DL360 SMP Install PowerShutsdown right after IDE detect during Boot

I am back again... on a different server...

I installed Debian 2.4.27 on a Proliant DL360 G1 server, 2x800Mhz PIII, 512MB Ram, onboard ATI Rage video, Compaq SCSI Raid Controller, IDE CD-Rom. everything works great. I did an apt-get install kernel-image-2.6.8-3-686-SMP to install the SMP version. When I try to boot the SMP kernel, it consistently gets to the IDE-Detect and then the power drops on the server and a few seconds later the server powers back up. The server then tries to boot again and the cycle continues. I did an apt-get install kernel-image-2.6.8-3-686 (note the non-SMP) and everything works great. I did a search for this issue but did not find any support, any help would be greatly appreciated.

The server is a G1 (1st generation) DL360. It has 512MB of Ram, two 800Mhz PIII/133 cpu's, 512MB of PC133 ECC Registered Ram, the onboard video controller, and the optional raid controller, as well as the floppy and the CD rom drive. It does not have the lights out board, nor any other cards installed.

I used the net-install CD to do the base install.

Any assistance would be greatly appreciated.

TIA!

Richard

0

Comment viewing options

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

Power Supply??? Sheesh...

I posted this question in a few places. The current common answer is that it is a powersupply problem. Apparently a batch of DL360 G1 servers got out with defective power supplies. I am not 100% sure I buy this, but here is what I have confirmed:

If I run with just one processor installed everything works fine.

If I install Windows server 2003 or Windows server 2000, both have the identical behavior during boot.

If I install other variants of Linux, they all work fine until I install an SMP kernel, then they exhibit the same behavior.

I have tried this on two DL360 servers, both have identical behavior (their previous usage did not utilize SMP). I have tried several CPU's, cpu combinations, memory configurations and combinations, always with the same results.

As I can't seem to find an SMP answer, and a new power supply costs far more than the servers are worth (even used ones from eBay, etc), I am looking at either scrapping the servers or relegating them to a duty which can be handled with a single 800Mhz CPU.

Thanks for anyone who did take the time to read the original post, I appreciate it.

Richard

that's a tricky one

So MS Windows was fine because it never really made use of the second processor? That's pretty bizarre. Then again a constant reboot is usually one of:
1. power supply
2. bug in kernel
3. bug in CPU
4. fault on MoBo

Have you seen the latest Centrino-based 'single-board computers'? Up to 1.6GHz at 15W - I wish I had money to buy about 20 of them - I'd build a Beowulf cluster that fit into a pretty small box. I agree with you - sometimes a machine simply isn't worth fixing. In fact if it also consumes huge amounts of power to do very little, buying a shiny new computer may be the more economical option - it all depends on how much money you have on hand and what you want to use the gizmo for.

test

I wonder... if you swap the cpu's over (if you can...) does the fault with the second processor (obviously now in the first slot) fail on normal linux 2.6 or windows?? Also are the CPU's identical pairs??

-----------
Luke Wilson

Makes sense that it would be the power supply...

Consider the fact that when when the kernel sees the second processor and attempts to light it up, you may see a sudden voltage drop at the supply (processors take a lot of current) - causing the first to drop off, reboot, ad nauseum.

A weak power supply could indeed cause such a problem.

Who sez you have to use OEM product? I'm sure something could be found to splice in on the cheap...
--
Jai yen

It is indeed the power supply.

It is indeed the power supply. I took the power supply out of a second DL360 and swapped them, the problem follows the power supply.

The power supply on the Compaq DL360 is definitely proprietary. It has a connector on it that slides directly in to a mating connector on the motherboard. It does not appear to be a standard ATX style connector. Also it is an odd square shape and designed to very specifically fit in to the space. It is cheaper to buy a used DL360 (they can readily be had for under $100 if one watches it) than it is to repair or replace the power supply in this one. For now I have it running single cpu handling MRTG, using Debian 3.6.1 with lighttpd as the web server. We'll see how this works out, it is monitoring 21 servers, a couple of switches and a couple of routers.

Thanks to everyone who read and commented as well. I guess that second PIII-800 must crank some serious current when it gets fired up. I even tried removing 1/2 the memory and one of the SCSI drives, no change. Ah well, it seems to be handling its new task thus far working with a single CPU.

Thanks again all!

Richard

Syndicate content