>	David Brownlee <abs@anim.dreamworks.com> wrote
>	|It uses as much swap space as it needs from the available swap
>	|devices. If you run out of all space... bad things happen.
>
>The machine remained (reasonably) responsive right up to
>the point where all swap was exhausted, whereapon the sis task self
>terminated. No other tasks crashed, and the machine has been up for
>about 22 days since then without a problem.
>

Well, that is what is supposed to happen, but what about in the case where
one of the daemons can't get enough memory to continue running? Say you've
got a rouge process which has chewed up all available mem + swap and is
still looking for more. Other programs will be competing for memory which
isn't there, and they too will terminate (gracefully or not). I've had this
happen where an apfd (from the netatalk package) process grabbed absolutely
all available memory, but didn't exit. It wasn't too long before important
stuff like ypserv died. It will really depend on how each process deals
with a no memory situation, but in general, it's a bad thing to run out of
swap.

Mark