Ext3 users - Nov 2001 - 7.2 might have ate my partition table.

Well I thought someone might be interested in this. I have a Compaq server
with a cpq array in a raid 5 configuration with some bog standard video card
no sound card, NO X installed, no applications/modules not included
with the distro, and 256 megs of memory. I upgraded it (yes I know 
not supported but...) from Roswell to 7.2 in the middle of 
last week and installed all applicable errata updates including the 
2.4.9-13 kernel. All seemed to be running well. Monday morning at 
4:02 am it stopped communicating with my network monitor. 
We were able to ping it but when trying to login it would hang 
after the username was entered (no passwd prompt). I had a guy push the 
"big red button" but it did not come up (the machine is 300 miles
away).
What he got was the grub prompt. We rebooted to rescue mode from anaconda 
and it was unable to detect any linux partitions. fdisk shows what 
appear to be correct partition sizes and types but ALL partitions start 
and stop on 1. Fdisk also complains about partitions overlapping and not
starting on correct boundries.

The good news is this is not a production machine. We were just getting
ready to start testing it in preparation for production. As a result it
is of no hardship if we have to reinstall. Can someone suggest a good 
way to try to figure out what happened? Is there any useful info 
to be gained for either the developers or myself. I would like to know 
what happened.

I cc the ext3 list on this because the partitions were converted
from ext2 to ext3 during the Roswell upgrade from 7.1. 

Ideas, suggestions and comments are welcome,

-- 
......Tom		Dysfunction The Only Consistent Feature of All
tdiehl@rogueind.com	of Your Dissatisfying Relationships is You.

On Tue, 2001-11-13 at 17:54, Tom Diehl wrote:
> I cc the ext3 list on this because the partitions were converted
> from ext2 to ext3 during the Roswell upgrade from 7.1. 
Its incredibly unlikely (in fact I'd say its impossible) for it to be an
ext3 issue.   Your most likely candidates are the cpqarray hardware,
which has some bizzarre behaviours occaisionally, or possibly the kernel
has scribbled on the partition table - VM fault of some sort maybe.

Do you have the compaq monitoring stuff on that machine?  If so thats
almost certainly your problem - compaq supply those as binary kernel
modules which they hack on install to match the kernel versioning.  I
suggest anyone considering using them not do so and tell the compaq
sales droids to provide source.  [I have evidence that they do scribble
on random memory - most likely due to a kernel or toolchain version
mismatch].

For the record, I have had compaq servers with cpqarrays running ext3
for well over a year now.

	Nigel.