Upgrading hardware on an 8 year old linux install

C

Cronyx

Limp Gawd
Joined
Mar 24, 2005
Messages
146
A client, a car dealership, has an 8 year old linux server they want to update.

I'm not a linux guru by any means, I've only been using Ubuntu for a few months on a laptop which isn't my primary system.

This old install has no GUI environment, it's all CLI, and I'm not sure of the distro. In fact, I'm not even sure there is a distro at all, it may have been a scratch build for this one purpose.

I'm wondering what kind of problems I might run into with upgrading this? I know windows will fall on its face if you image an existing install to an other box if, for example, the motherboard chipset happens to be different.

I'm guessing the first thing I need to know how to do, is how to install new drivers through the CLI of an unknown distro. I'd probably also need to update the kernel.

Or, if you guys think this would be the path of least resistance, installing a new one from scratch, and just copying the profile over?

Thanks a lot guys.
 
What is this thing doing? DB, web, etc server? If you only need to move a mySQL DB or web server this will be real easy.

uname -a will tell you what kernel its running and many times cat /etc/*release will tell you the distro.

Depending on how it's being used, the best choice will likely be a recent distro on new hardware.
 
Not actually sure what they're using it for, and they don't know either. They use this ancient dos software on the other computers for inventory and what not, and they just know "these programs dont work if the server is off".

I'll have access to it friday afternoon; going to get an acronis image of it to play with, and will load it on some hardware here to figure out what to do next.
 
Nice! I'm sure no backups as well, that would be a fun job when the HD dies!

"Uh, I don't know what it is, or what it does, but the entire company is crashed. Can you fix it?"

At least with an image you could ebay some similar hardware and get them going again.
 
If it's 8 years old and hasn't been updated in that time then it is unlikely to get on with modern hardware.

Assuming it isn't exposed to the internet my line of attack would be to try and get it running in a vm. Vms tend to emulate older hardware so there is more chance of things working than on new real hardware and it gives a level of abstraction from the hardware.

Still from the sound of your post it sounds like you might want to pass this off to someone with more linux experiance. You sound way out of your depth.
 
Build a newer machine to take place of the old one, and clone the drive... leave the old machine in place, and use the new one to stage everything. Trial and error... if something fails, you can drop in the old beast until you figure it out.

Stuff like this is great if you're planning on learning linux. Nothing like trial by fire.
 
Would I be correct in assuming that, if I went the route of doing a clean install of Suse 11 (that's the latest version right? The one on the server is Suse 9 I believe, I'll be looking at it again on monday), if I copied over /var, /etc, and /home to the clean install on the new box, it would act like nothing had changed? Is there more I need to copy than that?
 
Cronyx said:
Would I be correct in assuming that, if I went the route of doing a clean install of Suse 11 (that's the latest version right? The one on the server is Suse 9 I believe, I'll be looking at it again on monday), if I copied over /var, /etc, and /home to the clean install on the new box, it would act like nothing had changed? Is there more I need to copy than that?
Click to expand...

What is the issue with this machine, anyhow? Any particular reason why they want to upgrade?
 
Cronyx said:
Because it's a pentium 3 with 128 megs of ram :p
Click to expand...

If you absolutely have to upgrade...

Go with a faster P3, but stay on the same board. If possible, find the exact same board on eBay or something (possibly here on the forums), and keep it as a spare. Change to a beefier heatsink, and replace any fans. Odds are they're old, and we don't want a failure.

Replace the hard drive(s). Ghost over the contents of the original drives. These may be very old as well, and failure can happen suddenly.

Max out the RAM. PC133/DDR1 is cheap/plentiful.

Dust it out, and put filters on the fans. Pantyhose work too if you wanna be cheap.

Upgrading is in no way going to benefit them right now. As a matter of fact, I'd say old hardware is the best thing going for them if it works... if something breaks, a replacement can be easily had, and for very little money. If you upgrade to modern hardware... well, something failing is no longer a cheap fix.

Keep in mind that business customers are very rarely "wowed" by specs... all they care about is if it works, and how much it's costing them. Keep it running, and minimize their expense... and they'll love you. :)
 
They've already bought new hardware. Kinda have to do what they say because, well, they're the boss, being that they're the client.

They've just out grown this hardware. Company's bigger and there's just way more people accessing it now than there was when it was built. And they only want to upgrade this thing every decade it seems like.

Anyway I'll be at work in a bit to try playing around with it.
 
Alright, made an Acronis copy of the server's drive, put the server back untouched while I play with the copy. I tested the copy before I sent it back, in the original hardware. GRUB wouldn't load for some reason, but viewing the drive with Rstudio while slave mounted in an other computer showed all the data was there, so I figured I could do something akin to a repair install.

