Oracle RAC / Logical Data Guard causing network problems on VMware

Question

We have VMWare 5.0 cluster across the 12 blades (6 per chassis) running a mixture of Red Hat and Windows 2008 R2 vms. The Red Hat boxes are two times two node Oracle RAC (primary and secondary), also apache web servers and jboss application servers. The Windows servers are for AV/DC/Management/Monitoring.
The problem is that intermittent network connectivity to random Windows and Red Hat boxes occur when the Oracle RAC builds up archive logs and then ships / applies them to the secondary nodes, between ESX nodes either on different blades in the same chassis or across the chassis and even when all RAC nodes are on the same ESX host.
We are using NFS, Oracle 11g and Red Hat 6.2.
Sorry if this info is a bit vague, im not an Oracle expert! :-)
thanks,
Dave

Answer 1

Hi,
1.) The calculation for Standby RedoLogs is:
(Max Number of Logfiles per thread (Instance) +1) * Max Number of Threads (Instances))
So if you have 4 Redo Log Groups on your primary (which is 2 Redo Log Groups per Instance), then it ends up:
(2 +1) * 2 = 6
So actually you will only need 6 standby redo logs, not 8. But 2 more don't harm.
Your primary will need exactly the same number (6 or in your case 8). Which will be 3 per thread/instance or in your case 4.
2.) The SID List in the listener.ora is a listing of SIDs the Listener is listening on. It is not the listener name.
Hence it is not "lsnrctl guard_dgmgrl start" but only "lsnrctl LISTENER start", whereas the LISTENER is the default and "lsnrctl start" would be sufficient.
However since this is grid infrastructure with the listener running out of ASM home, be sure to have set your environment to GI Home not to DB_HOME for the listener.ora entries, but to DB_HOME for the tnsnames.ora entries necessary for data guard.
And since listener is running under clusterware you should use "srvctl stop listener" and start.
Last but not least the SID entries for dataguard have to use DGMGRL not dgmgrl.
3.) Here is the whitepaper you are looking for:
www.oracle.com/goto/maa
Also for client failover best practices.
(Here the direct link to the RAC whitepaper):
http://www.oracle.com/technetwork/database/features/availability/maa-wp-10g-racprimarysingleinstance-131970.pdf
However since this is 10g you should combine this with the 11g RAC standy paper (e.g. SCAN Listener setup).
Sebastian