Machine locks up - I suspect HD
 

I have a "server" at a job I just started. I use the server term VERY loosely. It's basically an old desktop sitting sideways in a rack (which Linux will run on and I'm a Linux fan/user by the way).

Here's the problem. We have a couple of web apps that are absolutely crucial to the business that run on this server and at first I tried to make a tarball of the entire directory to scp it to another machine. My brilliant predecessor already had everything he needed installed (it's Fedora release 14), but he didn't know that so he downloaded one of these "lamp" packages and installed it where everything is contained under one directory - apache, mysql, all the db's, etc.

Anyway, on my first attempt I tarred the entire lampp directory and gzipped it and attemped to scp it to another machine. Halfway through the copy, the machine locked up - as in no response from the keyboard at the machine at all. NOTHING.

My first thought was that the file was so big (3.2 GB) I wasn't sure if it was a 32 bit file size limitation issue - I don't know that much about LFS, I'll admit.

Next attempt, I tried plugging a thumb drive into a USB port on the machine and just running an rsync. It got along a short little while and then the same thing happened. Machine locked up. No response from keyboard (and obviously not from any of my shells I had to it with ssh).

This morning I though I'd try to tar each vhost at a time and scp them over to the new server. I got through a couple of them, then on about the fourth one, same thing. I am completely stumped.

I looked at dmesg to see if I could see anything that stood out and nothing does. I also looked and /var/log/messages and saw nothing really suspicious there either.

This is a VERY OLD computer (hence the reason for trying to get these sites off of here).

Does anyone have any ideas that I could try? I'm at a loss here.

TIA,

Adam

One thing that can give you these types of failures is a failing power supply. As power supplies get older the amount of power they produce begins to go lower. Your power supply may be slowly dying.

How you diagnose power problems varies with the hardware you are using. I used to have a Dell with a failing power supply. The Dell had small lights on the front of the case. The light pattern gave me a number which indicated which type of pre boot hardware failure I was having. I replaced a failing 75W power supply with a new 100W power supply and that solved the problem. Other computers that I have owned had hardware error codes in a special area of the BIOS.

I suggest that you find a hardware manual for your equipment and see if you can diagnose a hardware problem.

------------------
Steve Stites

I'd suspect heat, power supply, ram, and any other hardware. We can't say without some tests.
Take the cover off and see how much dust is in it. Running systems sideways has caused issues before so lay it flat. Be sure to watch out for esd and maybe clean and reseat stuff.



You might consider getting a usb to sata adapter and dd or clone the disk before you loose it all. You might be able to re-create this machine in a VM to keep it going. In any case, get a copy by some means.

Power supply is a real possibility.
I suggest No compression, if it locks up kill the cp process (get the id while it's working). It may recover.

Run memtest86+ first.

If you suspect the HDD, then you should try to get your data off ASAP. After you do, you can run a SMART long test.

Not trying to stop progress but fedora 14 has been outdated for 3yrs. Are these web apps going to run on todays fedora?
Are the apps available thru repos?

Thanks for all the replies. Good info there. I'll look into the power supply and the first thing I'll do is stand it up straight. The problem with this thing is that they're running a mission critical web app on it and I've been trying to go through and clean the directory up a bit (it's 5 GB). Someone who didn't know what they were doing downloaded one of these pre-built "lamp" packages, so EVERYTHING is contained in there. The DB's, apache, the libs, the code - everything...

The thing is that it will run along fine until I start trying to copy things out of that directory. Then, at random, it just locks hard. No response from keyboard or anything.

To EDDY1,

These are apps that were written in house, so I hope they aren't using too many old libs and will still run. We'll see. The It's business critical.

I'm thinking it is a bad RAM stick until it is ruled out using memtest86+.

I'm going to work with it this weekend (since it's a "production" box). I'll post the results...

Copy locking up could also be swap too small. Top command and swap tools may help.

Running fsck on it may help but I'd do that after I cloned it off. If this is critical then you should have had a copy or full backup already.....