I'm doing a big rsync to my 5 disk RAID 5 partition and am seeing an
uneven loading of my disks. The files are almost all huge video files.
Has anyone got any idea what I could have done wrong?
md6 : active raid5 sdi1[4] sdh1[3] sdg1[2] sdf1[1] sde1[0]
976783360 blocks level 5, 128k chunk, algorithm 2 [5/5] [UUUUU]
The filesystem is ext3 created with
mkfs.ext3 -R stride=32 /dev/md6
Disks are all identical 250G Hitachi SATA on a Marvell 8 port PCI-X SATA
controller.
"iostat -x 5" shows something like, (sorry about the wrap)
Device: rrqm/s wrqm/s r/s w/s rsec/s wsec/s rkB/s wkB/s
avgrq-sz avgqu-sz await svctm %util
sde 36.80 291.20 10.60 143.00 380.80 3486.40 190.40 1743.20
25.18 1.76 11.37 0.86 13.14
sdf 30.20 293.80 4.60 147.60 280.00 3531.20 140.00 1765.60
25.04 1.82 11.95 0.88 13.40
sdg 43.20 316.40 10.40 117.20 432.00 3521.60 216.00 1760.80
30.98 2.75 20.47 1.35 17.22
sdh 47.60 286.60 10.40 146.80 464.00 3470.40 232.00 1735.20
25.03 2.51 15.96 1.05 16.50
sdi 43.20 344.00 8.40 88.40 412.80 3425.60 206.40 1712.80
39.65 34.86 391.68 10.33 100.04
So sdi is 100% utilized with await 10x the other disks and the other
disks are not busy. When the array was building it was sdi that was
doing all the writes and the other 4 disks just doing reads but I
figured this was just the fastest way to build the array, now I wonder
if I have ended up with parity all on one disk like RAID4? Or is one
disk bad and is there any way to find out?
Scott
-
To unsubscribe from this list: send the line "unsubscribe linux-raid" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
