Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10.2 9/5/84; site spp5.UUCP
Path: utzoo!watmath!clyde!burl!ulysses!bellcore!decvax!ittatc!dcdwest!sdcsvax!sdcrdcf!trwrb!trwspp!spp5!brahms
From: brahms@spp5.UUCP (Bradley S. Brahms)
Newsgroups: net.rumor
Subject: Re: Computer Horror Stories
Message-ID: <108@spp5.UUCP>
Date: Tue, 25-Feb-86 19:22:15 EST
Article-I.D.: spp5.108
Posted: Tue Feb 25 19:22:15 1986
Date-Received: Sat, 1-Mar-86 00:46:24 EST
References: <14700001@hplabsb.UUCP> <476@mmm.UUCP>
Reply-To: brahms@spp5.UUCP (Bradley S. Brahms)
Organization: TRW, Redondo Beach  CA
Lines: 34

In article <1215@brl-smoke.ARPA> ron@brl-smoke.ARPA (Ron Natalie <ron>) writes:
>> If you kick off a shell that infinitely forks these babies, the only way
>> to kill the system in less than about 3 days is to drop it on the floor.
>> The bad thing is that anyone can do it, no special privleges are
>> necessary.  I won't post the source for obvious reasons, but it is a
>> wonderous way to get everyone competing for your resources to leave...
>> I have never done it in a working environment, but Man have I been tempted!!
>> 
>Most UNIX systems these days give a user process limit that is significantly
>less than infinity or NPROC.
>
>-Ron

There is another way to deal with something like this that works.
Presumably, the process that is doing the forking is running at know
higher a priority than anything else.  What then can be done is do the
following:

	Renice a su sh to a very high priority.

	Run a compute bound job at a lower priority than the su sh but
	higher then the problem process.

	At this point, the process will not get any run time.  You can
	then do a ps and kill the processes off.

	Kill the compute bound process and your done.

			-- Brad Brahms
			   usenet: {decvax,ucbvax,ihnp4}!trwrb!trwspp!brahms
			   arpa:   Brahms@usc-eclc

The opinions expressed above are my own, and may not reflect those
of my employer.