[wplug] Little Bo Peep Lost Her MBR...

Hagbard Celine hceline at fnordnet.net
Mon Jun 13 21:37:43 EDT 2005


I've been having a problem with lost Master Boot Records for some time.
Lately, some work I've been doing with large files has given me an opportunity
to collect some data in connection with the phenomenon.  Hopefully, some of
you folks might know what's going on?

Some system particulars:
Motherboards: ABIT K(x)-7 type motherboards, VIA chipsets all
Hard drives: Many models, most are Maxtors
Kernel: linux-2.4.19 through linux-2.4.29
C library: glibc-2.3.2
Filesystem: e2fsprogs-1.35
Boot Loader: lilo-22.5.x through lilo-22.6.1

Layout:
/dev/hda1: WinCrap98
/dev/hda2: An early linux-2.4.x system, once used for experimentation
/dev/hda3: The template system for my distribution, has had the versions above
/dev/hda4: swap
The boot loader on /dev/hda will boot into the systems on /dev/hda2 and
/dev/hda3, and chain to WinCrap or to the boot loader on /dev/hdb

/dev/hdb1: The root filesystem for my daily-use system; see above for versions
/dev/hdb2: /local for the daily-use system (/usr/local symlinks to this)
/dev/hdb3: /usr for the daily-use system
/dev/hdb4: swap
The boot loader on /dev/hdb will boot into either of two kernels, one with
SCSI emulation, the other without, otherwise identical.

The Problem:
It appears that when /dev/hdb1 gets to approximately 90% capacity, /dev/hdb's
MBR goes bye-bye every time I shut down (the command used seems not
significant; halt, reboot, shutdown all produce the effect).  The MBR on
/dev/hda always sticks around.  Strangely enough, though, when I boot into one
of the Linux systems on /dev/hda, mount up all the required partitions, and
write a new MBR to /dev/hdb, it will hold for the next reboot.  It just
disappears again when next I shut down.

Has anyone seen anything like this, and might anyone have an idea of the cause?

Many much thanks,
Hagbard



More information about the wplug mailing list