On 09/24/2013 02:18 PM, Karel Walters wrote:

 Hi Phil,

Thank you for all the great help so far!

quoted

Yes, that dmesg did the trick.  The drive that failed first was #3, and
the drive the failed second was #4.  You should create a list of which
drive serial number corresponds to which raid device role, with a third
column showing the current device name.

Serial no:              old name  ID     current name
WD-WCC1T1255024                     /dev/sdc1  new drive
W1F09XLV            /dev/sdb1 [3]    /dev/sdd1  failed drive 1
W1F0AXTQ           /dev/sdc1 [4]    /dev/sde1  failed drive 2
W1F0B6X6            /dev/sdd1 [0]    /dev/sdf1
S1F04BZT             /dev/sde1 [5]    /dev/sdg1
W1F0B9ER           /dev/sdf1 [2]    /dev/sdh1
WD-WMC1T2341606    /dev/sdg1 [1]    /dev/sdi1
S1F04CWH           /dev/sdh1 [6]    /dev/sdj1  (partially rebuild spare)

Ok, so your create operation will be:

mdadm --create /dev/md1 --level=10 -n 6 --layout=f2 --chunk=64
--data-offset=variable /dev/sdd1:2048 /dev/sdg1:4096 /dev/sdf1:2048
/dev/sdb1:2048 /dev/sdc1:2048 /dev/sde1:2048

I'm actually guessing that /dev/sdb1 and /dev/sdc1 need offset 2048 like
the original devices, not the 4096 of a device added later (newer
mdadm).  With the mixed offsets, you need mdadm version 3.3.

Use "fsck -n" to verify the array before mounting anything, just in case
one or both of those drives really does need :4096.

quoted

Then, to deal with the large number of pending events, you'll need to do
a "check" scrub with a very low speed limit.  To keep you from exceeding
the 10/hour read error limit in the MD kernel driver.

echo 1000 > /proc/sys/dev/raid/speed_limit_min
echo 10000 > /proc/sys/dev/raid/speed_limit_max
echo check > /sys/block/md0/md/sync_action

Would this be ok in such a case?

Looks ok.  You may want to experiment in progress.

Phil