Re: Fasten huge index rebuild - How To ?

From: Martic Zoran <zoran_martic_at_yahoo.com>
Date: Thu, 19 May 2005 07:18:40 -0700 (PDT)
Message-ID: <20050519141840.87839.qmail@web52609.mail.yahoo.com>

Hi Prem.,

I was thinking more about the events that/these sessions are experiencing while creating the index on the test system.

For example, when they are testing the creation of the index on the test system if possible to get the events timing for this/these sessions.

I suspect a few things: db file sequential read, direct path reads/writes and log file sync. also, it can be crazy CPU usage while creating the domain index.
Usually when something is slow it looks as disk I/O related thing, but CLOB/XMLType and domain indexes may be CPU related problems.

You need to look what the session are doing that long while creating the index on the test system.

Also it will be helpful to put SQL script for the index creation, the number of CPU's on the test/production box, ....

Regards,
Zoran

• Prem Khanna J <premjhere_at_gmail.com> wrote:
  > Hi Niall/Zoran , this was the reply i got from him.
  >
  

  ---

  
  > It maybe hard to tell because when our database is
  > running daily jobs,
  > the majority wait events are:
  > db file sequential read , direct path read
  > PX related events , i/o slave wait , db file
  > scattered read
  > SQL*Net message to/from dblink
  >
  > When we re-create the index, all other jobs will not
  > be running. We
  > can still assume the same major events as in running
  > daily jobs,
  > except the "SQL*Net message to/from dblink"
  >
  

  ---

  
  > thanx guys for all ur help.
  >
  > Regards,
  > Prem.
  >
  

  ---

Yahoo! Mail Mobile
Take Yahoo! Mail with you! Check email on your mobile phone. http://mobile.yahoo.com/learn/mail

--
http://www.freelists.org/webpage/oracle-l

Received on Thu May 19 2005 - 10:28:51 CDT