Re: Dealock on Rollback Segment (I think)

After carefully reading the other responses to this posting, I thought that I would weigh in on a slightly different take on the subject.

By splitting the update process into 8 separate processes, the consultant just forced this update process to be run in parallel. Since you are using Oracle 8i, why not just use parallel DML to perform the same operation? In the end, you might be better served since Oracle is handling the parallel operations for you. This means that you won't be getting a deadlock situation.

Just my 3.14159265 cents worth,
Brian  

buckeye714_at_my-deja.com wrote:
>
> We have oracle 8.1.6 on Solaris 2.6 and are implementing Oracle
> Financial Services Application. One process in this app updates every
> row in our customer table, currently four million rows. In an effort to
> speed this up, an Oracle consultant split the update process into 8
> separate processes. It now does the updates based on rowid, where
> process 1 does the first 500,000 rowid's, process 2 does the next
> 500,000 rowid's, and so on. These 8 process are run simultaneously.
>
> The process ran fine on the test box, but we get deadlocks on the
> 'production' box. Using OEM, it appears that the deadlock is on a
> rollback segment. Since the application is updating rows based on
> specific rowid's, I don't believe the deadlock is a result of two
> process
> updating the same set of rows. Also, I am positive that the
> sid's involved in the deadlock belong to the 8 processes being run.
>
> Does any one have an idea on how to fix this problem? Thanks in advance
> for your help.
>
> Patrick.
>
> Sent via Deja.com
> http://www.deja.com/
 

-- 
========================================
Brian Peasland
Raytheons Systems at
  USGS EROS Data Center
These opinions are my own and do not
necessarily reflect the opinions of my 
company!
========================================