Xref: utzoo comp.sys.dec:2811 comp.unix.ultrix:2981
Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!umich!caen.engin.umich.edu!mingin.engin.umich.edu!news
From: paul@speedmetal.engin.umich.edu (Paul Killey)
Newsgroups: comp.sys.dec,comp.unix.ultrix
Subject: Re: DECstation Paging (Swap ?)/Hang problems
Message-ID: <1990Mar9.153926.26579@caen.engin.umich.edu>
Date: 9 Mar 90 15:39 GMT
References: <1032@crk56.bnr.ca>
Reply-To: paul@speedmetal.engin.umich.edu (Paul Killey)
Organization: University of Michigan College of Engineering
Lines: 30

On the general topic of system hangs on 3100s ...

I have a system with 2 maxtor 780M drives, an internal maxtor 200M
drive, and a TK-50.

My system does hang occasionally with heavy (or the right kind of) disk i/o.  
I suspect paging i/o does this as well.  This isn't the problem where the
load average creeps up for a while, and then the system hangs.

In my case, I suspect a cabling problem and am waiting for some new cables.
At least, I won't look into this more until I get new cables.

I think the interaction between the dec bus and some peripherals can be
a bit touchy, and there is not a lot of room for error.  Minimal reflection,
whatever, on the chain of scsi devices, can be a problem.

For example, I can hang my machine by doing a dump at any level but level
0.  Level 0 works OK.  The others fail during the scanning, as the raw
file system is seeked and read.  But if I change dump to write anything
to the screen (either in multi-user or single-user mode) like a . or a
carriage return just often enough to slow things down a bit, the level
9 dumps will work.

It's always something.  But we do have machines that are configured with just
an internal maxtor 200M drive hang once in a while too and I suspect either
paging i/o or some other glitch in the VM code.  99% of the time, things work
fine.

--paul
"Don't call me baby when she's waiting in the car."