Subject: How to cause a melt down on a Multia
To: None <port-alpha@NetBSD.ORG>
From: Brad Spencer <brad@anduin.eldar.org>
List: port-alpha
Date: 05/15/1997 20:22:40
This appears to be a sure fire method of nailing a Multia.  It is
perhaps not all that suprising, however.

What you will need:

1) A Multia with 24MB of ram, 1.3GB disk, 150MB swap

2) NetBSD-alpha as of 4/19 or so

3) Majordomo 1.94 installed on the Multia

4) An over zealous brother who shot around a dozen separate messages
to a 'digest' type LIST on the Multia all AT THE SAME TIME


What you will see:

As the messages come into the Multia, and perl [btw, an older ECOFF
binary on a ELF system] is forked over and over to process the digest
and then sendmail likewise, the use of memory will rise greatly,
consuming around 70MB or so of swap.  Thrash.  Thrash.  Thrash.

In addition, the load of the machine will rise to between 35 and 45.

[rwhod kept running and reporting to the other machines on the ether]

The Multia ran with this load for around 2.5 hours before throwing
these messages many many many many times out to the console:

rmfree: map 'swap' loses space (32)
rmfree: map 'swap' loses space (64)


In a very short time the Multia panics with:

panic: rmfree swap


Thunk...




Brad Spencer - brad@anduin.eldar.org   http://anduin.eldar.org