New to RAC

Question

Hi,
I have been using Oracle 9i and 10g for several years and have worked extensively in all DBA tasks for single instance databases. But I have never worked with Oracle RAC. I have read a lot of RAC details on internet, but it seems it is not possible to setup an experimental RAC setup at home, because we need a shared disk storage.
I have 2 PCs (1 laptop, 1 desktop) at home, both have Windows XP SP2, and Oracle 10g (10.2.0.1) installed. I have now bought a separate 500 GB extenal USB hard disk, now my question is : Can I use the 2 computers and the external harddisk to set up a RAC installation at home? if yes, can somebody please give me some overview of the steps involved for this?
An early reply would be greatly appreciated. Thanks in advance.

Answer 1

you wrote:
to install RAC SWIM should have a user equivalence for grid owner and database owner.Now sure what you mean by RAC SWIM - not familiar with the abbreviation or acronym SWIM for RAC.
suppose every pc have same user in that case(in a two nodes home setup it could be a reasonable scenario),if the SWIM copy and paste the shadow and passwd file of one machine to others in /etc will that work?I would not recommend copying shadow and passwd files between machines (you forgot about the group file). As this assumes everything is identical. And ideally in a perfect world, it should be - down to the actual UID and GUID used (amongst other things). Invariable this is not the case (especially when you deal with other parties that prep the machines and test/commission the hardware).
I have my own rac shell script command - that enables me to run a command (or series of commands) on all nodes in the cluster, or on selected nodes only (using ssh). I prefer using this.
Another method I at times will use is KDE's konsole and open a ssh session to each of the 10 or more cluster nodes. konsole has a feature that enables all the tty input to one konsole session be directed to all other sessions. This enables you to, for example, enter and execute a command on node 1 and have that command also being entered and executed in the other konsole sessions.
another thing is that,this grid is implemented as a concurrent programming ie. by executing processes in different machine and those processes are communicating with each other... is there any such process who is responsible to run threads for certain job/ operations and those threads are communicating with same processes of different machine?Threading itself is not used by Oracle on Unix/Linux kernels. It uses multi-processing (the fork() model) instead.
Yes, IPC (Inter Process Communication) is used. That is one of the major tasks of the SGA - this shared memory area serves as the "+brains+" (common communication/nerve centre/data area).
Yes, some processes are "+process controllers+" - and will spawn and control and manage one or more other processes. For example, there are several process pools that are managed - where new processes can be started, failed processes restarted, minimum and maximum processes in the pool maintained, and so on. Think for example about the parallel query process pool, the shared servers pool, the job process pool and the dispatcher process pool.
Parallel query (PQ) processes can be used to service a single query - and these PQ processes can run on multiple RAC nodes concurrently for that single query.
Each of these processing pools are managed differently. If you want to read up on these, then I suggest trying the Oracle Database Concepts guide (you'll find the copy for your Oracle version via http://tahiti.oracle.com).
is there any short of Mesh Partitioning implemented in this clusterware or we need to go for other 3rd party clusterware ?? Depends on what you define as mesh partitioning. Remember that RAC is a shared-everything cluster architecture.
So there's no fundamental requirement to partitioning the cluster as the full set of database services are available on each node.
For management, performance, availability and other reasons, one may decide to run certain applications on a set of nodes, use another two nodes for web applications, one node for running batch processing and so on.
For example, job processes can be scheduled with node affinity. Oracle network services can be configured to be available on specific nodes. Etc.
The decision on how to structure and implemented it, is driven by your requirements and not by technical necessity.