Re: Recovery from RAID5 Failure Questions

On 23 Jan 1997 19:54:14 +0800, trwi_at_perth.DIALix.oz.au (Trevor Williams) wrote:

>I would appreciate hearing about DBAs' experiences with handling disk
>problems where a failing RAID5 disk holds database files. In particular,
>I am interested in RAID5 on Solaris.
>
>Do databases continue to function after a single disk failure? How well?
>
>How trivial is the replacing of the failed disk? How can the outage time
>be minimised?
>
>After replacing a broken disk and getting Solaris to synchronise
>it, is it necessary to do a database recovery?
>
>What other things are useful to know about the process of recovery from
>RAID5 disk failure?
>
>
>
>
>

Most of your questions are hardware specific - the RAID hardware. Some RAID hardware allows you to re-build the missing disk on the fly, some do not.

I have experiance with Oracle7 on Digital UNIX with a Digital RAID array. We had several single disk failures in our RAID 5 sets. Our users kept right on working and noticed very little slowdown in their work. We hot swapped new drives in and re-built during off-peak times and again the users noticed very little impact to the overall system performance. There was no Oracle recovery needed because Oracle was un-aware that anything had happened.

My strongest recomendation is that you study RAID and its various implementations, talk to several hardware vendors, and then decide. Andataco has a decent white paper on RAID at: http://www.andataco.com/genesis/segment1/techinfo/white_papers/RAID/raid_1.html although all of the links do not seem to work correctly it is raid_1.html through something like raid_15.html

Good luck,
Chris Received on Sat Jan 25 1997 - 00:00:00 CST