Re: HA Clustering with RAC

From: Jackson A. Krainer <jack_at_cade.com.br>
Date: 18 Mar 2004 09:33:00 -0800
Message-ID: <a3ef186b.0403180933.6ddbc0df_at_posting.google.com>


Ulf.Pralle_at_web.de (Ulf) wrote in message news:<57f997b6.0403180247.11eecb62_at_posting.google.com>...
> Hi guys,
> I'm writing my diploma thesis about Disaster Recovery. I try to differ
> types of cluster depending of the layer where they are realized.
> Here is my point of view:.This means that the OS, the database and the
> application (e.g. SAP R/3) must be active on the secondary site.
>
> Here are my questions:
> On which layer (OS,DBMS,Application) is the failover realized? Can it
> be realized on different layers?Which application does support this
> function? Or have all applications to support this function?
> Do you always need a special cluster software like HACMP?
>
> I would be glad, if you could answer my questions and give me links
> where I can read something about this.
>
> Bye

Let me try to explain my point of view. I don't agree with you when you said "The aim of a cluster is, that there is an automatic failover of resources to the secondary site that no transaction is lost". Transactions are always lost because threre is a time used by the high availability software to detect an transport all resources from one site to the other site.
The fail over occurs in different layers. Using HACMP as an example, it first stop all applications running on a given server, after that it change the IP address of the server and of the contingence server, after this the product starts all applications on the contingence server.
One more time is my understanding of how a cluster works. Regards,

Jackson Received on Thu Mar 18 2004 - 18:33:00 CET

Original text of this message