Luis Cerdas wrote:
<snip>
> The machine is a 1U rackmount chassis, PIII 733 Mhz, 256 MB RAM, 20 GB 7200
> rpm IDE HD (it is one of the firewall/DMZ boxes available from eRacks.com).

Unless you expected me to go to eRacks to buy one to help you for free
(and wait for me to receive it), it is nearly useless to me. (Do you
REALLY think the size of the case or the rotational speed of the drive
is more important than, say, the chip sets or the drive's model number
in resolving a problem?)

> It is running a fresh 3.0 install from an official December CD. The root
> partition is in fact over 8 GB (OS partition, swap partition, overlap [?]
> partition). The entire disk is used for OpenBSD.
>
> The thing is, after the initial install, the generic kernel has not been
> replaced, moved, or otherwise (so in theory, the LargeDrive issues should
> not yet be present).

That is not correct.

If the kernel was moved outside the 8G area, the boot floppy wouldn't
work. Since the boot floppy DOES work, the problem is with the /boot
file.

You have a Large Drive issue, I think. The little info you provided
only reinforces my belief.
 
> Before any flames, though, I do realize we should be separating the root
> partition and possibly others (/usr, /var, /usr/local)

yep.

The flames are for telling us it is urgent...and not providing
info...being told EXACTLY what info we need to help you on this
"urgent" problem...and AGAIN not providing it.

> and recompiling a new
> kernel specific for our boxes;

Nope.
Not necessary unless you are really stressing the thing, running on
unusual hardware, or just want to prove your geekhood or run -stable
or -current.

> however, due to the urgency of this
> deployment, this was not possible before the deployment. True that we have
> learned that it is better to tell our customers to wait a little, but
> nevertheless we would like to know what could have caused this.

Something moved or reinstalled /boot. As you had some kind of
traumatic event on this box, I rather suspect it happened in the
cleanup.

> BTW, the machine hangs right after the Boot 1.28 recognizes "boot: fd0 hd0"
> and the slash freezes; at this point all disk activity stops. It seems as
> though it is searching or waiting for something that doesn't exist.

Probably the tail end of /boot.

> If the
> BIOS settings pointed to SCSI booting, I think this would be the cause,
> however the BIOS is set for CD (IDE) and Primary HD (IDE) boot (currently in
> that order).

Nope.

> The installboot was done once we got it to continue booting from the hard
> disk after we booted from the CD. The command used was based on the one
> available in the FAQ:
>
> cd /usr/mdec; ./installboot -v /boot biosboot wd0
>
> and it finished without errors.

*grumble*
(I asked for the output...not your evaluation of the output..though to
be fair, the problem is fairly clear without it)

> We have four other boxes that are installed
> the same way, but have not developed the problem.

Fix 'em now.

You have five badly implemented systems. You need to rebuild all of
them properly, with / completely within the 8G range. If you want to
run 'em with one partition, odds are you can run your mail and web
server and firewall very nicely with a single 8G partition (that is a
LOT of mail and web). However, I would recommend keeping /var
separate to keep a lid on your mail, logs, and web space growth. Also
keep in mind that you are unlikely to need all 20G, you don't have to
allocate it all from the start, you can allocate 2G for /, 1G for
/var, 1G for /home (maybe you want to keep web sites under /home
instead of /var), and leave the other 16G unallocated. Need a larger
/home? make it! Copy the files, change /etc/fstab, remount, done.

You got bit on one of them already...the others are waiting to bite
you, too.

Nick.

-- 
http://www.holland-consulting.net

• Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]