This content has been marked as final. Show 5 replies
first of all Red Hat Cluster Suite cannot replace the RAC functionality. RAC is database Red Hat Cluster Suite, just provides clusterware. So all with what you will end up is a plain old failover cluster solution.
So in maximum Red Hat Cluster Suite can replace the Oracle Clusterware. (But not below a RAC, since this is not supported and even with integration with other clusterware Oracle Clusterware will still be there).
So what he will end up is also replacing RAC with single instance databases in a failover cluster solution, and with this will decrease a lot of RAC capabilities:
=> Rolling Upgrade (OS / Clusterware and some database patches)
=> Storage migration/consolidation
=> Fast failover (30 seconds vs. minutes).
See www.oracle.com and you will find whitepapers with differences between failover clusters and RAC.
But even if there are other justifications to reduce availablity (which actually can only be costs), I still wonder, why he wants to replace Oracle Clusterware - the clusterware shipped with Oracle is far more mature than the one from the Red Hat cluster Suite.
And there is no better integration for Oracle databases in clusterware than what Oracle clusterware can provide.
Even more so, OCW is a full capable clusterware and you can build your own failover cluster solution for free.
The same would have to be done if he replaces it with redhat cluster.
PS: If someone is providing failover scripts for Oracle in Redhat clusters, it will be Redhat (but I think you have to write them yourself).
user589812 wrote:Not a good idea. That is where RAC's forerunner, OPS or Oracle Parallel Server, was at.
Our boss will plan to configure Red Hat Cluster Suite to replace to update oracle RAC.
OPS came without a cluster s/w stack. The o/s vendor provided that.
This solution was often problematic. Not always that robust. Different vendor stacks behaved differently making diagnosing problems difficult. Problems had to be dealt with using 2 separate vendors support centres (that at best do not talk to each other, at worse point fingers at one another).
To address this mess, Oracle developed CRS (<i>Cluster Ready Services</i>)/Grid Infrastructure. This eliminates vendor issues (like implementing "standards" differently). Provides a common and tightly integrated platform for running RAC on. Enables you to get your RAC and Grid problems addressed and solved by a single vendor and support/maintenance contract. Etc.
Reverting back to the old approach of using different vendor stacks for grid and RAC? I fail to see any logic or sound reasoning backing this up.
Fact. It WILL cause problems. It WILL result in a lessor robust s/w stack on that server.
It WILL be a major mistake on your boss's side...
This long abandoned zombie of a thread, resurrected by a hijack post, is now locked.
Let it return to its slumber.
Grabbing someone else's thread for one's own question is very poor forum etiquette.