Wieers: Is 7 years of RHEL support still sufficient ? [LWN.net]

Dag Wieers wonders whether seven years of support is enough for Red Hat Enterprise Linux (RHEL). He looks at the increasing time span between RHEL major releases and notes that the support window may close before customers are ready. "Let me explain what I mean. When Red Hat released RHEL2.1, seven years of support was perfect, seemed more than one would want. RHEL3 came 18 months after RHEL2.1 and after one year of testing RHEL3 and 3rd party integration new systems could be deployed, giving you 6 years of support. Your hardware would usually not outlive the operating system support."

to post comments

Wieers: Is 7 years of RHEL support still sufficient ?

Posted Sep 7, 2009 22:20 UTC (Mon) by tialaramex (subscriber, #21167) [Link] (4 responses)

Realistically, the lifespan of RHEL is set by customer interest. The existing lifespan numbers are a commitment, nothing short of crippling financial problems would induce Red Hat to break it, but they don't prevent commercial interests from dictating an extension, even a very long one. This is the same as happens for Microsoft Windows. Microsoft has repeatedly extended the lifespan of popular products in the face of customer demand or even just to buy time for a delayed replacement.

Wieers: Is 7 years of RHEL support still sufficient ?

Posted Sep 8, 2009 4:07 UTC (Tue) by drag (guest, #31333) [Link] (3 responses)

The way it works for Microsoft, and probably is the correct form of action, is that if you want real support for items beyond the normal lifespan you pay out the nose for it.

Something like $5000 for a hotpatch to fix a problem. If a customer would not mind paying money like that for a bug fix I doubt many companies would turn them down. If a company does not think that it's worth it then it's probably worth it for them to upgrade their machines.

Mind you that I expect about 5 years is the about the reliable lifespan of hardware...

Wieers: Is 7 years of RHEL support still sufficient ?

Posted Sep 9, 2009 11:54 UTC (Wed) by nix (subscriber, #2304) [Link] (1 responses)

If five years is the reliable lifespan of hardware then it's gone down a *lot* recently. The mean age-on-failure of my hardware that survived infancy and was retired due to hardware failure (as opposed to, say, being too obsolete to be useful for anything) is about ten years. (Generally the disks or fans go first, but fans are easily replaceable. It is true that you tend to lose a fan or so in a five-year span.)

Of course hardware bought from about 2002-2005 is liable to fail *much* faster due to the electrolytic capacitor screwup, but hopefully that's going to be a one-off.

Wieers: Is 7 years of RHEL support still sufficient ?

Posted Sep 9, 2009 20:00 UTC (Wed) by Baylink (guest, #755) [Link]

Wow; that's weird. Not only did I just discover yesterday that a lot of used GX/270s I just bought suffer from that...

but I just re-read my RISKS posting from that period, entitled "The Great Capacitor Scare of 2003". Literally, not 5 minutes ago.

Does anyone know if Dell *did* do a long-term recall on those machines?

Wieers: Is 7 years of RHEL support still sufficient ?

Posted Sep 9, 2009 14:16 UTC (Wed) by k8to (guest, #15413) [Link]

5000 for a hot patch isn't really so great, unless you're providing the same fix to a lot of customers. 15 hours of support time to validate, clarify the problem, 10 hours of dev time to investigate and find the exact problem. 2-4 hours to do the commit build production dance, and then an hour or so to trigger the automated tests and qualify the thing (if you only produce the one build).

Let's say 18 hours. Probably your workers are making around 35 average. Employee overhead cost are around double that. So 1260 of your 5000 goes out the door before you you make a mistake. Then it's worse. You may have to repeat the cycle.

Factor in all the overhead to keep people trained, systems working, etc for a large number of old builds. It's profitable, but it's not a gravy train.

Wieers: Is 7 years of RHEL support still sufficient ?

Posted Sep 7, 2009 23:22 UTC (Mon) by mrecho (guest, #53167) [Link] (6 responses)

Version support to me is how long we keep patching old software to keep it secure, and to keep it at the same api/abi version.
That is a lot of work. Im just going to leave out Kernel and Driver stuff. Don't even want to get into that.

When you just keep patching security issues, your not fixing other issues. Your not giving the customer the fixed, updated versions. Your left with really old software.
The only good I see out of this is when you buy an application that is soo tied to a single version of an api, it will always work. I call that bad programming. If that program needs that very single version of the api, it needs to install its own copy of it, and use that one, not the systems. Now if the Distro is smart enough, it will have a way of having a few different versions installed, and not do a lot of sym links.
When it comes to php,mysql,apache,perl,python,postgres,xorg your stuck using multi year old applications, because there 'stable'. And now you have to wait a few years till the new version. Note the version of php for RHEL 5. I just don't get this 'stable' thing. I guess its because of 3rd party applications not knowing how to use libs, and or the distributions them self's for not providing a way to have a few different versions installed. I think its a combination of both.
So 7 years of having the same version of applications, reminds me of Microsoft. And 3rd party application makers need to change, along with distributions.

