Re: Issues with Triggers in an Oracle RAC

  • From: "Rumpi Gravenstein" <rgravens@xxxxxxxxx>
  • To: "Hemant K Chitale" <hkchital@xxxxxxxxxxxxxx>
  • Date: Wed, 9 Apr 2008 13:10:02 -0400

Thanks, that's good to know.

I should be clear in that the audit trigger I'm considering is not the
Oracle provided AUDIT SESSION.  Rather it is home grown pre-insert
and pre-update triggers on our application tables that sets the value of two
columns in a row that is being inserted or updated.  Not a lot of stuff
going on there.  We have a 5 node RAC.

On 4/9/08, Hemant K Chitale <hkchital@xxxxxxxxxxxxxx> wrote:
> I DID have problems with AUDIT SESSION that was enabled for a short while
> in an RAC instance.
> Although the AUDSES$ sequence had been altered to a high CACHE value,
>  there was some severe
> contention during a connection storm (connections coming in at about 1
> every 3 seconds) and one
> instance did crash.  We disabled Session Audit then.
> In my opinion, SYS.AUD$ tries to capture too much information.
> Hemant K Chitale
> At 11:19 PM Wednesday, Rumpi Gravenstein wrote:
> > I would like to add audit triggers to populate record create user and
> > timestamp columns ( :new.rcd_create_user_id := USER and :new.rcd_create_dt
> > := SYS_TIMESTAMP).  They have an unwritten policy that states triggers
> > should be avoided if at all possible.  I'm trying to find factual support
> > that their contention that triggers cause "library cache lock waits"  which
> > in turn impact the database resulting in unacceptable production problems is
> > misplaced, especially for the type of audit triggers I'm considering.
> >

Rumpi Gravenstein

Other related posts: