Unsolved
This post is more than 5 years old
3 Posts
0
1866
Oracle RAC problem
Customer have deployed NMO in RAC environment, and have half of his savesets done with client physical_node_1 and other half physical_node_2. The problem is rectified with modified RMAN script (send client environmental attribute while opening channels), but now the question is how to restore? Should we open one chanel for physical_node_1 and one for physical_node_2 ?
axavier1
76 Posts
0
April 14th, 2010 10:00
Hi,
Tell to customer configure only one client resource to RAC vip name ou cluster name, in RAC you don´t need to backup both instances just one.
Regards,
Anderson Xavier
BGuevara
101 Posts
0
June 1st, 2011 06:00
Hi Anderson,
Do you have that configuration up and running without issues?, how many licenses is consuming?
Thanks and regards.
ble1
14.3K Posts
0
June 1st, 2011 09:00
I use RAC (both classic one and geo RAC) with 4 and 6 nodes and it is not necessary to allocated channels towards different nodes if you don't want to.
BGuevara
101 Posts
0
June 1st, 2011 09:00
Hi,
According to NMO 5.0 documentation in a RAC system there should be one instance for each phisycal node in NetWorker:
'On a RAC system, traditional cluster failover is not available. If an instance or node
fails in RAC, another node detects the failure and recovers the failed node’s data. As a
result, the nodes in RAC carry on without the failed node. If a system failure occurs
on the RAC node used to initiate an NMO backup, the backup fails. In this case,
manual intervention is required to configure and restart the backup on a different
RAC node that is available:'
Regards.
BGuevara
101 Posts
0
June 2nd, 2011 00:00
Hi Hrvoje,
So if I understand correctly, you agree with Anderson that it can be used just one client instance pointing to VIP? And that when one node goes down backup still works with no user intervention since client instance is pointing to VIP adress?
Regards.
ble1
14.3K Posts
0
June 2nd, 2011 13:00
I do not use VIP address due to network model we have. Instead I used fixed host. If that host would go down I would just switch (manually) to another host. We plan to change this by the end of the year by using VIP and service names instead, but we might not do that at the end. Our biggest problem (backup wise) is that we use concept of oracle hotels where single RAC setup may host 70+ databases. Needless to say, index for such setup, if only VIP would be used, would be too big and would have impact for restores. This is why I prefer to have those spread fixedly across the nodes.
BGuevara
101 Posts
0
June 3rd, 2011 02:00
Your case sound interesting. But my concern is that probably configuring a client instance pointing to the VIP adress is going to work but i don’t know if EMC Support that configuration since in the documentation specify that phisycal hosts has to be used. Can any EMC member confirm this to us??
Thanks and regards.
De: Hrvoje Crvelin
Enviado el: jueves, 02 de junio de 2011 22:32
Para: Bruno Guevara Moya
Asunto: New message: "Oracle RAC problem" - Sender is probably forged (SPF Softfail)
EMC Community Network
Oracle RAC problem
reply from Hrvoje Crvelin in NetWorker - View the full discussion
ble1
14.3K Posts
0
June 3rd, 2011 02:00
It is RMAN which does backup. I had long talk about possibilities with Oracle DBAs and we did quite few of the tests and went for what I have described. NW uses RMAN so at the end it all comes down to what Oracle guys would prefer.