On 18 May, 2010 - Philippe sent me these 6,0K bytes: > Hi, > > The 4 disks are Western Digital ATA 1TB (one is slighlty different) : > 1 x ATA-WDC WD10EACS-00D-1A01-931.51GB > 3 x ATA-WDC WD10EARS-00Y-0A80-931.51GB > > I've done lots of tests (speed tests + SMART reports) with each of these 4 > disk on another system (another computer, running Windows 2003 x64), and > everything was fine ! The 4 disks operate well, at 50-100 MB/s (tested with > Hdtune). And the access time : 14ms > > The controller is an LSI Logic SAS 1068-IR (MPT BIOS 6.12.00.00 - 31/10/2006) > > Here are some stats : > > 1) cp of a big file to a ZFS filesystem (128K recordsize) : > ============================================================= > iostat -x 30 > extended device statistics > device r/s w/s kr/s kw/s wait actv svc_t %w %b > sd0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0 0 > sd1 0.3 0.3 17.6 2.3 0.0 0.0 19.5 0 0 > sd2 11.5 6.0 350.1 154.5 0.0 0.3 19.5 0 4 > sd3 12.5 5.7 351.4 154.5 0.0 0.5 27.1 0 5 > sd4 15.9 6.3 615.1 153.8 0.0 1.3 58.2 0 8 > sd5 15.1 8.1 600.4 150.7 0.0 7.6 326.7 0 31 > extended device statistics > device r/s w/s kr/s kw/s wait actv svc_t %w %b > sd0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0 0 > sd1 41.3 0.0 5289.7 0.0 0.0 1.3 31.0 0 4 > sd2 4.2 24.1 214.0 1183.0 0.0 0.5 19.4 0 4 > sd3 3.7 23.6 227.2 1183.0 0.0 2.1 78.5 0 12 > sd4 6.6 26.4 374.2 1179.4 0.0 10.1 306.5 0 35 > sd5 4.3 31.0 369.6 973.3 0.0 22.0 622.0 0 96 > extended device statistics > device r/s w/s kr/s kw/s wait actv svc_t %w %b > sd0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0 0 > sd1 17.1 0.0 2184.6 0.0 0.0 0.5 30.6 0 2 > sd2 1.6 12.3 116.4 570.9 0.0 0.6 41.3 0 3 > sd3 1.6 12.1 107.6 570.9 0.0 10.3 754.7 0 33 > sd4 2.1 12.6 187.1 569.4 0.0 9.4 634.7 0 28 > sd5 0.4 21.7 25.6 700.6 0.0 29.5 1338.1 0 96
Umm.. Service time of sd3..5 are waay too high to be good working disks. 21 writes shouldn't take 1.3 seconds. Some of your disks are not feeling well, possibly doing block-reallocation like mad all the time, or block recovery of some form. Service times should be closer to what sd1 and 2 are doing. sd2,3,4 seems to be getting about the same amount of read+write, but their service time is 15-20 times higher. This will lead to crap performance (and probably broken array in a while). /Tomas -- Tomas Ögren, st...@acc.umu.se, http://www.acc.umu.se/~stric/ |- Student at Computing Science, University of Umeå `- Sysadmin at {cs,acc}.umu.se _______________________________________________ zfs-discuss mailing list zfs-discuss@opensolaris.org http://mail.opensolaris.org/mailman/listinfo/zfs-discuss
