RAC -- Optimizing RAC Reconfiguration Time

  • From: Sriram Kumar <k.sriramkumar@xxxxxxxxx>
  • To: oracle-l@xxxxxxxxxxxxx
  • Date: Mon, 21 Feb 2005 17:27:52 +0200

Hi Guru's

We have a 3 Node Oracle9i RAC (9.2.0.4) cluster running on HP-UX
Itanium2 platform. Often the third node is evicted from the cluster
with reason 1 ( Node Monitor) and at times with reason 3 (
Interconnect).. We had fixed the interconnect problem but the other
problem persists

When this situation happens the other two nodes freeze with the
following message in the alert log
"
Waiting for instances to leave: 
3 
"

After this either 1 or 2 happens

1) This continues for about 10-15 minutes and then the third node is
shutdown automatcally(evicted!!) and then we startup the third node
2) After some more waiting, we will have to intervene and do a
shutdown abort in the third node and then startup

After this it takes another couple of minutes before the cluster
reconfiguration to happen and then things are normal.

In effect, we loose about 20-25 minutes of business window. I have
raised couple of TAR's with Oracle and no conclusive answers from
oracle..They are asking us to upgrade to 9.2.0.5.

We have tried 9.2.0.5 upgrade in one of the other sites running a
similar RAC setup and had a major problem post upgrade ( a lot of
Ora-600's, a lot of coredumps and you nameit) and finally we had to
downgrade back to 9.2.0.4.

With the present configuration (9.2.0.4), is there any way to reduce
the cluster reconfiguration time to say 5 minutes or less?..

I have tried metalink but in vain. Any pointers would be very much appriciated

Cheers

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

Other related posts: