The move to Linux, stymied by hardware...the server side...

If you thought installing Linux on a laptop was a fun discussion, have I got a new one for you. This comes straight out of the really, it should not be this hard category…the server side.

I have an HP Proliant 385 G2. It was running Fedora Core 7. It installed fine a year or so ago and was running fine, but I wanted to take it up to Core 10 and test drive the new features. Pretty simple. The server has a P400 RAID board, a pair of 72GB disks and four 146GB disks. The 72s are mirrored (RAID 1+0) and the 146s are in an RAID5 array. Pretty standard stuff right?

Boot up my Core 10 DVD, select my language and whamo, I get the message:

Unable to determine geometry of file/device /dev/cciss/c0d0. You should not use parted unless you REALLY know what you are doing.

My initial thought is that there was something wrong with my array configuration so I boot back into SmartStart and blow the arrays away and rebuild them, reboot and go through the process and we are back to the message. OK, maybe I can ignore it and move on, which I do. I spend about 45 minutes picking and choosing my packages and start the install and head home for the night.

The following day, I reboot the server and end up at a dead GRUB prompt. There is nothing you can do to move it beyond this point. It is at this point that I begin my Internet searching.

That was Tuesday. It is now Thursday afternoon, some 30 man hours later.

Since then I have read articles about large disks which, while interesting does not apply 1) because I don’t have a large disk in excess of 2 TB and 2) I am not using a DELL, but it is nice to see that I am at least not alone in having no immediate answers. I have read about bugs in Ubuntu and the various strings related to it and while interesting, does not solve my Fedora Core 10 issue, but it did pose some very interesting options to try.

It would seem that this fix, relating to the scsi_wait command is doing the trick, but only after a lot of machination after the fact. During the course of arriving at this article, I what seemed like several hundred pages of solutions, none really exactly what I needed. I patched, updated and flashed firmware, rebooted into Knoppix, partitioned with three other OSs, stood on one leg and generally cursed a blue streak. Good thing my office is in a bunker and removed from the rest of my coworkers.

The point here is that we keep trying to tell people Linux is the better mouse trap. I say it all the time because it has proven itself over and over again in real world use. But when something as simple as a commodity RAID controller in use in mainstream hardware chews up 30 hours of manpower time just to get the operating system installed and the process of getting there includes things like building and burning custom ISOs and enough byzantine commands to make perl readable, then maybe we need to step back a bit and ask ourselves what the goal really is? Do we want to provide a world class operating system, capable of supporting thousands of customers and millions of transactions, running on commodity systems or don’t we? Do we want to offer a stellar, rock hard, and truly viable alternative to Windows or do we just want to keep working in our little hobby shops?

This is very much about being smart. It is about the smart use of time and the money that time costs. I have seen, and believe, that Linux is the answer, that is is a world class operating system with all the capabilities that any IT shop could ask for. It is the better mouse trap. I have been using it since 1995 to solve a variety of problems that Novell and Microsoft could not do then and in many ways still cannot do today. I have seen Linux systems half the size of a Windows system run rings around the Windows box and still have CPU left over for other tasks. I am a believer. But after 30 hours of frustration, cussing, ISO burning, and hacking, even I begin to wonder if it is all worth it, when I know I can slap my copy of Server 2008 in the same box and be up and running in a couple of hours and management is willing to live with the shortcomings.

______________________

David Lane, KG4GIY is a member of Linux Journal's Editorial Advisory Panel and the Control Op for Linux Journal's Virtual Ham Shack

Comments

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.

