Re: find physical blocks/disks, mapped from Oracle file# and block#, block corruption

From: lopera <prlopera_at_techie.com>
Date: Thu, 09 Jan 2003 00:57:08 +0100
Message-ID: <3E1CBAD4.5030405@techie.com>

true indeed I didnt provide enough infos

Oracle version ==> 8.1.7.4
O.S ==> Suse Linux Enterprise 7
Server ==> Compaq (I dont know the exact model) woth 4 CPUs and 14GB RAM Disks ==> the ones which came with Compaq box, 16 disks in RAID 0 + 1 FS ==> Ext3 (Linux FS)

I did run DBVERIFY (in all 4 corruptions faced this month) and to my surprise it says there is no error!

I am thinking of enabling db_block_checking but I am not 100% sure what it does, I am reading some notes from Metalink and looks quite confusing to me.... some says this parameter can determine if we are facing memory corruption and others talks about different function. What I understand is it checks data blocks read from data block buffer before write to disk, if it's not logically healthy Oracle signals ORA-01578. (And it make the instance crash?) Anyone have experience with DB_BLOCK_CHECKING?

Gregor Theis wrote:
> Hello
>
> "lopera" <prlopera_at_techie.com> schrieb im Newsbeitrag
> news:3E1C7C00.9090402_at_techie.com...
>

>>Hi
>>
>>
>>I have some block corruption issues with one of our databases, it
>>happened 4 times in last 1.5 months, I am not sure if the underlaying
>>disk is bad, actually I am not even sure which can be the bad disk since
>>we are in RAID 0+1. Does anyone know how to find which physical disk is
>>the datafile located providing I have the file# and block#? Something
>>like a mapping from Oracle to physical hardware. I would like to know if
>>all four corruptions occurs in same disk
>>
>>Also at Oracle level is there a way to tell weather we are facing
>>logical block corruption or physical (media) corruption?
>>
>>Thanks in advance
>>

>
>
> I think that we need a bit more data here. Which OS are you running
> (e.g. Solaris?)? What kind of Raid Software / Hardware are you running
> (e.g. Veritas Volume Manager or Sun SDS, EMC)? What kind of
> file system are you using (e.g. UFS, VFS).
>
> The block corruption can actually be caused by ORACLE itself or if you
> have bad luck (like I had a custumer site), you have bad EMC
> box firmware.
>
> With your block size, the block# and file# you can check which file
> and which part of the file you have problems with. Did you already
> run the oracle "dbv" utility on this file? To physically check the
> file you might want to "dd if=the_file of=/dev/null seek=to_the_bad_block
> count=the_full_db_block_in_fs_blocks". To calculate the parameters
> you will have to check the OS block size and READ the dd manual
> (if you are using a UNIX).
>
> To actually find the disk in a raid config you will have to find
> the blocks "logical" address on the virtual disk (you need the
> file system information for this) and try to translate this into
> the physical disk address (data from the RAID system needed for
> this). This is complicated stuff. Maybe you find some tools on
> the net that can do the calculation for you.
>
> Have a nice day
>
> Gregor
>
> --
> -----------------------------------------------------------------------
> Dipl.-Ing. Gregor J. Theis (3x ORACLE OCP DBA 8, 8i, 9i)
> mailto:gregor.theis@ib-theis.de http://www.ib-theis.de
>
>
Received on Wed Jan 08 2003 - 17:57:08 CST