Return to BSD News archive
Path: euryale.cc.adfa.oz.au!newshost.carno.net.au!harbinger.cc.monash.edu.au!munnari.OZ.AU!news.ecn.uoknor.edu!feed1.news.erols.com!news.maxwell.syr.edu!news.bc.net!mongol.sasknet.sk.ca!tribune.usask.ca!rover.ucs.ualberta.ca!scanner.worldgate.com!not-for-mail From: Marc Slemko <marcs@znep.com> Newsgroups: comp.unix.bsd.bsdi.misc Subject: Re: Unexplained CPU usage on Apache httpd Date: 12 May 1997 04:54:25 GMT Organization: WorldGate Inc. http://www.worldgate.com/ Lines: 58 Message-ID: <5l67q1$irc$1@scanner.worldgate.com> References: <33731156.1E6F@chesapeake.net> <33734b1f.105758849@news.clara.net> <33774d27.92779299@news.callnet.com> NNTP-Posting-Host: valis.worldgate.com Xref: euryale.cc.adfa.oz.au comp.unix.bsd.bsdi.misc:6855 In <33774d27.92779299@news.callnet.com> matt@callnet.com (Matthew Ahrens) writes: >actually, this is a documented bug between apache and BSDI. if one of >the apache server processes is not used (ie. no hits) in a long time >(like 12 hours or something) then it will *appear* to get stuck and >add one to your load. however, it does not degrade system performance. No, this is almost certainly a different problem. The load average thing is caused by several possible things, including the kernel scheduling of the process and the load average calculation. With the load average issue it does not actually consume huge amounts of cpu. With the below report, it is consuming huge amounts of CPU. >the guys at BSDI said they are working on this but it is low priority >since it's just a "cosmetic" bug. (maybe it is fixed in beta 10, don't >know). Not unless it is some coincidental change that fixed it. The probable cause is the fact that the Apache parent process wakes up once a second to do some (very) minimal housekeeping. If the system gets into a state where Apache's wakeup is synchronized with when the kernel calculates the load average, Apache will always appear to be running to the kernel. Introducing a small random delay should eliminate the problem, but isn't really worthwhile. >--matt >On Fri, 09 May 1997 16:12:31 GMT, steve@clara.net (Steve Rawlinson) >wrote: >>support@chesapeake.net wrote: >> >>>For the past several nights our Apache web server has unexplainably >>>frozen (almost) due to an extremely high CPU percentage. Using the TOP >>>caommand I can see that it is using 98% CPU and not accepting and any >>>requests from inside our network or out. >>> >>>The only way to fix it is to SIGHUP the httpd daemon. That seems to >>>work until the next time. >> >>I had the same problem which suddenly appeared and then suddenly >>dissappeared (running apache 1.2b7). I still dont know what caused it >>but it looked like an infinite loop bug - the process took as much cpu >>as it could get its hands on and sometimes there were more than one >>doing it. >> >>steve >> >> >>------ >>Steve Rawlinson >>ClaraNet Ltd >>steve@clara.net >>0171 647 1000