Kafka is here....

[David Lesher <wb8foz nrk com>]

A friend and I are at wits end on this case. We had started
out on a NFS installation and it was going well, but we ran out of
time. So we later started over with a CD:

I pick the story up there.... suggestions welcome...




HW Environment:

* Nice Super EISA Rev 1.1 mainboard (SIS chipset) with AMI SIS EISA BIOS
(7 EISA slots and 1 VLB slot, 16 SIMM slots [4 banks of 4] with 32 MB
installed (8 ea 4MB 60 nS SIMM) AMD 486DX2/80 processor (CPU clock 40 MHz)
* Racal 6510 NIC
* Adaptec 1542B SCSI controller (BIOS & floppy jumpered out) attached to
Sony CD
* Promise EIDE4030Plus VLB caching controller, (cache disabled!) with a
Maxtor LXT-340A as master and an IBM DCAA-34330 (4.3 GB) as slave.  LBA
enabled at the controller for the IBM and disabled for the Maxtor.
* ATI Graphics Ultra Pro Mach32 video card w/2 MB VRAM
* 101 key keyboard & 3 button Logitech serial trackball

IBM DCAA-34330 partioned as follows using Linux fdisk: (cyl values = LBA
translation)
	Part		Cyl	Type 		Size	Use
	hdb1		1-33	Pri		~256 MB	ext2, /
	hdb2		34-46	Pri		~96 MB	swap
	hdb3		47-527	Ext		~4.0 GB	carrier for logical parts
	hdb5		47-142	Log		~750 MB	ext2, /usr
	hdb6		143-238	Log		~750 MB	ext2, /usr/src
	hdb7		239-383	Log		~1.2 GB	ext2, /home1
	hdb8		384-527	Log		~1.2 GB	ext2, unassigned

All partitions had file systems built with bad block checking (painfully
time consuming)

The Maxtor has DOS and Win3 installed on it and is a single partition

Tale of Woe:

Tried a number of different installations, both from the boot
floppy supplied with the 4.1 Official, floppies made from
the 4.1 CD, a CD based install from DOS using the #1 4.1 CD,
a CD based install from DOS using the 4.2 CD #1.  All failed in
essentially the same way!  They ran out of space and the package
installation would fail either because the expansion failed or
because "...device out of space".  Yet a check via VC#2 showed the
partitions mounted as /mnt/, /mnt/usr & /mnt/usr/src and a df on
each showed useage to be under 10% on EACH!  After considerable
time, frustration and bad language (sort of randomly mixed ;-})
formed the hypothsis that the problem was with the logical
partitions or the LBA translation in the "install2" package.

To test this, I tried another install making hdb5 the root (in
part to preserve the installation log and system snapshots we hope
are still on hdb1 from earlier install) working from a fresh set
of floppies made from the RH 4.1 CD using rawrite from DOS.

VC#1 had install running and showed succession of failed package
installations packages would appear to decomnpress (upper bar moved
left to right), but would then fail and require an acknowledgement

VC#2 ran mount and got the following
	/dev/root  /  		ext2  	rw 	00
	/proc      /proc	proc	rw	00
	/tmp/ram2  /tmp/image	ext2	ro   	00
	/tmp/scd0  /tmp/rhimage	iso9660	ro 	00
	/tmp/hdb5  /mnt/	ext2	rw	00
	
running df showed less than 4% used on /mnt/

VC#3 had install history (input to log file??), which showed in part:

...
installing initial packages (3 of them)
	installing setup
	installing pamconfig
	installing filesystem   <<<< suspect package ??
Installing packages
*followed by long listing the "error installing package xxxxxxx"

VC#4 long string of the following error messages:
"6 attempt to access beyond end of device
 03:45: rw=0, want 1077676661, limit=771088"
It appeared as if this was the same message for every failed install (at
least one screenfull of identical messages

VC#5  Had the following output from the e2mkfs that ran on hdb5 as part of
the install:
"193040 inodes, 771088 blocks
38554 blocks (5.00%) reserved for super user
First data block=1
Block size=1024 (log=0)
Fragment size=1024 (log=0)
95 block groups
8192 blocks per group, 8192 fragments per group
2032 inodes per group
Super block backups stored on:
		{many blocks listed apparently on cylinder boundries)
Writing inode tables: done
Writing superblocks and filesystem accounting information: done"

killed the install after ~30 minutes, when I got tired of acknowledging
each failure

Also tried to bring up a live CD file system per the instructions
in the official installation manual.  No "liveboot.img" to be
found, no readme in the live directory, no clear (or even murky)
indication of how to do it - was unable to bring up the CD ROM
live "test drive"  which I planned to use to recover the earlier
installation log and system snapshot stored on hdb1, so they could
be forwarded for review.  Installation instructions need to be
fixed!  No live file system on the 4.2 CD, so did not try with it.


Bottom line:  Murphy 10, Linux installers 0
		  + one shot weekend

Since now strongly suspect a problem with the use of logical
drives on a large EIDE disk (>= 2GB), may try to re-partition
the drive to create 3 primary partitions (/, swap and /usr)
plus one extended partition that will hold the balance of the
disk capacity.   Otherwise, am stymied!

I have installed earlier versions of RedHat on a Dell P133 with 2
WD 1.6 GB drives and had no problems - in the past also installed
Redhat (Mothers Day through 3.03), Slackware and Denebian without
any particular problem.  Installations have been on both Drive
0 and Drive 1 with no apparent difference

Would be grateful for any ideas or suggestions, the problem seems
to be in  "install2", since that was common to both the 4.1 and
4.2 attempted installs.  I was not able to come up with a work
around since there does not seem to be a way to get past either
the mount error or filesystem write errors from within the in
mmemory image used for installation


-- 
A host is a host from coast to coast                 wb8foz nrk com
& no one will talk to a host that's close........[v].(301) 56-LINUX
Unless the host (that isn't close).........................pob 1433
is busy, hung or dead....................................20915-1433