Fedora 10 aka FC10

Anonymous's picture

____I think the author's point was that there should have been better diagnostics available. Or a way to do a bare bones install to flash media with an incremental upgrade to full function.
____Keep in mind readers that while they've dropped "Core" from the long name it is still present in the short name (FC10). And while it may be an experimental branch of the Red Hat distro that does not excuse crippling choices in the package management (some people like to load packages from the DVD after the initial install) nor FireFox-happy browser patching (some of us run SeaMonkey, Opera, Dillo, and Amaya).
____I've been flamed for criticizing Fedora 9. I may be trying FC10 in the near future to see if they've caught up with Mepis and/or OpenSUSE. That's right - I'm a distro fan's worst nightmare - a Linux vs. Linux tester. And someone who thinks KDE4 should have been held back three to six months (I wonder when Konqueror 4 is going to catch up to Konqueror 3).

Fedora!?!? AW, COME ON!!

Rick's picture

Fedora and/or CentOS is the worst thing you can possibly install in a server. I've had numerous issues with it, including outdated packages, nonstandard configuration files, missing repository keys (which I had to google for), it's a horrible mess.

I really don't understand how RedHat/Fedora earned their "this is a distro specialized for servers!" reputation. They even go to the extremes of ignoring the de-facto standard that synaptic is, and replacing it with their own defective package manager. It's the closest thing to Microsoft Windows in levels of frustration and imposing their own technologies on you. Whenever I hear of Fedora/RedHat, I think: "It's like Microsoft, but for Linux".

On the other hand, debian-based distros are so much more user friendly (and let's not ignore the fact that Lenny just went out).

something not right about these articles

Anonymous's picture

The author wrote a previous article on how he would rather run Linux inside a VM on Vista on his laptop than spend $20 for a supported wireless adapter, then has this big drama with Fedora, and does not tell us the solution to the problem. The P400 RAID board is notoriously flaky on all platforms, and is notorious for giving Linux admins fits. Did the author ever call HP support? I bet not. Then he uses this one incident as an indictment against all of Linux. While HP has some good Linux support in various product lines, it is not universal across all of their products.

This is a poor article, as so many of LJ's web articles are these days. Nothing useful, just a content-free tirade. As someone else commented, it seems that LJ is going for flamebait instead of content.

Diagnosis:

Anonymous's picture

The error appears to be in the interface between the server keyboard and the chair sitting next to it.

In other words:

1. You NEVER upgrade server just to screw around with it unless you are prepared to spend the time needed to get it back up and running.

2. You don't use an "unstable" operating system like Fedora unless you understand that it CAN and DOES change regularly. Requirements should always trump software selection.

If you were working in my department and wasted that much time complaining about something I could have told you would happen if you'd simply asked, you'd have been fired a long long time ago.

Windows Server 2008

EarlyMorningHours's picture

I'm not sure what would be more frustrating - the ordeal that you describe in this article or the multiple phone calls to multiple people for multiple weeks, possibly totally over 12 hours, just to get my Windows Server 2008 volume license keys from Microsoft - after they were paid several thousand dollars in licensing.

I agree with "Jim" above in

Anonymous's picture

I agree with "Jim" above in the comments. Well written. I agree with one thing of in this article....Linux is not perfect, but the references about "30 hours, etc.." weren't fair in the least especially when comparing later to MS 2008 when that is "supposedly" a production ready system while Fedora 10 is purported as a hobbyist/community OS and NOT production while as Jim mentioned, CentOS (Red Hat Enterprise Linux) is.

Linux isn't perfect but let's compare apples to apples, please!

been there, done that

turgut's picture

I have experienced lots of similar frustations - for example, an ASUS M51VA's display - it took me ages to get 3D working well.

I tell everyone that Linux will be hell the first month, but then it will work for YEARS without fixing.

I still prefer it over the lesser alternatives.

-turgut

It is quite amusing to see

Joakim's picture

It is quite amusing to see how much offence people here take when someone dare to question the greatness in Linux-system! Is it so hard to see the shortcommings that sometimes do exist.

I do agree with the writer here, this is quite big issue in my mind, a big factor in getting people to switch is the old Apple thing, "It just works". In my opinion, this should be one of the biggest concerns today in development, ordinary people don't want to spend that much time to get things working.

People are not alwaya geeks, and Linux should not only be for geeks!

