Date: Tue, 29 Oct 2002 02:13:25 -0500 From: John De Boskey <jwd@freebsd.org> To: Current List <freebsd-current@freebsd.org> Subject: Poor 5.0/nfs performance Message-ID: <20021029071325.GA7231@BSDWins.Com>
next in thread | raw e-mail | index | archive | help
Hi, I have a 5.0 system from 10/27. In an attempt to improve performance I commented out the INVARIANTS/WITNESS options: #options INVARIANTS #Enable calls of extra sanity checking #options INVARIANT_SUPPORT #Extra sanity checks of internal structures, required by INVARIANTS #options WITNESS #Enable checks to detect deadlocks and cycles #options WITNESS_SKIPSPIN #Don't run witness on spinlocks for speed and then started a make release. Since doing this, the machine has become almost totally unresponsive. Command execution is measured in hours. A page from top which finally came up shows some very high load averages: last pid: 1892; load averages: 7.14, 6.00, 5.67 up 1+10:49:19 23:50:26 34 processes: 1 running, 33 sleeping CPU states: 0.0% user, 0.0% nice, 78.5% system, 0.8% interrupt, 20.7% idle Mem: 69M Active, 909M Inact, 214M Wired, 51M Cache, 112M Buf, 255M Free Swap: 4096M Total, 36K Used, 4096M Free PID USERNAME PRI NICE SIZE RES STATE TIME WCPU CPU COMMAND 99689 root -8 5 2928K 2236K biowr 0:59 0.00% 0.00% cvs The cvs is being executed by 'make release' updating the chroot area. The repo lives in /home/ncvs which is an nfs mount of a 4.7 system. A kernel with the above options does not exibit this behaviour. When I killed the cvs process, the machine returns to normal. I guess my basic question is: Are the INVARIANTS and WITNESS options required at this point? Thanks, John To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-current" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20021029071325.GA7231>