[PLUG] Re: Tape Drives and Power Supplies

Keith Lofstrom keithl at kl-ic.com
Sun Jan 9 21:27:01 UTC 2005


Russell Senior writes:
> Nope.  This is a bare box, with only one HD, a floppy and a SCSI card
> in a PCI slot.  LAN, Audio and Video are all on the motherboard, but
> of those only the LAN has anything connected.  Otherwise, only a
> serial console and the SCSI cable to the external tape drive.
> 
> I tried to induce a load with heavy disk seeking and it didn't seem to
> bother it in the least.  Also, I just removed two IDE drives and
> replaced them with one IDE drive.  You'd think the electrical load
> would have gone down from that.
> 
> The tape-related failures started immediately after the disk change.


Since this is after a drive change, and since you are presumably using
a +80GB drive to replace a couple of 10's or 20's, you may be taxing
something in the disk formatting, or in the BIOS;  backups may be
thrashing areas of disk space or memory that were not used before,
and are only accessed during backups.

You mentioned originally that you were using an Intel D815.  A bit of
Googling for "D815" and "shutdown" found a vague reference to the need
for a BIOS upgrade.  Look into that.  Perhaps older, smaller disks did
not stretch the BIOS quite as much, or tickle the memory space that
throws the power-off switch.

Of course, do an fsck -c to find/heal bad blocks, and check your
partitions.  Given your earlier comments about grub and disk setup,
there may a mismatch between the partition tables and the filesystems
stored on them.  

You did remember to do a "mkswap" on your swap partition, yes?  Is the
swap partition the same size as before?  What happens to available
memory as the backup program runs?  

Also try a dd if=/dev/hda of=dev/null and see if the drive gets 
unhappy as you read its entire length.

You mentioned that you reseated the connectors.  I would do that with
ALL the connections, including the RAM and CPU and mobo power connectors
and PCI slots.  Something might have pulled /almost/ loose during the
disk replacement adventure.  Or the flexing the mobo made an internal
wire intermittent. You may kill the mobo with the extra stretching and
squeezing, but better to spend $100 replacing a mobo than losing client
data.

Last, if nothing else worked, I would put the old disks back in for
a weekend, run the backups again, and see if the problem goes away.
I would also try the new hard disk on another system, and repeat some
of the above tests.

Fascinating problem you have.  I look forward to your exultant bragging
about a clever solution, preferably soon.

Keith

-- 
Keith Lofstrom          keithl at keithl.com         Voice (503)-520-1993
KLIC --- Keith Lofstrom Integrated Circuits --- "Your Ideas in Silicon"
Design Contracting in Bipolar and CMOS - Analog, Digital, and Scan ICs



More information about the PLUG mailing list