay to stay the same or try a migration with it's unknown level of risk -"
So, summarized: f**k off to the non-paying customers providing 80% of the fixes.
When those those customers leave the quality *will* plummet as it's obvious IBM demands profits and it's not going to invest a penny more to RedHat.
With millions of CentOS users missing, they have to write fixes themselves. That's not cheap.
Which means it's not "stay the same" as the company itself has totally different business model now: "Squeeze as much as you can from customers". Same as IBM. Or Microsoft.
No-one is a customer of either of them if they have a choice. CentOS users have several choices: Migrating from RHEL/CentOS to other distros is not *that* hard. Probably easier than migration from CentOS 7 to 8. I think commenter forgot that even staying with RH *is not trivial*.
Many of our customers are already doing that in development environments, piece of cake for developers. Those migrations will eventually flow into production and then it's goodbye RH: It was fun while it lasted.
Downward spiral for RH started at RHEL7 when they decided to incorporate systemd, a spawn from Microsoft hell with binary registry and all. This is just more of the same: microsoftication of whole company and their products, providing shitty, expensive products just by market share.
I used to prefer CentOS for legacy LAMP servers and stability and then Debian or Ubuntu for less stable but other more interesting projects. Like most sane I.T peeps you use whatever tool is appropriate for a job. Windows for games, Linux for servers, Wii-U for Zelda BOTW, Macs for pretending to know IT, etc, I have a toolbox full of all of the above and each are great for something different. CentOS' rug pull forced my hand in to the "never use RH again" camp despite being a yum trumps apt guy. Now firmly Ubuntu LTS. I have also sat here and consumed far too many beers. NN all you nerds, live long and prosper Luke
yum trumps apt ... I'll agree on that.
But RH basically committed a suicide when they sold themselves to IBM.
Owners obviously got a truckload of money but, all of the greedy vultures out there, to *IBM*?
As bad as selling whole thing to Oracle or Microsoft.
Anyway, the "we want profit and we want it *now*!" from the new owners is showing again. And, as usual, the beancounters in IBM have *no idea* how the relation between RHEL and CentOS was actually working.
Nor does the RH spokesperson here, either. Or he lies to present some made-up BS to cover the planned *money grab* behind this.
These morons obviously believe that most, or even significant part, of CentOS users move to RHEL and pay the license.
While it's trivial to move to other distro. They must be *really* stupid at IBM. And greedy as hell. But that's a given when it's the IBM.
"CentOS Linux was not open source, was not a collaborative project, and was not really a community in any meaningful way..."
Well, first, it certainly was open source; that's why it was able to exist in the first place. That it was neither collaborative nor a community is true, and gets us to the heart of the matter: what people want from software.
Enterprise IT shops want software that their bosses tell them to buy, and they want a golden support contract with it no matter the cost. They are on the hook for delivering reliable services and when things go TITSUP they want a throat to choke and they want it now. Evolving the software itself slowly isn't merely acceptable, it's preferable. Whether there's any community around it is largely irrelevant; they're paying for support so they don't need a disqus forum somewhere, and they don't care whether kids at home can play along.
The kids at home don't care about support because they couldn't afford it anyway, and it makes little sense to buy support for something that's not being used to generate revenue. A community-based alternative is nice when things go wrong, but there are a hundred other free, open source OSs out there so if CentOS were really broken (which of course meant that RHEL was broken too and those enterprise IT shops were busy choking RH throats), they could always use something else.
Everyone else really just wants it to cost nothing. If they can have something identical to RHEL that costs nothing, that's best -- they're freeloading on the enterprise IT shops when it comes to getting things fixed. But they don't have time to contribute anything anyway; these are mom and pop shops with one or two "IT guys" who are usually too busy to do anything but extinguish the biggest fire they happen to know about. They haven't got any budget so they aren't going to buy support from anyone, either.
This is the classic "we're losing out on all this business!" story we've heard time and again from pretty much everyone selling goods with a marginal cost of zero who find people freeloading. In reality, the mere fact that a lot of people were freeloading (in terms of support, where a support ticket led to a software fix) doesn't mean there was much or even any lost business there. Those people weren't going to pay for your support contract anyway, and by ending their freeloading all RH/IBM have done is convince them to use someone else's (also free) product instead.
