Hi,

We are planning to implement RHEL Cluster for Oracle 11g. In past, we have installed Oracle on stand alone. Also, We have used AIX HACMP/Windows clusters. But RHEL Cluster looks a bit confusing. So Consider me a newbie, I have some queries to start up.

1) RHEL Cluster requires any extra License?
2) Which File system ext3, ext4, xfs or gfs2?
3) Which is Device Fencing?
4) I have seen multiple tutorials, none mention VIP, HeartBeat, Qurrum Disk. What are alternative names for them?

Thanks for reading the question.

Clustering for Oracle DB is done using Oracle RAC/GRID which does its own clustering setup so you don't need RHEL clustering. This requires use of Oracle ASM as well so doesn't use a filesystem as ASM utilizes devices directly. In the past you could use OCFS (Oracle Cluster Filesystem) but they've since recommended using only ASM instead. You'll need to read up on the Oracle documentation for this.

You didn't mention which version of RHEL you're using.

1 members found this post helpful.

Thank for the reply.

Oracle RAC/Grid is paid software, you need extra license with oracle database. Also, Oracle RAC ensures Active/Active cluster which is currently not our requirement. We are looking for basic recover-ability on Linux

Just curious, are you talking about Oracle Cluster-ware or RAC? I believe OCFS2 under OEL, with Oracle Clusterware is a free to used (Still checking with Oracle to come Back on this query).

I was thinking about building cluster over RHEL 6.7.

Thanks again for the reply. It really helped to do some more research before actual implementation.

OEL is derived from RHEL although they do some differences from RHEL including kernel customizations - which are open source so can be applied to RHEL but we don't use those on our RAC/GRID. I don't know that there are different licensing requirements for RAC/GRID than for active/passive. I think Oracle takes the position you have to pay for a license on each server where Oracle is installed but I'll admit I'm not directly involved in licensing.

One thing I've done in the past (on Solaris, not Linux) was a log shipping setup. In that we had Oracle live as Production database on one server then had a "standby" Oracle copy on another server. The DB was up on the other server but not accessible. We did full archiving on Production and transferred all archive logs to the standby. On the standby we'd apply logs on a 6 hours delays (to give us time to stop propagation of bad things like accidental table drops). If the Production server went down we'd move the IP to the Standby server and go ahead and apply any pending logs (i.e. not wait the 6 hours) then bring up that Standby as the Production database.

On HP-UX we did do an Active/Passive cluster setup years ago for Oracle wherein Oracle binaries were installed separately on the two cluster nodes but the Production instance was only running live on one node on SAN disks. If the Production node went down we'd bring up the SAN disks with the actual database on the failover node and start Oracle there. I guess that is still possible but we haven't done it on Linux.

1 members found this post helpful.