From owner-freebsd-questions Fri Oct 19 15:53:27 2001 Delivered-To: freebsd-questions@freebsd.org Received: from globalrelay.com (h216-18-71-77.gtcust.grouptelecom.net [216.18.71.77]) by hub.freebsd.org (Postfix) with ESMTP id C127B37B403 for ; Fri, 19 Oct 2001 15:53:22 -0700 (PDT) Received: from [10.2.0.6] (HELO hpvl4002) by globalrelay.com (CommuniGate Pro SMTP 3.4.7) with SMTP id 720298; Fri, 19 Oct 2001 15:53:20 -0700 Message-ID: <039e01c158f1$168c9880$0600020a@internal.globalrelay.net> From: "Eric Parusel" To: "Ryan Thompson" , "Gary W. Swearingen" Cc: References: Subject: Re: Loads on a Web/Shell Server Date: Fri, 19 Oct 2001 15:53:16 -0700 MIME-Version: 1.0 Content-Type: text/plain; charset="Windows-1252" Content-Transfer-Encoding: 7bit X-Priority: 3 X-MSMail-Priority: Normal X-Mailer: Microsoft Outlook Express 6.00.2600.0000 X-MimeOLE: Produced By Microsoft MimeOLE V6.00.2600.0000 Sender: owner-freebsd-questions@FreeBSD.ORG Precedence: bulk List-ID: List-Archive: (Web Archive) List-Help: (List Instructions) List-Subscribe: List-Unsubscribe: X-Loop: FreeBSD.ORG I've had this before as well, and I was told that it was possibly some sort of odd result of the algorithm that figures out the loadavg... Here's an email I received after inquiring about this previously on -stable a few months ago: =============================== From: Ian Dowse iedowse@maths.tcd.ie =============================== >Twice today I've had this problem, and it's "weirding" me out... > >On two servers, both with a load average of normally 0.01, all of a >sudden got load averages of 1.1...! I did a ps, top, and there were >no processes that could account the high loadavg... ... >Once I did that on both servers, the load average promptly dropped >back down to 0.01...! This problem has been reported a few times; the last time I tried to investigate it I came to the conclusion that there is a weird synchronisation effect that confuses the load measurement. The code for calculating the load average periodically samples the set of runnable processes - a bit like doing ps -o state -axl | grep ^R and counting the lines, except that the kernel does the sampling so that you don't see `ps', `grep' and maybe `xterm' processes that occur with the above. The kernel performs a sample like this exactly once every 5 seconds, and averages the result to produce the load. My guess was that some process (system or user) also does something exactly every 5 seconds, and occasionally they get synchronised so that the load average sample always sees that process running. I'll probably eventually produce a patch that randomises the timing to avoid this effect. In the meantime you can probably consider it as a harmless bug. Ian ================================== ----- Original Message ----- From: "Gary W. Swearingen" To: "Ryan Thompson" Cc: Sent: Friday, October 19, 2001 1:11 PM Subject: Re: Loads on a Web/Shell Server > Ryan Thompson writes: > > > The load averages are, at best, a comparative indication of the change in > > load of one system over time. Unless your system is really unresponsive, > > you needn't pay much attention to the load averages. If your system IS > > really unresponsive, make a note of the load average, and see what is > > eating all of your resources. > > My desktop system usually has load < 0.10, but under 4.2 & 4.3, it would > a few times per day jump up to hover around 1.0 for about 20 minutes. > Network device lights show no internet traffic, "ps" and "top" show > nothing using unusual CPU or memory. Xosview (which runs continually) > shows nothing unusual. In several weeks with 4.4, I've only seen this > once, IIRC. Any ideas what could cause that behavior? (I'm fairly sure > that my system monitoring tools (or anything else) have not been cracked.) > > To Unsubscribe: send mail to majordomo@FreeBSD.org > with "unsubscribe freebsd-questions" in the body of the message > To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-questions" in the body of the message