RE: ** commit or rollback - diff

From: Yavor Ivanov <>
Date: Wed, 18 Feb 2009 14:27:08 +0200
Message-ID: <BD17E2E69E17C64A9684C940EB580E030111A7AECE1F_at_stemodc1.stemo.local>

Two points here:
  • as you have noticed, even when you issue only SELECT statements across a database link, there is a transaction lock is placed on an undo segments. IMHO, when you have no DML, there should be no difference between COMMIT or ROLLBACK - either will just rerelease the lock.
  • generally when there is some real DML, COMMIT is much cheaper than ROLLBACK (in Oracle). COMMIT just flushes a little redo and that's all. ROLLBACK does much more operations.

Yavor Ivanov

From: [] On Behalf Of A Joshi Sent: Thursday, February 12, 2009 11:47 PM To:; Subject: RE: ** commit or rollback - diff


   Thanks. Yes, I see that from safety point of view. From performance point of view and resource consumption : which is faster? Or does it make no diff? I know commit is expensive operation : however : is that only if there are changes. Thanks

  • On Thu, 2/12/09, Goulet, Richard <> wrote: From: Goulet, Richard <> Subject: RE: ** commit or rollback - diff To:, Date: Thursday, February 12, 2009, 4:38 PM Rollback is safer just incase you did a DML transaction without knowing it like inside a procedure.

Dick Goulet

From: [] On Behalf Of A Joshi Sent: Thursday, February 12, 2009 4:32 PM To:
Subject: ** commit or rollback - diff

    If I have not done a dml transaction in a session : no update, delete or insert etc. I have only done select and some of the objects can be over a db link. So I can do a commit or rollback so that no transaction is pending in my session. My question is : is there any difference in such case between the behaviour of commit and rollback. When there is no data as such to commit or rollback. I am thinking it is better to do rollback since it has to do less. Am I wrong. Any observation. Thanks for help. Thanks

Received on Wed Feb 18 2009 - 06:27:08 CST

Original text of this message