Oracle FAQ Your Portal to the Oracle Knowledge Grid

Home -> Community -> Mailing Lists -> Oracle-L -> Re: Data Mirroring on two data centers -- How to use ASM ?

Re: Data Mirroring on two data centers -- How to use ASM ?

From: Tanel Põder <>
Date: Sun, 21 May 2006 20:39:45 +0800
Message-ID: <0b9a01c67cd3$a57f60d0$6401a8c0@porgand>

I wondered if anyone brings the HARD up ;)

I haven't really seen nor heard anybody using it. HARD basically means that storage subsystems know where the checksum bytes and relative data block address bytes reside within a block plus few extra trivial checks. However the situation isn't as bright as Oracle's doc states.

There are several limitations and fundamental problems, like what to do with detected blocks which checksum bytes don't match the content of the block. One option is to write the block to disk and alert storage administrator/DBA for post-diagnostics. This means that the corruption will get replicated to DR site - and that's a problem.

Other option would be to reject the corrupt write, causing Oracle to retry it. If we are lucky and the error is transient, the processing continues, but if not, then Oracle has to take the datafile offline. And if the datafile happens to be in system or undo tablespace, your services (and instances) are pretty much dead.

EMC offers a discard option as well, that the corrupt block write are just silently thrown away, but that's the last thing we want to happen..

Another issue is that there must be no abstraction layers such filesystems between storage and Oracle. You have to use RAW devices and have to make sure that the blocks written to storage aren't moved around in higher layers as the correct block address checking relies on comparing the physical address on raw volume to the RDBA stored within block header.

Read more:

Also, it seems that the interest in HARD initiative has slowly died, the last certified DB version is 9.2:

Kevin probably knows stories from the field, if anyone actually has used successfully... Kevin, wake up ;)


To solve this kind of problems, oracle has created Oracle Hardware Assisted Resilient Data (HARD) Initiative, and according to the all large storage providers are members of this initiative, so „smart" storage does know about Oracle block format and can catch block corruptions, before they are written to disk.

On 5/21/06, Tanel Põder <> wrote:
> Btw, one more thing if talking about redolog level replication versus
> storage replication is that as storage doesn't normally know anything
> about Oracle's block formats then it doesn't have any choice but replicate
> all corruptions to remote site as well. Oracle's redo apply mechanism
> would detect most of such issues...
> Tanel.

Best regards,
  Edgar Chupit

Received on Sun May 21 2006 - 07:39:45 CDT

Original text of this message