Anyway, mounted that here in the new hardware and ran a repair with a SUSE 11 disk, repair finished, but the login/password that I verified worked before, isn't working now. Is there an equivalent to a Trinity boot cd for windows for resetting passwords or something of the sort?
 
Izza said:
If it ain't broke, don't fix it.
Click to expand...
To a certain extent, sure. The problem you run in to with this mentality taken to the extreme is two fold;

1) Hardware WILL fail. You can either choose when it fails ( planned upgrades ) or you can leave it up to chance. Murphy's law is alive and well, the later is likely to happen right when it shouldn't.

2) There is a knowledge drain, and the longer you go between upgrades the harder it becomes to upgrade/replace.

In this specific instance, before you do anything else I would observe normal behavior. Watch the number of connections, what the processor usage. Who connects when, catalog what starts up on boot. All of it.

And I would avoid SUSE like the plague. Personal preference would be CentOS, but almost anything is better than SUSE. Especially now that it's future is uncertain.

EDIT: Oh, you are trying to mirror the install. Well, you can boot up off of a live disk and run grub to reinstall the boot record ( grub-install ). That *might* work, but then again it *might* not ( probably won't ).
 
I'd love to ditch SUSE. I just don't know how.

They don't even know what they use this server for, they just know they need it. I'm not even sure what purpose it's serving, myself. It's been there for 8 years, and it has something to do with inventory or payroll or something. They use some old dos-like programs on their other machines, which don't work if the server isn't turned on.

That's the primary reason for an in-place update, I don't know what software they're running, how it's configured, or how to put it back if I did need go from scratch. I know that this box was put together before FSSTND was as ubiquitous and established as it is now, so I'm not sure where everything is set up to be located on this install.
 
Find out what software they are using and find a moder replacement.

Maybe Oracle or SAP perhaps?
 
I think I've been playing around enough with this to explain a problem I'm having now.

I forgot how linux handled hardware, everything being a file or directory under /dev.

When I cloned the disk contents and restored to an other disk, that didn't account for the fact that the old piece of hardware was mounted as hda1 (I think I'm explaining that right). This new piece of hardware has a different hardware ID, so it's trying to mount it as sda1 or something, and GRUB shits itself not knowing what to do.

Running through the repair install options on a SUSE 11 disk encounter the same problems, because it's a different drive, and from its point of view, an other piece of hardware already called dibbs on that alias so it isn't allowing this new hard drive to take over.

Am I right about what's going on there, at least in superficially, and if so is there a work around?
 
Met-AL said:
Find out what software they are using and find a moder replacement.

Maybe Oracle or SAP perhaps?
Click to expand...
I would try to find out what they have installed, I'm just not familiar enough to do that. I don't know what the linux analog for "add/remove programs" is. As far as director structure, isn't /var the analog to c:\program files? If not, what is? I might be able to figure out what they have installed by just looking at the directory structure if I could figure out where things get installed.
 
Oracle or SAP would be a complete replacement for what ever software they are running. You need to find out what they are using it for.. does it run the entire company? Payroll, inventory, production, accounts recievable, accounts payable, etc.

Or does it do one thing and they use other software for the other tasks? Can they all be combined into one software package? If so.. time to find a consultant and get some quotes. Maybe even something simple like Quickbooks would do all of it.
 
Problem is they're pretty stubborn about what they want. They aren't going to change anything without basically getting down right aggressive enough with them that they'll just go ask someone else to do it. Their mentality is, this hasn't broken once in eight years, we are not changing anything, end of discussion. There's no negotiation to be had. They want modern hardware -- the hardware they already bought -- and that's it. No switching to a different distro of Linux, no changing their software. I appreciate there might be better options in theory, but they are non-options for this scenario with the mentality of this client.
 
There is a 99.999999999% chance a 10 year old kernel (you said the box was 8 and i'm sure they didn't load a fresh kernel that day) will not boot that new hardware.

You will either be upgrading various parts of the software or buy eBay hardware.
 
longblock454 said:
We still need the output of uname -a to see how old the kernel is.
Click to expand...
Damnit. I did do that, friday, when I had physical access to the server (it was here at the shop, when I made the acronis clone), and I wrote it down, but I forgot to post it, and now the paper's gone. The image I have here won't boot on anything either.

I had it friday because they closed early and didn't mind leaving it here. The rest of the time, they can't be without it.
 
