on Wed, May 23, 2001 at 11:33:45AM +0700, Oki DZ (okidz@bdg.pindad.com) wrote: > "Karsten M. Self" wrote: > > I usually try to track down process relationships with 'pstree', then > > try killing related process with 15, 1, 2, and, if all else fails, 9. > > killall -9 <program name> would be much more efficient, right? Using kill -9 on a process means you may have to clean up the pieces. Signals 15, 1, and 2 (TERM, HUP, and INT), are generally considered to be polite requests to jobs to get the hell over it already, but to clean up on the way out. SIGKILL is nonmaskable, and a process *can't* perform cleanup or garbage collection even if it wants to. > > True unkillable zombies are rather rare. > > Usually, it's pretty difficult; even though they are already > half-dead. Most zombies are waiting for a resource to close. Hitting the other end of the resource (parent or child) generally does same. > BTW, speaking about killing processes... > Once I had a daemon that couldn't be killed because the client program > that once connected to it didn't close the ports properly. So the daemon > was just there sitting, waiting for a time-out. But it didn't happen. > Magically, kill -9 didn't work; the daemon materialized as a zombie, yet > it was a strong one. This is a case where you may have to shut down. Sometimes you can get the buggers if you shoot at 'em enough ways though. > It was pretty amusing... how long would a TCP connection time out given > that any other side doesn't close the connection (and none is reading > the ports)? Or, it just simply wouldn't time out? > > Oki -- Karsten M. Self <kmself@ix.netcom.com> http://kmself.home.netcom.com/ What part of "Gestalt" don't you understand? There is no K5 cabal http://gestalt-system.sourceforge.net/ http://www.kuro5hin.org Disclaimer: http://www.goldmark.org/jeff/stupid-disclaimers/
Attachment:
pgpqYhIrzKfOt.pgp
Description: PGP signature