Return to BSD News archive
Path: euryale.cc.adfa.oz.au!newshost.anu.edu.au!harbinger.cc.monash.edu.au!news.rmit.EDU.AU!news.unimelb.EDU.AU!munnari.OZ.AU!news.mel.connect.com.au!news.uwa.edu.au!disco.iinet.net.au!news.uoregon.edu!vixen.cso.uiuc.edu!newsfeed.internetmci.com!in2.uu.net!world!coopnews.coop.net!hops.entertain.com!stout.entertain.com!not-for-mail From: dwatson@stout.entertain.com (Darryl Watson) Newsgroups: comp.unix.bsd.bsdi.misc Subject: MAX_USERS Date: 23 May 1996 12:07:52 -0600 Organization: ABWAM, Inc., full service ISP, voice: 1+ 303 730-6050 Lines: 37 Message-ID: <4o29ho$1q9@stout.entertain.com> NNTP-Posting-Host: stout.entertain.com Hi all! I have BSDI 2.0 with patches. Our main server is quite busy with dialups through a Digiboard, primary nameservice, Apache httpd, and ftp service. Right now, my kernel is set with MAX_USERS=64, and we typically have 25-30 users telnetted or dialed into the server. In addition, we usually have at least 10 ftp sessions going at any time. Lately, as our ftp traffic has increased, our server will occasionally start failing to serve POP requests, new telnet or ftp sessions, and other inetd services. Inetd goes into runaway mode, and the syslog reports that inetd can't create any new processes, or there are too many files open. My understanding is that the MAX_USERS kernel parameter controls more than the max number of users you can have logged in, but also gives the kernel hints as to how many files can be open on the system, how many processes can be created, etc. Questions: Will raising the MAX_USERS parameter from 64 to, say, 96 tend to get rid of the inetd race situation? Is there a practical maximum on the MAX_USERS parameter? Am I messing with a parameter that could seriously affect the performance of the system in mysterious ways? Am I on target in my analysis of what the inetd race problem is caused by? Any info appreciated! I won't be able to read netnews for about a week, so please email dwatson@abwam.com with suggestions. Thanks! P.S.: We are preparing to increase RAM on the system from 32megs to at least 64megs. We've got plenty of swap space (200 megs or so!) so we expect better performance, but maybe putting MAX_USERS at 96 will require us to do the memory upgrade first, practicially speaking...