Date: Fri, 08 Feb 2008 18:37:32 +0100 From: Frank Staals <frankstaals@gmx.net> To: doug@fledge.watson.org Cc: Jerry McAllister <jerrymc@msu.edu>, FreeBSD <freebsd-questions@freebsd.org> Subject: Re: buildworld releng7 exterme performance loss Message-ID: <47AC935C.7060909@gmx.net> In-Reply-To: <47AB4985.7040607@gmx.net> References: <47A3720D.1080702@gmx.net> <47AA08E8.1020502@gmx.net> <20080206193916.GA20244@gizmo.acns.msu.edu> <20080206144920.G12870@fledge.watson.org> <47AB4985.7040607@gmx.net>
next in thread | previous in thread | raw e-mail | index | archive | help
Frank Staals wrote: > doug@safeport.com wrote: <message snipped> > > As suggested I turned on all.log and also captured the output from > vmstat -i and top -S as 2 others suggested. From the all.log output I > don't seem to see anything out of the ordinary if I compare it to a > snapshot of all.log on my working environment but maybe I am looking > for the wrong thing ? The same applies for the output of top -S. > vmstat's output may be more interesting though. In both cases I the > information was taken just after boot with a login on tty0 and 1 > remote session. vmstat shows the amount of interrupts for my network > interfaces are a lot higher with todays build compared to my working > env from 3 weeks ago. There is one difference though: I did a rm -rf > /usr/obj/* through ssh on the todays build. Could that be the reason > for the difference or did I find the source of the problem ? > > info : > > all.log > todays build : http://fstaals.net/junk/rena/all.log.2008.07.02.1140 > working env: http://fstaals.net/junk/rena/all.log.stable > > top -S output (links so the output is still properly formatted ) > todays build: http://fstaals.net/junk/rena/top > working env: http://fstaals.net/junk/rena/top_stable > > vmstat -i on todays build: > frank@Rena# cat /root/vmstat > interrupt total rate > irq1: atkbd0 610 2 > irq9: acpi0 1 0 > irq12: psm0 6 0 > irq14: ata0 58 0 > irq17: wpi0 bge0+ 151862 706 > irq19: cbb0+ 2 0 > irq20: uhci0 uhci+ 1 0 > irq21: pcm0 uhci1+ 8 0 > irq22: ehci0 uhci4 5 0 > cpu0: timer 179565 835 > cpu1: timer 169770 789 > Total 501888 2334 > > frank@Rena# cat /root/vmstat_stable > interrupt total rate > irq1: atkbd0 130 1 > irq9: acpi0 1 0 > irq14: ata0 58 0 > irq17: wpi0 bge0+ 3544 44 > irq19: cbb0+ 2 0 > irq20: uhci0 uhci+ 1 0 > irq21: pcm0 uhci1+ 7 0 > irq22: ehci0 uhci4 5 0 > cpu0: timer 155701 1946 > cpu1: timer 145928 1824 > Total 305377 3817 > > > to see if the values from vmstat were normal I tried a couple things again. I once more recompiled the latest sources and disabled my wpi wlan interface (since if the vmstat value was incorrect the most obvious reason would have been something with wpi I thought. ). Unfortunately that didn't help much. Second I again cleared /usr/obj/ through an ssh session. The vmstat value for bge0 (my LAN interface) was >160 000. Lastly I recompiled my normal sources again and removed /usr/obj/* again through ssh. Allthough the number of interrupts is still quite big it is only half of the amount it shows with todays sources. So now I am once again lost on what to do to determine the cause of my problem: frank@Rena# vmstat -i interrupt total rate irq1: atkbd0 1898 2 irq9: acpi0 1 0 irq12: psm0 84 0 irq14: ata0 58 0 irq17: wpi0 bge0+ 83570 110 irq19: cbb0+ 2 0 irq20: uhci0 uhci+ 3 0 irq21: pcm0 uhci1+ 2217 2 irq22: ehci0 uhci4 5 0 cpu0: timer 1484618 1956 cpu1: timer 1474830 1943 Total 3047286 4014 Anyone ideas ? -- -Frank Staals
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?47AC935C.7060909>