Is that good for the business? Bad? Mostly it's probably neither. Some will make the case that it's bad because it shrinks the body of professionals familiar with the RH tools and ecosystem, but the reality is that most IT pros adapt to whatever the shop that's hired them is using. I suppose it's good in that they don't need to bother building CentOS any more, but by creating Stream instead (which no one wants) they're actually investing more without getting anything from it, so that doesn't seem like much of an argument either. The revenue needle isn't going to move, because everyone who wanted RHEL support and was willing/able to pay for it was already doing so.
In the limit, new rebuilds will exist again and life will go on for everyone pretty much as it has, because most people care about neither support nor community: they're just using whatever seems most convenient to do their jobs, as long as it doesn't cost anything. It's a pile of packages of free software, not a church.
"Enterprise IT shops want software that their bosses tell them to buy, and they want a golden support contract with it no matter the cost. They are on the hook for delivering reliable services and when things go TITSUP they want a throat to choke and they want it now."
No. That's old school thinking and obsolete now when "golden support contract" from *anyone* doesn't provide actual support whatsoever. Those exist only to make money and only MBAs buy them, not IT people.
You get actual support from Microsoft (or RHEL) if you are DOJ or DOD. Otherwise: Good luck.
Even for enterprise level of users: If internal support can't solve it, it's not solved.
Then you just want reliable services from lowest bidder and CentOS has been by far the best on that since 1990s as tens of thousands of enterprises use RHEL and CentOS is 1:1 clone.
It's confirmed to be a stable platform which really doesn't *need* a support contract. Which obviously is something RH is going to change by forcing the Streams, i.e. Fedora, i.e. alpha level packets to business users too.
Unless they pay RHEL-tax. One of my client calculated that whole development team can do a month work for migrating to other distro by the yearly price of RHEL support contract.
Obviously, it won't take a month of work to migrate from CentOS. RH/IBM bean counters haven't realized that, at all.
They will.
" The revenue needle isn't going to move, because everyone who wanted RHEL support and was willing/able to pay for it was already doing so."
Yes. I think this is a major point RH/IBM morons aren't getting at all. Not one CentOS 7 user is *migrating* to RHEL 8, it costs too much. As money and as work.
Streams is by definition unstable stuff, like Fedora: No-one wants that even near to production.
What's left? Migrate to Debian, perhaps.
..."golden support contract" from *anyone* doesn't provide actual support whatsoever.
Of course it doesn't, but the big IT shops with big budgets still pay for them. We all laugh at IBM's ever-shrinking revenue, but people are still giving them $75 billion a year and most of that is for support contracts. Almost all of the RH reporting segment is support contracts and that's been growing. So it's not what reality is but what the customers think it is that matters. They have money and they want to spend it on this, so they do. The support is nothing (how could it be? IBM fire their best engineers and hire mostly the dregs) but the throat to choke is worth a lot to an IT director. That they could get better support for 1/10 the price by hiring a few generalist engineers is not within their ability to consider, much less accept. Obsolete thinking? Sure is!
The reality that I think most rank and file operations staff have come to understand is that if you are not a government behemoth or one of "the hyperscalers", you don't get any kind of customer service at all from anyone for anything ever, whether you pay for it or not. It's no more true in IT than anywhere else; that's pretty much just life in the 21st century: no one is willing to work for anything if it's not going to result in a tremendous immediate windfall. Everyone else exists only to be locked into some kind of useless subscription and sucked dry. The idea of patiently caring for thousands of customers who give you a profit and expect value in return just sounds too much like work.
" If they can have something identical to RHEL that costs nothing, that's best -- they're freeloading on the enterprise IT shops when it comes to getting things fixed"
You seem to forget that said "freeloaders" provide almost all bug reports and most of the bug fixes as community service to *CentOS*. RHEL gets them as a byproduct.
For free and as such, it's the RH which is actual freeloader and community does most of the work here. Now RH is ditching that work and it will bite them in the bottom later.
No thanks.
I'm migrating off of CentOS starting in Jan 2023. I've been using CenOS, unsupported, since version 5.5.
icon isn't a burning bridge, but it'll do.
And while I may not have been able to contribute code, I've opened my share of bug reports against CentOS in the last decade and I hope that helped someone other than just me.