Re: Oracle database corruption on HP-UX & Andataco(Micropolis) disk.

dgv_at_teleport.com (Don Vinton) writes:
>
>>Random data corruption errors with ANDATACO 2 or 3 GB SCSI disc drives
 

>>Copying the data at the HP-UX file level (e.g. cpio -p) will copy the random
>>errors and will then make them a permanent part of the receiving file.
>>SCSI terminators, cable lengths, addresses have all been checked.
>
>Here what the error says.
>
>>ORA-01578: ORACLE data block corrupted (file # 22, block # 372)
>>ORA-00600: internal error code, arguments: [3339], [1476395384], [1476395380]

We had exactly the same kind of error about 1 1/2 years ago on a Dec 5900 running Ultrix. The problem turned out to be a frayed connection on the SCSI cable. We replaced it and never had another problem.

I cant tell you that this is exactly your problem. It could be a connector on the I/O cabinet, could be the internal ribbon cable or it could be a ton of Things I dont even know exist in a I/O subsystem.

I can tell you that as a Oracle DBA dont let them take try and give you any lines about needing to replace the disk drives. You already know what I wish I knew when I had the problem. "The data on the drive is good and it is the xfer that is bad"

I didn't have any permanent data problems until I let a hardware tech convince me I should get my data off that drive. I wish I had gone with my first instinct instead. - "Replace the entire Subsystem. Cabinets and all but leave my drives the hell alone."

-- 
Regards, 

Lee E. Parsons                  		
Systems Oracle DBA	 			lparsons_at_world.std.com