Wieers: Is 7 years of RHEL support still sufficient ?

Posted Sep 7, 2009 23:41 UTC (Mon) by dag- (guest, #30207) [Link] (2 responses)

Third party applications are not the essence of the problem though, they only make it slightly more difficult. Even when you use 3rd party Open Source applications, you have to integrate and test your solution and it is not uncommon that this takes up a year.

And while virtualization helps to extend the life of a system, the operating system (in this case RHEL) is becoming the limitation. Not that any other Linux distribution is doing any better currently :-)

So I expect a customer demand for extending support for RHEL.

Long-lived guest OSs

Posted Sep 8, 2009 14:09 UTC (Tue) by dmarti (subscriber, #11625) [Link] (1 responses)

You can still license VMS. Somebody is going to be maintaining early RHEL versions, at least as guest OSs, for a long time.

Long-lived guest OSs

Posted Sep 9, 2009 0:50 UTC (Wed) by quotemstr (subscriber, #45331) [Link]

I can imagine, one hundred years hence, a hacker being hired to fix a bug in some industrial control system that's "always worked until now", poking around, and finding a ten-system deep stack of virtual machines with DOS 3.22 at the bottom.

Wieers: Is 7 years of RHEL support still sufficient ?

Posted Sep 8, 2009 0:54 UTC (Tue) by dowdle (subscriber, #659) [Link]

Businesses invest a lot of money in IT hardware and software... and often it takes a while to get a return on that investment. Who cares if software is "old" as long as it works, performs well enough, has the features that are needed, and is secure. If you have to keep upgrading that is a sign the software didn't meet your needs in the first place.

Maintaining an existing version should be much less work than adding new features and adapting to newer libraries... or moving to a new OS/OS version... unless what you were working with was sub-optimal. Hardware is fast enough now, operating systems are reasonably mature, as are development environments.

Ideally, hardware and software should be able to last between 10-15 years... but maybe that is just a pipe dream of mine. Unfortunately most of the software industry operates on a business model where they have to sell their customers something new every few years... even if their customers don't use a significant amount of features in the current version... just keep adding more features trying to avoid approaching unmaintainability.

Wieers: Is 7 years of RHEL support still sufficient ?

Posted Sep 8, 2009 12:59 UTC (Tue) by qg6te2 (guest, #52587) [Link] (1 responses)

your vs you're

(yes it's pedantic, but I can take only so much sloppy spelling in one day)

Wieers: Is 7 years of RHEL support still sufficient ?

Posted Sep 9, 2009 13:43 UTC (Wed) by michel (guest, #10186) [Link]

http://www.apostropheabuse.com/2009/07/confused-about-you...

Wieers: Is 7 years of RHEL support still sufficient ?

Posted Sep 8, 2009 7:18 UTC (Tue) by wzzrd (guest, #12309) [Link] (2 responses)

Dag probably has a good point: businesses like this. Longer support, a longer status quo is good is you run a database (or some other enterprise app) on this platform. The only problem is that though stretching Red Hat support to nine years is a good idea, it is moot unless companies like Oracle and SAP do the same.

For Red Hat, it does not financially matter whether or not people buy the new version of the OS. What matters is subscriptions, which cost the same for all versions, I think. For Oracle and SAP though, it does matter financially whether you (e.g.) keep running Oracle 10g for a few more years before migrating to Oracle 11. Therefore, I think the odds other companies will follow suit are pretty small. And hence, I'm not sure it will matter much whether or not Red Hat does this...

Versions and Oracle

Posted Sep 8, 2009 7:37 UTC (Tue) by lakeland (guest, #1157) [Link]

I don't believe this is correct.

If you've purchased and maintained an Oracle license, then Oracle really don't care whether you
choose to use that license to run Oracle 8 on an extra CPU of an old server, or install Oracle 11
on a new single CPU (dual core) server.

I know SAS is the same. I don't know SAP.

Wieers: Is 7 years of RHEL support still sufficient ?

Posted Sep 8, 2009 8:18 UTC (Tue) by lmb (subscriber, #39048) [Link]

While it doesn't matter from a revenue perspective which software version a user runs (SLES is the same), the older the software, the more costly the maintenance and support side becomes due to the widening gap to mainline. People working on maintenance for zombie distributions also don't work on new functionality and improved current fixes.

So from a P&L perspective, it matters.

