On Mar 10, 2007, at 23:40, Ken Raeburn wrote:
AC3 came out just in time for me to be reinstalling a couple of the
test machines at work.
[... AlphaStation 400 stuff omitted ...]
On another machine (PWS500au, only 128M RAM) it started up okay,
but when it got to "Installation Starting" ("Starting install
process. This may take several minutes..."), it crunched for a
while, displayed "install exited abnormally", and went through the
same shutdown sequence. No messages were displayed about why it
exited, and root/install.log is empty.
Any suggestions?
I've finally gotten around to dealing with this machine again.
Sergey Tikhonov suggested checking other virtual consoles. Jay
Estabrook suggested I try to get my hands on more memory.
I did get my hands on some more memory from another machine. (I
figure, worst case, I install with gobs of memory and then put the
other memory back, and see if I can run an already-installed system
with only the 128M.)
First thing I tried was the Debian installer, since I didn't recall
if I'd tried that on the PWS before; my email had only mentioned
trying it on the AlphaStation. With the extra memory, at least, that
seemed to run fine, at least far enough to finish the first pass of
installation, eject the CD, and reboot. So I went to retry the AC3
install, and remembered I should check for more info logged anywhere
with just the 128M, so I pulled out the extra memory and booted the
AC3 CD.
This time, it seems to be working somewhat better.
It has, a couple of times, seemed to freeze up partway through the
installation of packages. The installer display stops changing, and
the disk seek noise has stopped, for 10 minutes or more. But if I
switch VCs and run "top" or something at the shell prompt, that seems
to prod it into activity; the disk noises, at least, start up again
right away. I can quit top and go back to the installer virtual
console, and it's moving on to other packages.
It got as far as starting on the second CD before giving me "install
exited abnormally".
So I checked the different VCs (couldn't run any commands because the
OS has shut down), and found some stuff logged by the kernel:
<6>SELinux: initialized (dev hda, type iso9660) uses genfs_contexts
<6>sd 0:0:1:0: SCSI error: return code = 0x8000002
<6>sda: Current: sense key: Medium Error
<6> Additional sense: Unrecovered read error - recommend rewrite
the data
<6>Info fld=0x10bb9f0
<4>end_request: I/O error, dev sda, sector 17545712
So, it appears I've got some kind of disk problem on the disk I'm
installing to, and the install system hides it from me.
I think at this point it was still running from the file system
pulled from the CD, not running code from the hard drive. If that's
true, it would've just been an error reading or writing data in the
install process that isn't handled properly, rather than an error
loading pages of the executable image, which might cause the process
to simply be killed...
Ken