[ale] Software RAID1 rebuild after HDD failure (How to do?)

Jim Popovitch jimpop at yahoo.com
Fri Jun 30 20:59:37 EDT 2006


David Corbin wrote:
> On Friday 30 June 2006 08:42 pm, Jim Popovitch wrote:
>> Joe Steele wrote:
>>> By way of comparison:  when first building a new raid array, you
>>> partition the drives, then you run mdadm to assemble the array.  Only
>>> then do you format the assembled array.  The device that gets formated
>>> is /dev/md0 (the raid device), not /dev/hda1 (a partition that is part
>>> of the array).
>>>
>>> If a drive fails, its partition(s) are removed from the array, a new
>>> drive is installed with partitions of the same size as those of the
>>> failed drive, and the partition(s) are added back to the array.  At that
>>> point, the raid system begins to rebuild the data on the partitions,
>>> which includes rebuilding whatever formatting belongs there.
>> Good point Joe.  Ryan, just make sure you don't re-format /dev/md0, let
>> mdadm sync the data from the first drive to the second drive. ;-)
> 
> and once you do, you can monitor progress with "cat /proc/mdstat".

One other thing that comes to mind is that to increase the raid recover 
speed you need to cat a few values to /proc like this:

echo 40940 > /proc/sys/dev/raid/speed_limit_min
echo 60000 > /proc/sys/dev/raid/speed_limit_max

Google for speed_limit_min|max to get an idea on acceptable values.


-Jim P.







More information about the Ale mailing list