Path: dp-news.maxwell.syr.edu!spool.maxwell.syr.edu!drn.maxwell.syr.edu!news.maxwell.syr.edu!news.glorb.com!transit.nntp.hccnet.nl!transit1.nntp.hccnet.nl!reader.nntp.hccnet.nl!not-for-mail From: Sybrand Bakker Newsgroups: comp.databases.oracle.server Subject: Re: rollback space Date: Wed, 02 Feb 2005 22:35:32 +0100 Reply-To: sybrandb@hccnet.verwijderdit.nl Message-ID: References: <1107294990.249825.185300@c13g2000cwb.googlegroups.com> <1z4Md.12981$qJ3.7230@bignews1.bellsouth.net> <1107369583.318529.264990@z14g2000cwz.googlegroups.com> X-Newsreader: Forte Agent 1.93/32.576 English (American) MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Lines: 29 Organization: HCCnet Nieuwsgebruiker NNTP-Posting-Date: 02 Feb 2005 21:35:31 GMT NNTP-Posting-Host: 62.251.33.219 X-Trace: 1107380131 reader10.nntp.hccnet.nl 769 62.251.33.219:4283 X-Complaints-To: abuse@hccnet.nl Xref: dp-news.maxwell.syr.edu comp.databases.oracle.server:234843 On 2 Feb 2005 10:39:43 -0800, manish.nevrekar@gmail.com wrote: >1) Each row has 10 columns. Approximately 70 B each row. >2) No lobs >3)No other transactions are running, this is a batch load, any other >transaction if running at all is select. >4) there are no on isert triggers on the table. > >I beg to differ on commiting after every 'x' rows. I always prefer to >commit once at the end of the transaction. Commiting often is >definitely more work plus there is always the chance of the dreaded >ORA-01555 "Snapshot too old" error. >In my ETL procedures every insert and update is time stamped and hence >it can be identified. >Thanks >Manish Seems like a case for logminer. Be aware though, as soon as a block needs to be reorganized -and as far as I know rows are always stored adjacent- pretty much the whole block could need to be rewritten. Logminer will identify the actual insert including the redo size, and any 'internal' operations associated with the SCN, usually index manipulations. -- Sybrand Bakker, Senior Oracle DBA