| Oracle FAQ | Your Portal to the Oracle Knowledge Grid | |
 |
 | |||
| Oracle FAQ | Your Portal to the Oracle Knowledge Grid | |
|
| |||
Home -> Community -> Mailing Lists -> Oracle-L -> Re: Split Blocks on Instance Crash
Hi Mark,
Yes I am thinking about the O/S block writes. If I understand you correctly, you're saying that since DBWR schedules these as packaged O/S writes they are likely to either succeed or fail in their entirety and unlikely to lead to split blocks - although that could happen under some circumstances.
So very occassionally you may indeed see a server requiring media recovery following a crash.
The difference with backups is that a crash is fairly instantaneous so you've only got a tiny window in which the split block will possibly occur - so it's rare. During backups the file copying/writing mechanism runs scanning files for a long time while DBWR is potentially busy - so there's a much bigger window for it to pick up a split block - hence it's not so rare and the need for BEGIN/END BACKUP to accommodate this.
Have I got it now?!
Thanks for your help
Charlotte
> I think the questioner is asking about the 512 byte
components of the write
>of the Oracle block.
>
>In a non-RMAN backup, the operating system utilities
have no particular
>interest by default in copying chunks (n of the
usually 512 byte pieces) in
>anything like an
>alignment matching Oracle's blocks. DBWR, on the
other hand, will definitely
>submit Oracle blocks as integral sets of the pieces
that make up the Oracle
>blocks from the underlying OS pieces.
>
>So the risk goes way, way down. I'd quibble slightly
with Tanel's "always"
>remark. If you restart and it tells you media
recovery is required, then a
>file really did get crashed and one of the cases of
>a file getting trashed is that it is not marked fuzzy
and a block's header
>and tail don't match. Then you would have to get a
previous backup of the
>file (or block) and roll it forward. So the recovery
is just the same as if
>you lose a file for any other reason. When I've
gotten a file trashed
>(...never on a client of mine with at least duplex
plex images, but
>frequently in the dim past when lots of folks used
single plex images) I
>never saw a block that looked complete except for a
mismatch, so I'm
>thinking this is very rare.
>
>Regards,
>
>mwf
-- http://www.freelists.org/webpage/oracle-lReceived on Thu Dec 22 2005 - 19:11:52 CST
 |
 |