bonnetf@apo.esiee.fr (bonnet-franck) (06/04/91)
Hi, Is there a way to limit the number of processes running on a specific machine or not ? -------------------------------------------------------------------------------| bonnetf@apo.esiee.fr | | Frank Bonnet | Surfing ... | E.S.I.E.E | | BP99 93162 Noisy le Grand cedex.FRANCE. | the rest is details ! | Fax : 33 1 45 92 66 99 | | -------------------------------------------------------------------------------|
mottram@viper.ntl02.decnet.nokia.fi (Peter Mottram) (06/04/91)
bonnetf@apo.esiee.fr (bonnet-franck) writes: > Is there a way to limit the number of processes running on a > specific machine or not ? I personally would be very wary about doing such, whether possible or not. What happens if you have a problem with a node and are unable to create that one extra process required to fix it?? Hit the big red button?? Not me - I prefer the simple life. Peter Mottram tel: +44 223 423123 System Support fax: +44 223 423139 Nokia Telecommunications Ltd inet: mottram@ntl02.decnet.nokia.fi UK
dpassage@tornado.Berkeley.EDU (David G. Paschich) (06/05/91)
In article <MOTTRAM.91Jun4170602@viper.ntl02.decnet.nokia.fi> mottram@viper.ntl02.decnet.nokia.fi (Peter Mottram) writes: >bonnetf@apo.esiee.fr (bonnet-franck) writes: > >> Is there a way to limit the number of processes running on a >> specific machine or not ? > > >I personally would be very wary about doing such, whether possible or not. > >What happens if you have a problem with a node and are unable to create >that one extra process required to fix it?? Hit the big red button?? >Not me - I prefer the simple life. We ran into this problem a while back under 10.2 on a 4500, when the 64 process limit still existed. We were using the 4500 to gateway mail for around 500 active users, and it would regularly wedge itself with large numbers of sendmail processes. To unwedge the machine, we wrote a daemon which would watch for the existence of a file, and if it existed, kill a certain class of processes, such as sendmails. We could then touch the file from another node. Since the daemon process was always running, we didn't need to fork a new process to fix the problem (or at least keep things running). David G. Paschich Open Computing Facility UC Berkeley dpassage@ocf.berkeley.edu "Everybody wants prosthetic foreheads on their real heads." --They Might Be Giants