Wieers: Is 7 years of RHEL support still sufficient ?

Posted Sep 8, 2009 9:56 UTC (Tue) by rahulsundaram (subscriber, #21946) [Link] (4 responses)

If 7 years is not sufficient, go for a 10 year plan

http://www.redhat.com/promo/mc_program/

If not, vendors will extend the lifecycle for some customers depending on the premium.

Wieers: Is 7 years of RHEL support still sufficient ?

Posted Sep 8, 2009 20:23 UTC (Tue) by dag- (guest, #30207) [Link] (3 responses)

That's good information. I feel ashamed to say that I didn't know this (Advanced Mission Critical Program) was an option today, otherwise I would have mentioned it in the original article. (I will probably update it anyway)

But the fact is that the longer release cycle affects everyone the same way, regardless if it is mission critical or not. And for the companies I work(ed) for I can see this as a real problem. Installing new servers to be reinstalled in 3 or 2.5 years has a lower ROI than 5 or 6 years.

On the other hand the fact that Red Hat also offers this (expensive) 10 year support program could be a (strategical) reason for not extending the normal offering. Which could slow down their growth.

Fact is that it's hard to see what the causes and effects are when Red Hat is rapidly growing the way they are.

Wieers: Is 7 years of RHEL support still sufficient ?

Posted Sep 8, 2009 21:44 UTC (Tue) by rahulsundaram (subscriber, #21946) [Link] (2 responses)

Note that the program isn't just about a longer lifecycle. If your concern is about the gap between the releases rather than how long it is maintained, then you need to highlight that specifically.

Red Hat like other vendors would be willing to extend the lifecycle for specific customers on a case by case basis. If the regular subscription should be itself extended for a few more years, that demand needs to be from paying customers. Are they asking for that?

Wieers: Is 7 years of RHEL support still sufficient ?

Posted Sep 8, 2009 22:22 UTC (Tue) by dag- (guest, #30207) [Link] (1 responses)

It makes sense that a Red Hat customer in this situation should probably talk to their Red Hat contact. I will do my part to talk to our global team to communicate to Red Hat before my contract ends.

But as a freelancer, I only do one contract (one customer) at a time. And I am usually not the person dealing with Red Hat (except for technical stuff). Blogging is a way to create the awareness before it becomes a problem.

In the end I am the person manoeuvring within the boundaries and limitations of hardware, operating system, applications, vendors, processes, policies, people... But I do love my job :-)

Wieers: Is 7 years of RHEL support still sufficient ?

Posted Sep 13, 2009 13:26 UTC (Sun) by rahulsundaram (subscriber, #21946) [Link]

The recent Red Hat Summit presentation repeatedly points out that custom extended lifecycle for specific customers is available

http://www.redhat.com/f/pdf/summit/tburke_1050_rhel_roadm...

Wieers: Is 7 years of RHEL support still sufficient ?

Posted Sep 8, 2009 15:39 UTC (Tue) by gbailey (subscriber, #58) [Link] (3 responses)

I think the point about the "increasing time span between RHEL major releases" hits the nail on the head for me. As a maintainer of both RHEL and CentOS servers, I'm hit more and more with the need for something I can run more recent applications on--not necessarily things like Oracle, but open source applications. All one has to do is troll through the CentOS or RPMForge mailing lists to get a sense of all the apps with newer versions than what's included with RHEL that people are having to port: subversion, PHP, MySQL, PostgreSQL, python, etc., just to name a few. I need something that's stable and has long term support (hence going with RHEL over Fedora). But it's getting harder to deploy open source applications that require newer versions of software than that which comes with RHEL.

Waiting for RHEL 6...

Wieers: Is 7 years of RHEL support still sufficient ?

Posted Sep 9, 2009 3:39 UTC (Wed) by motk (subscriber, #51120) [Link] (2 responses)

EPEL to the rescue!

Wieers: Is 7 years of RHEL support still sufficient ?

Posted Sep 9, 2009 16:15 UTC (Wed) by jbardin (guest, #54917) [Link] (1 responses)

epel has software that isn't included in the base RHEL release, but it doesn't provide newer versions of base packages.

Also, epel packages often commit to a set version. For example, git was added last year to epel, and will probably stay at version 1.5.5, but I want to use 1.6.4+, and so I'm back to packaging it myself.

Wieers: Is 7 years of RHEL support still sufficient ?

Posted Sep 13, 2009 13:22 UTC (Sun) by rahulsundaram (subscriber, #21946) [Link]

Read

https://www.redhat.com/archives/epel-devel-list/2009-Sept...

In particular, refer to

https://www.redhat.com/archives/epel-devel-list/2009-Sept...