[CentOS] An mdadm question

Thu Jul 2 16:00:09 UTC 2015
Chris Murphy <lists at colorremedies.com>

On Thu, Jul 2, 2015 at 8:24 AM,  <m.roth at 5-cent.us> wrote:
> CentOS 7.
>
> I have a server with four drives. 1 is /, and the other three are in
> RAID5. I need to pull a drive, so I can test whether the server can read >
> 2TB drives. I've been googling, but don't want to screw the server up....
>
> I think I'd like to
>   1. stop the RAID
>   2. pull a drive
>   3. put in a large drive, and run parted, and mkfs
>   4. pull the large drive
>   5. replace the RAID drive
>   6. fire up the RAID.
>
> So, can I do it in that order? Do I need to fail something?

Almost invariably that RAID will assemble degraded, and the drive you
pulled will now be out of sync. When you add it back in, it will have
to be, block by block, resynced. So if you want to avoid that, you
have to avoid the autoassembly at boot time, or pull at least one more
drive to prevent assembly.

Even live media may assemble the RAID, degraded, if enough drives for
degraded operation are found.

-- 
Chris Murphy