Oracle corruption after RAID 5 recovery

[Declan Hogan]

Hi, all,
I have a server which has a NetServer Storage Rack/12 which contains 1 Hot Swap Drive and 2 RAID-5 Disk Arrays (1x8 and 1x3) attached to an NT Server (SP6a).
Last weekend we had a Drive failure and when the Drive was replaced and the RAID recovery was done I tried to open the Oracle DB (spread across both arrays) I got a message to say one of the Tablespace files was corrupted.
As a result I had to restore the DB and do a recovery, which went fine and all data was recovered.
The same thing happened before when I had added the 4 drives to fill the Array early last year. The rack is connected to a NetRAID 3Si Card.

I have done a similar drive replacement on 2 other machines both of which have Oracle DBs residing on the RAID5 array and they both recovered without any problems (which is what I expected in all cases).

Has anyone else had any similar experiences of this.

Thanks in advance.
Declan Hogan.

[Martin Johnson]

Are you doing asynch I/O? With asynch I/O that application does not wait for confirmation of the I/O request. When hardware problems occur, there is an increase in risk of data corruption.

HTH
Marty

[Declan Hogan]

Hi, Marty,

Thanks for your reply. I have checked the Write Policy on the Server which had the failure, it is set to Write Back. The other two servers (where no failure occured) have a Write Policy of Write Thru. All three have a read policy of Adaptive Read Ahead.

When you say async I/O does this "translate" to Write Back policy.

Regards,
Declan Hogan.