"Operating system not found" [Solved. Sort of.]

[Sivar]

I opened my server and a SCSI cable bumped what must be an internally loose molex connector to my server's boot drive as I made a fan cable a bit more tidy. The boot drive lost and regained power several times in a brief period of time.
I shut the server down and took a photo. It hasn't booted properly since.

I started off of an install disk, and found that the partition is in perfect health, there are no bad sectors on the entire drive, and a long SMART test with smartmontools says the drive is in perfect health. I would expect this, since the drive is rated for more than 50,000 power cycles.

For some reason, the bootable flag of the partition was set to "off", so I used fdisk to re-enable it, and I reinstalled grub to the partition just to be safe. Grub should overwrite the MBR. It installed fine--not a hiccup.

The controller is set to include the hard drives in the BIOS scan, system BIOS is set to boot the appropriate drive first, and the partition is marked bootable and has had grub installed several times, yet the system still doesn't seem to be particularly enthusiastic about booting up from the hard drive.

System hardware:
Tyan S2466 motherboard (760MPX chipset)
Adaptec 29160 controller
Boot drive is a Seagate Cheetah X15-36LP
Power supply is a PC Power & Cooling Silencer 400
Again, the drive is in perfect health; no bad sectors, no SMART errors or warnings, no file system damage, and GRUB runs and installs just fine. I can mount all partitions on that hard drive with no problems.

At first I thought that perhaps fdisk and grub were not fully overwriting the MBR, so I manually zeroed it using "dd if=/dev/zero of=/dev/sda bs=512 count=1", and then manually rebuilt it with fdisk (after noting the start and end locations of each partition).
I've also tried installing LILO instead of grub. It still refuses to boot.

I am about out of ideas here. Any help would be greatly appreciated.
_________________
The greatest deeds are still undone, the greatest songs are still unsung...

[buffalo]

are you installing grub in the MBR of the disk or that of the partition itself?
I don't remember well, but it seems to me that i had a similar problem one time and that i was in fact installing grub on the partition MBR (which was bootable), but my bios only wanted to boot from the disk MBR.
_________________
--
Forgive your ennemies -
nothing annoys them more
O.wilde

[Sivar]

Thankyou for the reply.

Grub is going to the main disk MBR.
_________________
The greatest deeds are still undone, the greatest songs are still unsung...

[Sivar]

I don't believe it.

It was the power supply. The power supply was causing the system to refuse to boot from the hard drive.

What the hell.

I never even considered the power supply. For one thing, it is a PC Power & Cooling unit, which is generally regarded as a brand of power supply that is rather unlikely to be flaky. Additionally, the power supply was powering all hard drives at the same time (and there are quite a few of them) without problems, yet even when I stripped the system bare as WinCC suggested (e.g. only the boot drive, no NIC, ancient low-power PCI video card, no CD-ROM or tape, no non-critical cooling fans), it still wouldn't boot. Period. But the drive worked wonderfully.

In the process of stripping the system bare, I replaced the power supply with an old Seventeam ST400GL, and it worked fine. It boots.

I don't like unsolved mysteries, though. Are there any ideas as to how on earth a power supply that can power 7+ drives somehow prevents a system with 1 drive from booting, yet the drive itself works fine, mounts fine, self-tests fine, and is detected without problem?

I would have figured that this was just a coincidence, but I tested the system every time I removed a component, and retested it with the PC Power & Cooling power supply (using different cables and molex connectors). The only factor that differs between "works" and "doesn't work" is the power supply.
_________________
The greatest deeds are still undone, the greatest songs are still unsung...

[buffalo]

i don't understand at all...
why would a power supply cause this problem?
it's a mystery to me...
_________________
--
Forgive your ennemies -
nothing annoys them more
O.wilde

[justanothergentoofanatic]

Unequivocal proof of demonic posession? Have you tried exorcism?

-Mike

[Sugarat]

I've had similar goings on. Its bizarre, but now I always check the PSU's fairly early on.
_________________
--[ UltraSPARC ]--
Accept no imitations