05-07-2015 02:44 AM
Hello Experts,
I have an issue with oracle database backups which is in active active cluster. The details below
Nodename : Node1 Nodename : Node2
Oracle cluster name : Orclus1 Oracle cluster name : Oraclus2
databases : oradb1 & oradb2 databases : oradb3 & oradb4
Node1 & Node2 are in active active cluster and oracle cluster names are Oraclus1 & Oraclus2
At present Node1 have Oraclus1 and Node2 have Oraclus2. Incase of one node failure both oracle clusters would be in single node.
I have created two policy for each oracle cluster and keep the client name as oracle cluster name. When i start the backup first it failed because backup was triggeting with nodename. so i changed client_name in the bp.conf file as the oracle cluster name. After this my backup completed successfully. I have done this on bothe the nodes. So in Node1 bp.conf file CLient_Name : Oraclus1 and in Node2 bp.conf file Client_Name : Oraclus2
But when Node2 down and Oraclus2 moved to Node1, Node1 have both Oraclus1 & Oraclus2. In this situation when backup triggered for Oraclus2 policy it would fail because in the Node1 bp.conf file have Client_Name is Oraclus1.
How to overcome this ?
Solved! Go to Solution.
05-07-2015 03:08 AM
Best if dba hardcodes the NB_ORA_CLIENT in the rman script to the Oracle cluster name (Oraclus1 or Oraclus2) and leave CLIENT_NAME in bp.conf to be the local nodename.
This has always worked well for our customers with Oracle clusters.
Remember that restores should be treated as redirected restores and proper altnames entries are needed on the master server.
05-07-2015 03:08 AM
Best if dba hardcodes the NB_ORA_CLIENT in the rman script to the Oracle cluster name (Oraclus1 or Oraclus2) and leave CLIENT_NAME in bp.conf to be the local nodename.
This has always worked well for our customers with Oracle clusters.
Remember that restores should be treated as redirected restores and proper altnames entries are needed on the master server.
05-12-2015 10:35 AM
Thanks Marianne, i have done the same and working fine now.
05-13-2015 03:05 AM
But that isnt the correct way to do it.
That would be to have a service address as part of the cluster resource group so the storage, the ora server, the listener and the ip address are in the same group. Then if you fail, the rg moves , the ip address moves. So regardless of which node the rg runs on , the ip address / service address is always the same and you always back it up with that address. Jim.
05-13-2015 04:03 AM
Sorry but it is the correct way for Oracle databases on clusters, and has been since Netbackup 3.4.
And as I understand the setup descibed the OraClus1 and OraClus2 is the service adresse for the two Oracle Cluster Groups.