THAT'S WHAT YOU GET FOR USING PROPRIETARY HARDWARE

Anonymous's picture

Why has no one here placed the blame where it should be? The problem was created by the use of a proprietary, closed piece of hardware. Using Windows on this machine would only compound the problem!

Release Notes - Bugs/Common Issues

Anonymous's picture

I don't know why Linux Journal allows you to post on this site. Your articles are always flamebait. Or maybe that's what Linux Journal is looking for now.

Try reading the Release Notes and the list of Bugs/Common Issues it points to. You sound like a total moron.

Well... Every time linux

Mr. Pink's picture

Well... Every time linux craps out let's get together and talk about Windows being so bad.

use centos or something else

cornel panceac's picture

ok, i know somebody said it before, but where x distro is failing, try another distro until you find one that works. there are problems in all bleeding edge linux distros like fedora (wich i use a lot). on hp dl 380 centos 5 should work fine but i'm sure there are other distros working. of course, you can use whatever microsoft server os, as long as you enjoy conficker and friends and all the other well known benefits :)
but if you _have_ to use microsoft (like i have, sometimes), go for it.
anyway, returning to linux, developing fast means testing little. so if you want a stable linux, try in this order: slackware, debian, centos. just my 2 cents.
happy hacking!

Grub

FrankR's picture

David,

This sounds like a fairly common issue. When grub is setting up your PC, it enumerates your disks, their startup order, and where it should be booting from (/boot). Sometimes the order that grub detects is different than the Bios's boot order.

What you likely need is to set grub up properly.

Boot any linux CD
Go to the CLI
type "grub"
root {hd0,1}
setup {hd0}

Where "0" is the disk and "1" is the partition you wish to boot from.

Frank

I Agree With the "Windows Installs, Then What?" Post

C. Whitman's picture

This doesn't really show much of anything, except maybe that the leading edge is the bleeding edge and that laptop and server installs in particular tend to be crapshoots.

I have a machine in my office that was put together from components by a colleague and former colleague of mine to run Windows Server 2003. For some reason, Windows Server 2003 will just not run properly on it for any length of time. It develops problems every week or two, or simply crashes altogether. It was taken out of service and replaced with other hardware which does what it was doing with few issues.

I installed FreeBSD on this machine for a little while (because if fully supported the semi-hardware RAID controllers), and it seemed fine (although admittedly it wasn't doing much). I have now had Xubuntu Linux on it for quite some time with no significant issues.

I have seen other hardware that made installing Windows XP, Server 2000, or Server 2003 a big pain in the neck, or just wouldn't run properly with them, and worked fine with various Linux distributions. This doesn't really prove anything except perhaps what I said in the first paragraph. If you want everything to run smoothly, use established hardware with established software and you have a decent shot at it. If you want to experiment with leading edge hardware or software, remember why it's often called the bleeding edge. (Also remember that any combination of OS and RAID controller may lead to new and interesting results.)

That's Funny....

Anonymous's picture

I never have these sort of issues when I install Linux (and usually, Windows) on my servers. Of course, I don't use desktop Linux distros, either. I also make sure that I have the correct drivers available. And I tend to use Dell and Supermicro servers-- I've had too many "issues" with HP servers, the most glaring being their enterprise tech "support" when you do have issues.

Oh, and I also have the correct drivers available when I install Windows servers. And I don't try to install desktop Windows versions on server hardware, either.

Why not try the latest beta version of Windows Server?

Josh's picture

David,

Let's compare Apples to Apples here. Did you try to install the beta of Windows 7? That would be comparable to installing Fedora 10. Yes, it may be simple to install Windows 2008. However, installing Windows 2008 is like installing RHEL 5.3 - they are both commercially supported operating systems that claim to support your hardware. Did you try to install RHEL 5.3?

If you attempt to install something that is a development and testbed environment, not supported, what do you expect?

Also, why did you spend 30 hours trying to resolve these issues - did Fedora 10 provide some compelling feature that you had to have, that a supported distro does not provide and that you could not get on any other machine?

Ok, maybe spending 30 hours in and of itself isn't a bad thing, if you wanted something to do with your time. But I don't understand why you would write an article like this to complain about it. If it was RHEL or SUSE or some other enterprise distro, and the company claimed support, and they weren't helping you, then I could see why you would complain... but using something that is designated for 'testing purposes only' and then complaining that it took you 30 hours to make it work does not make sense.

best,
-JK

FUD?

Linux_Loner's picture

Yes. All that and more.
For starters, you tried one hobbyist OS...wait you briefly mentioned Knoppix.
Next, you did not try other enterprise class OS' (my Ubuntu Server installs/upgrades just fine on all of my DL 385 models). You also did not name the hardware correctly as it is not a HP 385...it is actually a HP DL385 and a older one at that.

Bottom line - articles like this do nothing for Linux. It neither proves TCO, ease of use or any proof point you were attempting to make other than Fedora 10 is hobbyware and not nearly enterprise ready.

Ubuntu Server on the other hand works very well on my HP servers, my IBM servers, and my custom built servers. It is FAR cheaper to operate than any other OS even with world class support from Canonical and I experience NO issues during upgrades.

Now...go back to you lab and do your little experiment as a real system administrator would...with REAL enterprise software and expertise.

Oh and by the way...you are fired.

OK, Windows installs, then what?

Anonymous's picture

First of all, did you file a bug report on Fedora? Since Fedora is the test platform for Red Hat, I would assume you expected a few problems especially with server hardware. Secondly, comparing a bleeding edge Linux distribution with a production ready OS is grossly unfair. But since you did it, let's look at another installation.

For a custom at the customer's request, I just installed a Windows Server 2008 standard edition on a Dell PowerEdge T300 Quad Core Xeon machine, with 4 SATA drives, configured RAID 5 etc, etc.

This was an approved platform for Windows Server 2008. And yes, 2008 installed just fine. Then came the problems . . . .

The network performance was so poor on this machine, with it's Broadcom Gigabit network adapter that it was unusable. Downloaded new drivers, did updates, spent a whole day reading tech articles about network performance issues, found dozens of articles and other suggestions, none of which helped, changed just about everything we could think of, and still SLOW. (I mean 56K Internet slow on a gigabit ethernet). Called Dell and Microsoft Tech support . . . nothing helped.

At the point of almost giving up on the thing, I popped in a spare drive and installed CentOS 5.2, just to make sure the hardware was OK. And guess what. The network speed was about as fast as I've ever seen with Samba. Not the hardware. Buggy drivers?, not according to Dell, Broadcom (although Broadcom blamed the Cisco switch, but the switch was fine with CentOS) or Microsoft. Screwed up installation, maybe, but I though we were talking about, it should just work out of the box. Anyway, finally ordered an Intel network adapter and put it in. And it works great (which pretty much vindicates the installation).

The point is Windows didn't work out of the box and CentOS (Red Hat Enterprise) did.

Still have an open ticket with Dell and Microsoft on this. The Broadcom network problem is still not resolved, but they are aware of the problem . . . now.

What I learned from this is, don't trust that a "tested" platform will actually work.

You should have learned from your experience is don't use a "test" platform on a production box. If it was your intent to be a tester, you did a good job. You found a bug, you worked through the bug to get it working and you have something to report to the developers. File a bug report and move on. Just don't whine about it. If you want production ready Linux, next time just get an official Red Hat Enterprise Linux license. They're cheap enough, they'll help with installation issues and it will just work.

Jim

So much an expert that...

Anonymous's picture

you didnt know there is no such thing as Fedora Core 10. This exposes the lie my friend, you're a shill, if you KNEW what you were doing, you'd know that Fedora dropped the core name a while back.

It's a stupidly small point, rather piddling little point to be honest, but not even being able to get this tiny detail correct, how can anything else in your article be correct? I mean, it's like saying your a maths professor, but then saying that 2+2=3. Getting that wrong, even how stupid and small it is, proves that actually, you havent got a clue what you're talking about.

try harder next time.

GUI error - Gross User Incompetence

Anonymous's picture

It warned you... but you carried on anyway, with a production mindset on a bleeding edge distro release.

There are a huge numbers of drivers, and believe it or not, not every single one works perfectly all the time. Noone complains about too many testers. Of course you prepared the machine, for this test, so the Fedora 10 was an alternative experimental boot, and Fedora 7 could still be operated, didn't you, and you had a backup.

"This is very much about being smart. It is about the smart use of time and the money that time costs."

No, this is about the sys admin, messing around on a whim, not being smart, then whining when he discovers that testing & developing fixes for OSes is time intensive. I've seen other reports of problems with this driver ccis (in openSUSE) so you just didn't do basic homework.

You really should withdraw this Blog article before your Boss reads it!

re: this was experimental from get go

Anonymous's picture

This effort was experimental from get go. The user obviously wasn't installing this on a mission critical production system but was rather using it to test out new features and functions.

If the author is a professional and was setting up a production system, he obviously would have chosen to use a less cutting edge but production proven stable system such as redhat/centos/suse etc.

What the author did in this article was akin to installing a beta, pre production version of windows to see how it looked, felt and what features benefits it added.

Now really.... I mean really!!

AdamS's picture

If you were going to upgrade a production windows environment to a new os in ANY enterprise I've ever been in... you'd have spend 30 hours + PER PERSON over the span of 30 or 40 people, built a test environment, tested every possibility for failure, implemented a plan for the install and generally have spent a TON more money just preparing... Why? Because you KNOW that the MS upgrade would have caused problems, and you wanted to be prepared. You did NONE of that for linux upgrade, and you want to compare it with linux. You missed the Forrest for the trees man!

Distros of choice

SwiftNet's picture

I too have encountered situations where a certain distro would not work while another would, even when they were approximately the same age. It pays to be distro agnostic; ). SmartStart and a supported distro are the way I would've attacked this problem, but hindsight is 20/20.

As for a poster's comment stating that Linux TCO is higher and that quote "Plain and simple, MS enterprise servers can and do, do things that Linux has yet to even come close to (high integrated, extremely redundant and yes, reliable" unquote....please. Linux is more scalable, offers better storage varieties, offers better clustering, the list goes on and on. Linux does not use Exchange, which is the least reliable enterprise mail system I've encountered. So many Windows admins use Exchange as the "Windows enterprise killer app" argument. Posts like this are so scary because there are ignorant people who actualy believe that MS crud.

Oyyy

Oh my. I once did it also

Anonymous's picture

Always, always start HP installations with HP Smart Start media. Not with Windows/Linux media so you get correct mass storage drivers to installation phase. Trust me. Been there, done that many times with Windows 2xxx and RHEL 4.x and almost every time with HP RAID installation failed if I did not use "Smart Start".

an interesting comparison

Anonymous's picture

So why not slap Win2008 server on it and see what happens?

It installs.

David Lane's picture

The first time. No errors.

David Lane, KG4GIY is a member of Linux Journal's Editorial Advisory Panel and the Control Op for Linux Journal's Virtual Ham Shack

Have you read a single comment?

Rohn Wood's picture

Congrats that Window's runs -- strange that this is the only comment you replied to--so you could, do what? Rub ignorance in our face? Take off the Red Hat, resign your membership in advisory board.

What are you doing with...

Anonymous's picture

Fedora on a server system... 10 even which is barely a quarter old...

Why not use CentOS 5? Do you really need to be so much on the cutting edge on the server side?

BTW: I do not like the kernels above 2.6.25 because it does not work well at all from my experience with an M3N78-Pro motherboard... USB.. BUT for server there is no chance in hell that I would use "new" Fedora in production... I also believe that they discourage the use.

Side note: Upgrading php/mysql/apache/postgresql and other components from a major version number is much more than just a simple update... one needs to check the code in productions as well... this part has "nothing" to do with Linux and would be the same for all OS'... like going from MSSQL 7 to MSSQL 2008 and the like... Oracle8 to Oracle11i... most of these kind of upgrades I wouldn't even try to reuse the same machinery. But that's me, mileage may vary.

TCO MA

Curtis Cooley's picture

I realize that it's just a comment on a rant article, but when you say something like "Congrats, you've learned that the linux (sic) TCO (Total Cost of Ownership) is much higher than Windows (and probably Novell?) and it's not something that should surprise you as it's been well known for years" please provide references. Every TCO study I've seen that places Windows above Linux was funded Microsoft.

And let's not forget that Bob Lewis calculated the TCO for a day planner at $5000 a year :) (http://www.infoworld.com/articles/op/xml/03/01/13/030113opsurvival.html)

See, that's a reference ;)

That's TCO for ya

Anonymous's picture

Congrats, you've learned that the linux TCO (Total Cost of Ownership) is much higher than Windows (and probably Novell?) and it's not something that should surprise you as it's been well known for years. Sure, you pay for a license up front, but what does 30+ man hours cost? And support when you need it?

Linux is an amazing development system, a light weight embedded system, and simple server (ie: bind, apache, etc)...but so many people think that because it's different than what they've been using it's the answer to everything wrong with the world. Plain and simple, MS enterprise servers can and do, do things that Linux has yet to even come close to (high integrated, extremely redundant and yes, reliable [provided they're not setup by an idiot -- but the same could be said of any flavor of *nix]). True that it can be said that Linux does things Windows doesn't, but that's simply that they have different purposes. I would choose to build a linux os for an embedded system over windows xp embedded (I have done both), but I have also seen every linux desktop in an organization completely freeze up due to homedirs loading off an NFS server that goes down while the Windows approach (FRS + DFS) relocates the share to another server that has real-time data replication allowing the users to not notice anything happened. It is possible the wonderful world of open source has provided a replacement for the IP based NFS mount technique, but I have yet to see it and this is just one of many examples. In this situation, the windows admin fixes the server that went down and the linux admin has to first update the desktops NFS mounts (and hopefully has some sort of real-time replication to another storage device) before even thinking about the server that went down. TCO.

Linux does have a place in the world and it is a great system to use provided you realize it's strengths and can forgive it's weaknesses, which can equally be said about MS products. In the real world in which I live, I've always found a reason to use both and they've generally worked fairly well together. How's that for averaging the TCO + Benefits?

All in all, great article...and remember that upgrading is never a simple task.

Ironically

David Lane's picture

I tried to do the upgrade and it failed...so I actually slicked the box and started again.

David Lane, KG4GIY is a member of Linux Journal's Editorial Advisory Panel and the Control Op for Linux Journal's Virtual Ham Shack

Webinar
One Click, Universal Protection: Implementing Centralized Security Policies on Linux Systems

As Linux continues to play an ever increasing role in corporate data centers and institutions, ensuring the integrity and protection of these systems must be a priority. With 60% of the world's websites and an increasing share of organization's mission-critical workloads running on Linux, failing to stop malware and other advanced threats on Linux can increasingly impact an organization's reputation and bottom line.

Learn More

Sponsored by Bit9

Webinar
Linux Backup and Recovery Webinar

Most companies incorporate backup procedures for critical data, which can be restored quickly if a loss occurs. However, fewer companies are prepared for catastrophic system failures, in which they lose all data, the entire operating system, applications, settings, patches and more, reducing their system(s) to “bare metal.” After all, before data can be restored to a system, there must be a system to restore it to.

In this one hour webinar, learn how to enhance your existing backup strategies for better disaster recovery preparedness using Storix System Backup Administrator (SBAdmin), a highly flexible bare-metal recovery solution for UNIX and Linux systems.

Learn More

Sponsored by Storix