> >ZFS will write buffer ALOT. Your settings of vfs.zfs.arc_min and vfs.zfs.arc_max will likely affect how bursty this >is. I've been running with Min, 32 GB arc, and see very bursty writes (of course, my systems have 64GB). >Also, use 'zpool iostat 1' and not the system iostat. Check gstat as well. >also try using CAM_ATA (ahci.ko) instead of ad devices, so your can get the benefits of tag queuing. >There is also a possible undocumented issue with, I believe it was the WD green drives, where the vibration of the >drive actually slowed it down. This was demonstrated by setting a pound or two of silly putty on the top of the >drive, and its performance (significantly) increased. Experiment with putting some heavily, non-conductive weight on >top of each drive - a book or something. Even with hardware the FreeBSD Handbook comes Handy ;-) > ensure that is its securely mounted as well. >--- >David P. Discher >dpd_at_bitgravity.com * AIM: bgDavidDPD >BITGRAVITY * http://www.bitgravity.com >On Dec 18, 2009, at 12:04 PM, Alexander Zagrebin wrote: >>>> The disks are used for mirrored ZFS pool. >>>> I have noticed that the system periodically locks up on >>> disk operations. >>>> After approx. 10 min of very slow disk i/o (several KB/s) >>> the speed of disk >>>> operations restores to normal. >>>> gstat has shown that the problem is in ad6. >>> >>> Want you try this patch and report?: >>> http://people.freebsd.org/~attilio/lockmgr_lostwake.diff >>> >>> It should apply to STABLE_8 and if not, just report and I will help. >> >> Thanks fo your attempt to help! >> I've successfully applied this patch to RELENG_8_0, recompiled >> kernel and required modules (krpc, opensolaris,zfs), >> but nothing has changed. The problem still exist. >> >> -- >> Alexander Zagrebin >> >> _______________________________________________ >> freebsd-current_at_freebsd.org mailing list >> http://lists.freebsd.org/mailman/listinfo/freebsd-current >> To unsubscribe, send any mail to "freebsd-current-unsubscribe_at_freebsd.org"Received on Mon Dec 28 2009 - 07:21:31 UTC
This archive was generated by hypermail 2.4.0 : Wed May 19 2021 - 11:39:59 UTC