Date: Sun, 31 May 2009 14:25:36 -0400 From: grarpamp <grarpamp@gmail.com> To: freebsd-fs@freebsd.org Subject: ZFS and idprio Message-ID: <d2e731a10905311125j1dd7f0ads15b475394e33eb26@mail.gmail.com>
next in thread | raw e-mail | index | archive | help
On an i386 with 1GB ram and P4 2.x GHz single core with arc_max at 64MiB and everything else from RELENG_7_1_0_RELEASE left at default, with this disk layering: PDC20269@100 > PATA > geli > zfs [gzip sha256]. The system become unusable slow when doing heavy sequential disk IO. Such as copy 50G data, zpool scrub, etc. These two processes [*] eat the cpu. I did idprio 31 on them and got 'ki-1' in the nice column [was '-'] and seemed some responsiveness back after a few minutes. Should I be able to nice these kernel threads? Any other way to drop their priority? What is 'ki-1'? Top doesn't seem to show that field right as with say nice -20 sleep 60? Sorted by time: 11 root 1 171 ki31 0K 8K RUN 25.2H 93.90% idle: cpu0 * 594 root 1 -8 - 0K 8K geli:w 321:27 0.68% g_eli[0] ad4 * 695 root 1 -16 - 0K 24K tq->tq 252:33 0.00% spa_zio_intr_1 * 597 root 1 -8 - 0K 8K geli:w 181:52 0.00% g_eli[0] ad5 * 729 root 1 101 - 0K 24K tq->tq 131:01 0.00% spa_zio_intr_1 758 iso 1 44 0 172M 49572K select 62:43 0.00% XFree86 740 root 1 -16 - 0K 24K tx->tx 44:00 0.00% txg_thread_enter Not under load: load averages: 0.12, 0.40, 0.51 up 2+01:11:50 14:19:22 164 processes: 4 running, 141 sleeping, 19 waiting CPU: 0.8% user, 0.0% nice, 15.0% system, 0.8% interrupt, 83.5% idle Mem: 147M Active, 146M Inact, 265M Wired, 24M Cache, 47M Buf, 410M Free
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?d2e731a10905311125j1dd7f0ads15b475394e33eb26>