General
cancel
Showing results for 
Search instead for 
Did you mean: 

Block corruption on Oracle data file

 
Highlighted
Honored Contributor

Block corruption on Oracle data file

here is one of the update from Oracle about the oracle data file corruption.(data file on a unix file system;not a raw file).The file system is on a SAN disk. The update is telling the data block got overwritten.

*******************
I reviewed the data in the two corrupted blocks. Both blocks were corrupted in teh same way. This is, blocks were overwritten with exactly the
same data at the same offsets.

For block 73 we have:
------------------------------
$ diff 73_bad.hex 73_ok.hex
130,131c130,131
< 0004020 000b 0000 0c00 0000 0100 0100 0b00 000c
< 0004040 0000 0001 0002 041d 1825 3002 c10d 0180
---
> 0004020 0202 c108 ffff ffff ff02 c102 02c1 15ff
> 0004040 02c1 0207 7864 041d 1825 3002 c10d 0180

and for block 272:
------------------------

$ diff 272_bad.hex 272_fix.hex
130,131c130,131
< 0004020 000b 0000 0c00 0000 0100 0100 0b00 000c
< 0004040 0000 0001 0002 0000 0000 0000 0000 0000
---
> 0004020 0000 0000 0000 0000 0000 0000 0000 0000
> 0004040 0000 0000 0000 0000 0000 0000 0000 0000

(*.hex files have the output from od -vx *.dd command)

We can see that in both cases, block were overwritten with the following 22 byte pattern:

000b 0000 0c00 0000 0100 0100 0b00 000c 0000 0001 0002

starting at offset 2064 (4020 octal)
===

Any idea what is this 22 bit pattern means?what could cause a data corruption(either a h/w issue or some one messing up with that file accidentally).

Corrupted blocks have been fixed and the databse is currently up now.


Could some clarify in a better way.
We have not done a fsck on this FS(2TB) as it require to unmount and we want to avoid it at this moment.
5 REPLIES 5
Highlighted
Honored Contributor

Re: Block corruption on Oracle data file

how would you handle if you have a similar scenario?any comments?
Highlighted
Acclaimed Contributor

Re: Block corruption on Oracle data file

If this were me, I would search the HP-UX patch database for "data corruption". I would look for and install the latest LVM, VxVM, and VxFS patches and read the release notes and search for anything that might suggest possible data corruption. Since you haven't bother to list your OS version, platform, or Oracle version, I'm certainly not going to bother looking for patches.

Now having said all of this, while it is possible that this is an OS problem, it is far more probable that this is an application error (Oracle). I would be spending most of my time at MetaLink. It almost certainly is not a file system corruption since you are able to extract the
data from a cooked file using dd. This is an extremely strong indication that the file system internal integrity is intact.
If it ain't broke, I can fix that.
Highlighted
Honored Contributor

Re: Block corruption on Oracle data file

I have the almost recent patches.The server was already pathed.

9000/800/S16K-A;B.11.11 U; 10.2.0.2.R;
Not yet done any dd/fsck till now.
Highlighted
Honored Contributor

Re: Block corruption on Oracle data file

I have the almost recent patches.The server was already pathed.

9000/800/S16K-A;B.11.11; Oracle 10.2.0.2.R;
Not yet done any dd/fsck till now.
Highlighted
Honored Contributor

Re: Block corruption on Oracle data file

closig with the current informations. Thanks to all.