arnold@emory.UUCP (Arnold D. Robbins {EUCC}) (05/13/86)
I am the (relatively new) sys admin for two vax 780's running 4.2 BSD. Our version has had almost none of the bugs or problems fixed that have been posted to the net. We seem to have a problem whereby, when there are a large number of stopped jobs on the system (~20-40% of the total number of processes), the load average goes way up, and response time goes way down. Yet, to my understanding (which is probably very limited), they should not contribute to the load. I had conjectured that maybe the system was slowing down because Unix is doing a linear search of the process table every time it schedules another job to run. Yet, supposedly, these jobs should be like jobs waiting on i/o, and not even eligible for scheduling. Any enlightenment, in clear terms, would be appreciated, preferably by mail, since I am behind on my netnews reading. If this has been discussed before, I missed it, or skipped it, since I did not used to be quite so interested in these kinds of things, and I apologize. In that case, a summary would be appreciated. (I realize that having such a large number of stopped jobs is not a Good Thing, but it can be very hard to properly educate the users about how to correctly use job control. I am asking about this to find out how the system actually works, and what, if anything should be done about it, besides killing the jobs off and teaching the users what to do next time.) Thanks in Advance, -- Arnold Robbins CSNET: arnold@emory BITNET: arnold@emoryu1 ARPA: arnold%emory.csnet@csnet-relay.arpa UUCP: { akgua, decvax, gatech, sb1, sb6, sunatl }!emory!arnold "All this digital stuff is just a fad. Analog is the way to go." -- William M. Robbins, 1984