Subject: Re: Stability of NetBSD/VAX
To: None <port-vax@NetBSD.ORG>
From: Mike Young <young@ecn.purdue.edu>
List: port-vax
Date: 04/11/1997 07:50:47
-> From: Boris Gjenero <bgjenero@undergrad.math.uwaterloo.ca>
-> Subject: Stability of NetBSD/VAX
-> 
-> I'm just wondering how stable is NetBSD/VAX supposed to be.  I've had
-> some unexplained and bad things happen on my two VAXstation 2000s. 
-> First of all, it seems that if junk (eg. a binary file, accidentally) is
-> sent through the console port there is almost always a crash.  The
-> VS2000 sometimes even reboots but then doesn't work right until power
-> cycled.  I've also had panics due to filesystem errors such as corrupt
-> directories.  This was just some filesystem on /mnt.  I've also just had
-> fsck on boot crash the system repetedly until I booted into single user
-> mode and fsck completed without finding errors.  Is this normal or is
-> something wrong here?  Should things be better on a MicroVAX II?

	I have noticed exactly the same thing. My impression is that things
got more dicey with the 1.2C kernel, compared to the 1.1 code, for the VS2000.
Especially the observation about fsck on reboot. I have three data partitions
on an RD54: root on rd0a, /usr on rd0e, and /var on rd0f. On reboot, the
fsck -p will succeed for the root filesystem, mark it as clean, then ALWAYS
take a panic at the exact same place. Guaranteed. I then reset and B/3,
manually fsck the other partitions, and I'm on my merry way. Next time it
crashes and/or I reboot, same sequence. This kind of thing never happened under
1.1 . I'm not entirely satisfied that the kernel code is the problem, though -
there could be some sneaky incompatibility between kernel and some system
utility I suppose. Like maybe fsck? Did it change recently?

	Perhaps this is a divine hint that I should get more familiar with 
db :-|

	I think my next move is going to be putting the 1.1A kernel back on
the system, so it will stay up, then pull down the most recent kernel
sources and think about why this is happening. Hints as to where to start
looking will be appreciated :-) Also, many thanks go to all the thoughtful
persons who continue to post to this list. Taking everyone's comments in
total, it's much easier to get the big picture as to where problems are, and
what steps need to be taken to make things better. Maybe I can make a
helpful comment once in a while too.

_________________________________________________________________________
|   Mike Young     young@purdue.edu     Purdue University ECE Dept.     |
|   W. Lafayette, IN 47907-1285    http://dynamo.ecn.purdue.edu/~young/ |
_________________________________________________________________________