RAID5 Upgrade

Thanks for the reply, Alex. I should have mentioned in my original post that this machine is maxed out with disks; there are no available disk slots to install the new disks. The vendor wants about 4x as much as the disk costs for a new cabinet to hold the disks, and we are not going to pay that, since this machine is going to be replaced in the next year or two. So before we can add a new disk, we have to remove an old disk.
Returning to my original question, has anyone done anything similar to what is proposed below?

>>

Why not install new array, connect it to the same box, create file systems and directories
on it, put tablespaces one by one into backup mode and copy files to the new raid array including control files. Also copy oracle software and admin directories if they are also on old array. Then shutdown the database (not abort), change everywhere in parameter file old directories to new directories if needed, open database on the old array in mount mode and change file names to new directories - it can be scripted. Then you can open database and take away old array. Downtime should me minimal.

Alex Hillman
>>

-----Original Message-----
Sent: Wednesday, December 20, 2000 4:56 PM To: Multiple recipients of list ORACLE-L

Hello,

Someone made a clever(?) proposal regarding a disk upgrade at our site, and I'm wondering what to make of it.

First, we run Oracle 8.1.6 on Sun Solaris; filesystem type is UFS (not Veritas) with an 8K block size. Most of the Oracle data resides upon two RAID5 arrays; each array is composed of five 18GB disks. The upgrade consists of replacing the RAID5 18GB disks with 36GB disks.

The usual way we do this is to shutdown Oracle, take a full cold backup of the system, remove the 10 18GB disks, install and configure the new 10 36GB disks, and restore from backup.

The problem with this process is that we will be down for 8-10 hours or more.

The proposal is that we upgrade the disks one-at-a-time by making use of the automatic RAID5 rebuild mechanism. That is, each day we will simply remove one 18GB disk from a RAID5 array, replace it with a new 36GB disk, and let the automatic RAID5 rebuild mechanism kick in and restore the disk. Poor performance for a few hours, but no down time.

Has anyone done anything similar? I suspect this method will only reclaim 18GB of space on the new 36GB disk; can anyone verify that? If true, does anyone know of a way to increase that to the full 36GB on all disks without taking the machine out of service for any length of time?

Thanks to any responders.
Bill "I don't actually do this stuff, I just propose it" Becker Received on Thu Dec 21 2000 - 07:56:00 CST