Software or hardware RAID1. (gstripe)

pete wright nomadlogic at gmail.com
Wed Oct 27 21:20:08 PDT 2004


On Fri, 27 Aug 2004 21:40:22 +0000, Thordur I. <thib at mi.is> wrote:
> I´m currently trying to decide upon a sultion on how to go about keepin
> my data safe.
> 
> Now, a RAID1 setup sounds good in this position. But I can´t seem to
> find an answer to one question.
> 
> Let´s say that I have 2x 120Gb (ad1 and ad2) disk. I stripe them wich
> should give something like /dev/stripe/foo  (either with gstripe or a
> hardware raid controler.)
> 
> Now I fill this striped volume with data, and ad1 dies.
> What happen´s ?
> Does gstripe alert my that ad1 has died and will I still be able to
> access /dev/stripe/foo ?
> And when I shut the box down, remove ad1 and insert another (identical
> disk) to restore the stripe, what then ? (assume the same bus ofcourse)
> 
> I have tryed to google this up but to no avail.
> 
> Any help is well appriceated.
> 

here is a good link regarding raid levels in general

http://www.acnc.com/04_01_00.html

what it sounds like is you want to create a mirror of disks for
maximum data integrity.  depending on how much disk i/o you are going
to be doing i think using a hardware RAID solution on one level or
another is probably the best choice.  not only does it free the OS up
from having to manage the RAID but you also gain benefits of onboard
cache of the disk controllers, and if you are lucky decent management
firmware to maintain the array(add new disks, verification, etc).  the
idea with striping is to be able to survive a failure of one of the
disks while being reasonably sure that the data integrity on those
disks will be intact.  most hardware raid controllers will alert you
when a disk fails, as i'm sure gstripe will as well altho i have not
used it personally.
     the next step is rebuilding, this is usually also managed by the
hardware controller or via the RAID implementation on the OS level
(vinum etc..) and should be automatic when a new working disk is
available.  If you are lucky to have hot-swapable disks a power cycle
is not even neccessary.

now if you have the money/space for multiple disks you it may worth
looking at RAID 5, which will give you decent i/o and decent fault
tolerance.  or you can start playing with stripping disks togther in
LUN's then mirroring set's of the LUN's together but that is usually
an expensive route (for disks, controllers and high-speed
interconnects).

as always the specific man pages for which ever implementation are the
best sources for info.

HTH

-p

-- 
~~o0OO0o~~
Pete Wright
www.nycbug.org
NYC's *BSD User Group


More information about the freebsd-questions mailing list