Date: Sun, 9 Mar 2008 00:12:13 -0700 From: "Edward Sanford Sutton, III" <mirror176@cox.net> To: freebsd-geom@freebsd.org Subject: gvinum raid5 consuming entire cpu on FreeBSD 7.0 Message-ID: <200803090012.13711.mirror176@cox.net>
next in thread | raw e-mail | index | archive | help
Upgrading from 6-stable to 7.0-release has lead to a couple of immediately noticeable changes. The striped swap across 3 disks, with 1 disk having been removed and reinserted could finally be brought to an up state again with a start command; was not available when last tested on 6-stable. Within top, I've found the system process "gv_p usr.p0" taking up all cpu time on my system when there is a lot of write activity to usr.p0. The result is the system becoming almost entirely unresponsive during much of the write. An example case is when trying to build the editors/openoffice.org-2 port; when archives are being extracted, services time out connections to clients, screens stop refreshing (and top shows the system process at 100% once it starts to get a little bit of time again), mouse/keyboard input becomes bufffered and all goes through at once when some cpu time is free. The openoffice.org build running with nice 20 still causes major service interruption. The only good response I still get is switching between virtual text consoles (alt + F2, alt + F3, etc). Upgrading lead to a few changes being required in the kernel; do any of the following sound like optional changes I made that I can remove to try improving the performance: options STOP_NMI # Stop CPUS using NMI instead of IPI options AUDIT # Security event auditing options SMP # Symmetric MultiProcessor Kernel device cpufreq device uart # Generic UART driver wlan additions, although I have no wireless hardware on this machine. Any other ideas how to get the load under concrol? I remember being able to bogg down my system with an excessive mysql database fro disk load before, but do not recall reaching such cpu load with anything that it could seem the computer froze for input/output for more than a few seconds max; freezes can now go for over a minute. Thanks again, Ed Sutton
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?200803090012.13711.mirror176>