zpool scrub speed

Steve Costaras stevecs at chaven.com
Fri May 20 13:59:21 EDT 2011


Thanks, ok, so you created the pool with all drives initially (didn't add another vdev or anything later). 

The drive you mention may show 512b sectors however this line:

"Drive Configuration 512 bytes per sector with Emulation" at http://www.samsung.com/latin_en/consumer/monitor-peripherals-printer/hard-disk-drives/internal/HD154UI/CKW/index.idx?pagetype=prd_detail&tab=specification
seems to imply that it may be 4K internally and just 'lying' to the os. (same thing I see with a lot of newer drives so it's hard to tell).

A couple things that may help. First you can try decreasing the zfs_vdev_max_pending to 1 (default is 10) (assuming Brian has this as a passable variable?)

echo "1" > /sys/module/zfs/parameters/zfs_vdev_max_pending

Other item would be to modify or pull a new version of zpool and create a new pool with ashift=12 opposed to ashift=9 (4KiB opposed to 512b). (unfortunatly this would require re-creating the pool from scratch).



-----Original Message-----
From: Nils Bausch [mailto:nils.bausch at googlemail.com]
Sent: Friday, May 20, 2011 11:44 AM
To: 'zfs-discuss'
Subject: Re: zpool scrub speed

....
 pool: wtf
 state: ONLINE
 scan: scrub repaired 2K in 14h54m with 0 errors on Thu May 19
10:27:15 2011
config:

	NAME STATE READ WRITE CKSUM
	wtf ONLINE 0 0 0
	raidz1-0 ONLINE 0 0 0
	ata-SAMSUNG_HD154UI_S1XWJ1KSB33767 ONLINE 0 0 0
	ata-SAMSUNG_HD154UI_S1XWJ1KSB33765 ONLINE 0 0 0
	ata-SAMSUNG_HD154UI_S1XWJ1KSC11001 ONLINE 0 0 0
	ata-SAMSUNG_HD154UI_S1XWJ1MSC01749 ONLINE 0 0 0

errors: No known data errors

> And what type of drives are these (i.e. are they 4KiB sector sized drives internally?)

Model Family: SAMSUNG SpinPoint F2 EG series
Device Model: SAMSUNG HD154UI
1.5 TB and they have 512 bytes as sector size.

Thanks,
Nils

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://list.zfsonlinux.org/pipermail/zfs-discuss/attachments/20110520/1676b6c8/attachment.html>


More information about the zfs-discuss mailing list