[CentOS-docs] New page on installing to software RAID

Tue Apr 28 15:04:22 UTC 2009
PJ Welsh <pjwelsh at gmail.com>

On Tue, Apr 28, 2009 at 4:15 AM, Karanbir Singh <mail-lists at karan.org>wrote:

> PJ Welsh wrote:
> > mdadm can resync, but so can the /proc with:
>
> I must be missing something. What makes mdraid fail when a new kernel is
> installed ?


I focused on the re-syncing part not the kernel part.


>
>
> > "echo check > /sys/block/md#/md/sync_action" for CentOS 5.x where "#" is
> > the number. In fact, if you don't do this on a regular basis, you *ARE*
> > asking for trouble!
>
> umm.. Do you have some more details to back this statement up ? Are you
> saying that the image code is somehow broken in mdraid ?
>
> - KB
>
>
UDE's (*undetected disk errors) and URE's (unrecoverable read errors) are
growing very common with large capacity drives! These are silent killers of
storage systems. ZFS is the only one at this time that automatically
validates parity and checks for controller or drive issues actively. Every
commercial product has a parity scrubbing process and so does Linux.

http://en.wikipedia.org/wiki/Standard_RAID_levels (look in the "*RAID 5 disk
failure rate" sectinon) and lots of recent articles and postings:

It is essential to periodically validate the actually data on disk with a
parity scrub!
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.centos.org/pipermail/centos-docs/attachments/20090428/df538576/attachment-0006.html>