Comments on: TNS Listener Configuration for Oracle RAC

By: Andrew Kerber

Andrew Kerber — Mon, 18 Jul 2011 14:36:00 +0000

Admittedly this is rather late, but you should be sure and use the fqdn when using a name rather than ip for the remote and local listeners. There is an oracle bug referencing this. What it amounts to is that if you dont use the fqdn for the local listener, when you connect to the database from a different default domain, the oracle client will have trouble connecting if the local listener setting doesnt include an fqdn.

By: Dan Norris

Dan Norris — Fri, 04 Mar 2011 21:57:00 +0000

For 11gR2, all the stuff above is automatically configured when the DB is started. No manual configuration needed for any of it. However, some of the status commands are helpful for troubleshooting and debugging.

There’s exactly one SCAN allowed, so if your database server has more than one interface used for client connections, one can use the SCAN, the other will have to use the pre-11.2 method that lists all the VIPs (which you’ll have/want to create too).

I have no experience with the cold failover, but if you use RAC One, it is mostly automated.

By: Steve

Steve — Fri, 04 Mar 2011 21:31:00 +0000

Hi Dan, Nice post!

Have couple of questions in reading your post. Most of our implementations at customer sites are done with DB in cold failover mode where we have DB vlan which only App nodes can talk to and OAM vlan or control vlan where back office application can only talk to. In these cases, we generally have the listener endpoints to listen on both DB VIP and OAM VIP and they fail over to other node in the cluster, if primary node fails for some reason. All works great. However, if we had to support RAC with similar support i.e having back office apps connect to DB via control vlan only, what would you recommended setup?

And how does this work with SCAN listeners on RAC?

Cold failover protected by 11gr2 Oracle Clusterware is even worse cause during failover, we fail just the DB and not SCANs or restart local listeners. In a typical cold failover solution, we failover db,listener and storage resources along with it, but with clusterware, we have asm clustered but DB running in cold failover mode.

Appreciate your feedback.

By: Dan Norris

Dan Norris — Wed, 13 Oct 2010 17:38:00 +0000

Let’s keep the Q&A on listener issues here for all to benefit.

Shoot me a mail on whatever else you need – my schedule is tight for the
next 2 weeks, so a call is probably less likely to get answered than an
email.

By: Jake Ramirez

Jake Ramirez — Wed, 13 Oct 2010 17:16:00 +0000

Hey Dan. This is Jake Ramirez in Vancouver WA. Wanted to know if I could call you about Cluster Listener config in 11gR2. Also need to talk to you about one more thing

By: johannabartley

johannabartley — Tue, 20 Jul 2010 14:28:59 +0000

Thank you for the code. It saved me a lot of time in front of the computer. Johanna

By: Roberto Pinel

Roberto Pinel — Fri, 24 Apr 2009 00:11:11 +0000

Hi, Good article, I have Oracle 11.1.0.7 with Rhel 5, 3 nodes, but I have a problem, I got the tns-12520 3 times a day, no one cant connect and I have to restar the database, The parameter process is 800 and I have the local_listener with the same address from tnsnames.ora in each server, the workaround that I use was to change the local_listener valur for the remote_listener value, and work fine, no more connection problems, but I know that is not correct, have you seen this problem before.

Thanks

By: Roberto Pinel

Roberto Pinel — Thu, 23 Apr 2009 17:11:11 +0000

Thanks

By: Dan Norris

Dan Norris — Sat, 11 Apr 2009 08:53:13 +0000

A few ideas come to mind. To answer your question, you could use a tns
entry that references all instances/servers to check overall
availability and also separately check each individual instance to
determine availability.

You might also consider using clusterware callouts to alert you.
Scripts placed in ora_crs_home/racg/usrco will be run whenever
clusterware detects a resource offline. That may be easier than
writing your own custom polling method.

Please excuse my typos, I sent this from my iPhone.

By: vnkt

vnkt — Sat, 11 Apr 2009 00:59:17 +0000

Hi I have a situation where i need to develop a monitoring solution for checking the availability of oracle database instances on a rac system.
I need to alert not only when the overall rac system is unavailable but also when one of the instances on a node is not available while overall the rac system is available(thro other isntances).
I already have developed a monitoring solution for single instance oracle database where the script connects to an oracle instance and if unable to then raises alert. Can i have a similar approach for rac — where in i can specify the overall sid of the rac to check the overall avaialability and individual sid of each instance to check the individual availability ?