Huh... file is strangely missing. All that's in that directory is:


device.map
e2fs_stage1_5
fat_stage1_5
ffs_stage1_5
jfs_stage1_5
menu.lst
minix_stage1_5
nbgrub
pxegrub
reiserfs_stage1_5
stage1
stage2
stage2.netboot
vstafs_stage1_5
xfs_stage1_5


Now that MIGHT be because I've already gone through some failed repair install attempts from the SUSE 11 disk. If you want, I'll restore fresh from the acronis image again and see if it's there when unchanged.
 
# Modified by YaST2. Last modification on Tue Jun 22 12:32:01 2004


color white/blue black/light-gray
default 0
timeout 8
gfxmenu (hd0,1)/boot/message

###Don't change this comment - YaST2 identifier: Original name: linux###
title Linux
kernel (hd0,1)/boot/vmlinuz root=/dev/hda2 vga=0x31a splash=silent desktop showopts
initrd (hd0,1)/boot/initrd

###Don't change this comment - YaST2 identifier: Original name: floppy###
title Floppy
root (fd0)
chainloader +1

###Don't change this comment - YaST2 identifier: Original name: failsafe###
title Failsafe
kernel (hd0,1)/boot/vmlinuz root=/dev/hda2 showopts ide=nodma apm=off acpi=off vga=normal nosmp noapic maxcpus=0 3
initrd (hd0,1)/boot/initrd

###Don't change this comment - YaST2 identifier: Original name: memtest86###
title Memory Test
kernel (hd0,1)/boot/memtest.bin
 
going to be leaving for the day in a minute, work on it more tomorrow. Thanks to everyone that continues to help! On the bright side, I'm getting some very useful real world practical linux experience.
 
Cronyx said:
going to be leaving for the day in a minute, work on it more tomorrow. Thanks to everyone that continues to help! On the bright side, I'm getting some very useful real world practical linux experience.
Click to expand...

Sorry for the delay! Not sure what other distros use Yast but SUSE defiantly does/did. Fedora adds the kernel version to the end of the vmlinuz name, was hoping that would be an easy way to pull the kernel version.

Looks like for SUSE:

/usr/src/linux-{kernel-version}.SuSE
/boot/vmlinuz.config
/boot/vmlinuz.version.h

In one of those you should see something like 2.6.12 or some similar numbering.

Then lets see whats in this file if it exists:

/etc/SuSE-release
 
Met-AL said:
Find out what software they are using and find a moder replacement.

Maybe Oracle or SAP perhaps?
Click to expand...
Meh, I'm not sure I'd recommend anyone to Oracle, especially in an environment like this where you won't have a tech onhand.

Here's what I'd do; contact the software vendor, tell them what you need to do. You will likely be facing a wipe/reload scenario, you will need the app vendor one way or another. If you can't get in touch with the software vendor, then I'd begin looking at a migration scenario ( as they don't want to be running mission critical software that is unsupported ).

Either way, this isn't simple. I would make it absolutely clear that this isn't as simple as a hardware upgrade, and ask them if they want to proceed. I would also caution them that regardless of their answer, this needs to be handled soon or they risk their data and business processes.
 
UPDATE: _These fraking people._

They're running it as a NAS. THERE ARE NO PROGRAMS INSTALLED.
 
Their clients are all windows, so NFS. And yeah, going to do a clean install of an other distro on the new hardware, copy the data over, set the network name the same, share the direcotries, etc.

They actually are running software from this thing, but it's not being run server side. They're .exe's that they have shortcuts to, and they save the output of these files to the server /also/.
 
marley1 said:
just raw files? replace with windows =)
Click to expand...

They got more than 10 people connecting (XP limits it to 10 simultaneous inbound connections), so that means you have to get a more beefy (and expensive) Windows licence. They don't want to pay for software, so it's gotta be linux.
 
Poking around with a Cent 5.5 live cd. Feels uncomfortable as I'm not familiar with it, I use ubuntu on my laptop.

Any reason why I can't just use ubuntu for this? Why cent?
 
Cronyx said:
Poking around with a Cent 5.5 live cd. Feels uncomfortable as I'm not familiar with it, I use ubuntu on my laptop.

Any reason why I can't just use ubuntu for this? Why cent?
Click to expand...

Cent 5.5 is pretty old looking, RHEL 6 is already out but Cent hasn't caught up yet, version 6 is MUCH newer. With RHEL/Cent you have at least 7 years of security patches. I don't know Ubuntu but you probably only looking 1 to 2 years at most.

RHEL/Cent is hardened, Ubuntu like most of the other fast development distros are not as secure. Saying that I do run Fedora in several production environments with no issue, however I am very familiar with Linux and know when I can get away with it!

I would have no issue running Fedora as a samba server for anything of my own, but would not do it for a customer.
 
You must log in or register to reply here.
Back
Top