Date: Fri, 4 Feb 2011 15:47:14 +0000 (UTC) From: Alexander Motin <mav@FreeBSD.org> To: src-committers@freebsd.org, svn-src-projects@freebsd.org Subject: svn commit: r218281 - projects/graid/head/sys/geom/raid Message-ID: <201102041547.p14FlE3x089371@svn.freebsd.org>
next in thread | raw e-mail | index | archive | help
Author: mav Date: Fri Feb 4 15:47:14 2011 New Revision: 218281 URL: http://svn.freebsd.org/changeset/base/218281 Log: When volume is not idle, use shorter rebuild requests and truncate size of requests that already running. This should be a compromise between faster rebuild and lower payload performance degradation. Not exactly solution I would like to see, but still better. Modified: projects/graid/head/sys/geom/raid/tr_raid1.c Modified: projects/graid/head/sys/geom/raid/tr_raid1.c ============================================================================== --- projects/graid/head/sys/geom/raid/tr_raid1.c Fri Feb 4 15:45:48 2011 (r218280) +++ projects/graid/head/sys/geom/raid/tr_raid1.c Fri Feb 4 15:47:14 2011 (r218281) @@ -68,7 +68,7 @@ SYSCTL_UINT(_kern_geom_raid_raid1, OID_A &g_raid1_rebuild_fair_io, RAID1_REBUILD_FAIR_IO, "Fraction of the I/O bandwidth to use when disk busy for rebuild."); -#define RAID1_REBUILD_CLUSTER_IDLE 10 +#define RAID1_REBUILD_CLUSTER_IDLE 100 static int g_raid1_rebuild_cluster_idle = RAID1_REBUILD_CLUSTER_IDLE; TUNABLE_INT("kern.geom.raid.raid1.rebuild_cluster_idle", &g_raid1_rebuild_slab); @@ -578,6 +578,8 @@ g_raid_tr_iostart_raid1(struct g_raid_tr !(bp->bio_cflags & G_RAID_BIO_FLAG_SPECIAL)) { if (--trs->trso_fair_io <= 0) g_raid_tr_raid1_rebuild_some(tr, trs->trso_failed_sd); + /* Make this new or running now round short. */ + trs->trso_recover_slabs = 0; } switch (bp->bio_cmd) { case BIO_READ:
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?201102041547.p14FlE3x089371>