Re: FTS, db file sequential read on undo

  • From: Tim Gorman <tim@xxxxxxxxx>
  • To: mathias.magnusson@xxxxxxxxx
  • Date: Mon, 13 Apr 2009 11:32:23 -0600

Not strange at all.  Consistent read images are constructed by retrieving data from the undo segments.  Your analyze and export operations are querying from a table and/or associated indexes which are being updated frequently during the lifetimes of those queries, so as the queries continue it becomes more and more frequently necessary to retrieve the data from the undo segments containing the point-in-time before-image when the query started.



Mathias Magnusson wrote:
Sorry. First reply and didn't watch how the "reply to" was set so it missed the list.

Vladimir,

That sounds very strange indeed. Is this a situation you see with any table or with just one? If you do a CTAS on dba_objects and export it, does it too suffer from this effect?

Have you just completed a large update to this table when you run this analyze/export or has the table not had any recent transactions when you run this? Do you suffer the same problem if you run your command twice in a row or is it just the first invocation that shows this effect?

Mathias

On Mon, Apr 13, 2009 at 5:36 PM, Vladimir Barac <vbarac@xxxxxxxxxxxx> wrote:

When accessing table for analyze or to export table, we are seeing "db file sequential read" wait event. V$session_wait file# is 2, that is – UNDO tablespace. Because of this analyze or export take forever to complete.

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

Other related posts: