Oracle-L: Re: Snapshot too old in undo tablespace in 9i?

From: <Stephen.HODGKINSON_at_eaguk.com>
Date: Fri, 15 Aug 2003 06:49:22 -0800
Message-ID: <F001.005CAEB1.20030815064922@fatcity.com>

Daniel, that make sense , thanks.

Stephen Hodgkinson
Oracle DBA
Total Gas & Power Ltd

Phone: 01737 27 5564
stephen.hodgkinson_at_eaguk.com

                                                                                                                                              
                    Daniel Fink                                                                                                               
                    <daniel.fink_at_s       To:     Multiple recipients of list ORACLE-L <ORACLE-L_at_fatcity.com>                                  
                    un.com>              cc:                                                                                                  
                    Sent by:             Subject:     Re: Snapshot too old in undo tablespace in 9i?                                          
                    ml-errors_at_fatc                                                                                                            
                    ity.com                                                                                                                   
                                                                                                                                              
                                                                                                                                              
                    15/08/03 15:26                                                                                                            
                    Please respond                                                                                                            
                    to ORACLE-L

Stephen,

Tim's statement is correct, but can be construed incorrectly if you read it and think of TEMP segments. AUM still uses undo segments (same basic structure as rollback segments). However, one of the space management steps is to allow an undo segment to 'steal' extents from another undo segment. This means that all extents (other than those currently in use or extent 0 (and perhaps 1)) are available to any other segment should it require them.

A single transaction cannot start in undo segment #1, allocate space in it and then move to undo segment #2. However, the transaction can cause undo segment #1 to allocate space currently allocated to #2.

I hope this clears up the disparity between the statements.

Daniel

Stephen.HODGKINSON_at_eaguk.com wrote:
>
> Daniel,
>
> I have just finished reading your document on UNdo Internals and Tims
> "Cats, Dogs and ORA-1555s".
>
> Thanks for the documents they were both great.
>
> There is something I don't understand and I am not sure about it.
>
> You have said below:
>
> "When a transaction is bound to an undo segment, it allocates a slot in
the
> tx table."
>
> I thought that transactions were no longer bound to UNDO segments and
> this was one of the improvements in 9i.
>
> I have pasted an extract from Tims document:
>
> Into the future: Oracle9i UNDO tablespaces.
> As you may have observed, one of the reasons space management for
rollback
> segments is so difficult is due
> to the fact that a transaction is assigned irrevocably to a single
rollback
> segment.
> Each rollback segment can only handle a finite number of transactions
> (due to block-level contention for the transaction table in the header
> block),
> so there must be multiple rollback segments to handle potentially large
> numbers of transactions.
> UNDO tablespaces in Oracle9i allow an entire tablespace to become a
single,
> large pool of undo blocks for use by any and all transactions.
> Instead of having available space carved up into many smaller rollback
> segments,
> a single transaction can utilize all of the space in the UNDO tablespace,
> if necessary. Many, many transactions can share that space also,
> because the controlling transaction table is no longer contained in a
> single database block,
> avoiding contention for this important resource.
>
> I guess I am jumping to the wrong assumption in Tims extract - can you
> clarify it for me.
>
> thanks, stephen
>
> Phone: 01737 27 5564
> stephen.hodgkinson_at_eaguk.com

This insert confirms that this email message and all associated attachments have been swept by TotalFinaElf using Network Associates VirusScan for the presence of computer
viruses during the receipt of this message.

(See attached file: daniel.fink.vcf)

This email contains information which is confidential. It is for the exclusive use of the addressee. If you are not an addressee please contact us immediately on 01737 275500. Please note that any distribution, copying or use of this communication or its contents is prohibited.

This footnote also confirms that this email message and all associated attachments have been swept by Network Associates VirusScan for the presence of computer viruses.

-- 
Please see the official ORACLE-L FAQ: http://www.orafaq.net
-- 
Author: 
  INET: Stephen.HODGKINSON_at_eaguk.com

Fat City Network Services    -- 858-538-5051 http://www.fatcity.com
San Diego, California        -- Mailing list and web hosting services
---------------------------------------------------------------------
To REMOVE yourself from this mailing list, send an E-Mail message
to: ListGuru_at_fatcity.com (note EXACT spelling of 'ListGuru') and in
the message BODY, include a line containing: UNSUB ORACLE-L
(or the name of mailing list you want to be removed from).  You may
also send the HELP command for other information (like subscribing).



application/octet-stream attachment: daniel.fink.vcf

Received on Fri Aug 15 2003 - 09:49:22 CDT