Home » Server Options » RAC & Failsafe » Rac and loading flat files (Oracle 10G)
Rac and loading flat files [message #381413] Fri, 16 January 2009 15:39 Go to next message
oakstream
Messages: 3
Registered: January 2009
Junior Member
Hello,
We are moving our database from a single instance to a 2 node RAC system on Solaris 10 and not sure how to handle the loading of flat files that get delivered to our file system.

In our previous system we had 2 seperate database servers which both received input files and both database servers loaded the data. This loaded to lots of inconsistencies, however, provided for redundant data in case of a failure.

We envision 1 node handing the loading using the job scheduler and services and the other node performing queries and other tasks. But the second node needs to be able to process/access the files if node 1 goes down and vice versa. (picking up where the other left off)

The issue we have is that how can Node 2 access the files if needed and pick up where the other left off. I'm wondering if a clustered file system would work? Such as OCFS? We could also send redundant files to both nodes, however, Node 2 wouldn't know where Node 1 left off in the load process? And not sure how to manage this. I would prefer to "share" the file system if possible thru clustering or (would NFS work?). I'm only thinking shared because then both nodes could tell where the other left off because it would just go to the inbound directory and look for files?
I'm not looking for details here, just a high level architecture. Any help would be appreciated.

Re: Rac and loading flat files [message #381414 is a reply to message #381413] Fri, 16 January 2009 16:02 Go to previous messageGo to next message
Mahesh Rajendran
Messages: 10672
Registered: March 2002
Location: oracleDocoVille
Senior Member
Account Moderator
Not sure I understand your requirement.
In RAC, Oracle datafiles are kept in a shared storage.
You can can use an NFS mount accessible from both nodes as your "flat file" area.
All rest (resuming operations etc) depends on how you are exactly loading the data. Shed more light on that, probably we can address specific issues.
Re: Rac and loading flat files [message #381415 is a reply to message #381413] Fri, 16 January 2009 16:08 Go to previous messageGo to next message
Mahesh Rajendran
Messages: 10672
Registered: March 2002
Location: oracleDocoVille
Senior Member
Account Moderator
>>both database servers loaded the data. This loaded to lots of inconsistencies, however, provided for redundant data in case of a failure.
What kind of failure are you talking about here?
All Oracle RAC does is, it will protect from a single node/instance failure. Using RAC, you will have many instances of database spread across many servers.
If one of the node/instance goes down, others will serve.
If the database goes down, RAC is if no use.

With your old architecture, it seems you have a redundant database.
To do the same in a much efficient method, you need a Standby database using DataGuard.
Re: Rac and loading flat files [message #381421 is a reply to message #381413] Fri, 16 January 2009 18:29 Go to previous messageGo to next message
oakstream
Messages: 3
Registered: January 2009
Junior Member
Thanks for responding.

My issue is not with Oracle data files but with input data files that will be loaded into Oracle (sales records). Files get delivered into a directory such as /input and I need the Rac instances to process these files. (Load them into the database).

With an NFS mount, Can I have both instances read and write into the NFS mounted file system concurrently?... or do I need something like a clustered file system for these files...or would this even work?

The other option is to send files to RAC1 and duplicate files to RAC2. Just use RAC1 to process all the files and if it goes down then RAC2 can pickup. My problem is that when RAC2 picks up it will have all the files in the directory (including files that were already loaded with RAC1). So I will need to put some logic in to maintain the file system as well as delete or not load files that were previously loaded.



Re: Rac and loading flat files [message #381425 is a reply to message #381413] Fri, 16 January 2009 19:43 Go to previous messageGo to next message
BlackSwan
Messages: 25046
Registered: January 2009
Location: SoCal
Senior Member
It appears you misunderstand what RAC is & how it works.

With RAC, only 1 collection files exists.

You need to RTFM found at http://tahiti.oracle.com on RAC.
Re: Rac and loading flat files [message #381426 is a reply to message #381421] Fri, 16 January 2009 20:06 Go to previous messageGo to next message
Mahesh Rajendran
Messages: 10672
Registered: March 2002
Location: oracleDocoVille
Senior Member
Account Moderator
Quote:

The other option is to send files to RAC1 and duplicate files to RAC2. Just use RAC1 to process all the files and if it goes down then RAC2 can pickup.

As BlackSwan indicated, You have totally misunderstood how RAC works. There are NO TWO databases. Only two instances.
Both RAC1 and RAC2 will talk to the same database. Period.
If RAC1 goes down, RAC2 will serve the data from database.

>>but with input data files that will be loaded into Oracle (sales records).
That is what I am talking about.
>>With an NFS mount, Can I have both instances read and write into the NFS mounted file system concurrently?
Yes. I use it all the time. Look in TAF options.
You can use any supported cluster file system also.
But i would keep it away from the filesystem used by database.

[Updated on: Fri, 16 January 2009 20:13]

Report message to a moderator

Re: Rac and loading flat files [message #381535 is a reply to message #381413] Sat, 17 January 2009 18:08 Go to previous message
oakstream
Messages: 3
Registered: January 2009
Junior Member
Thanks for the responses. I understand RAC just having a tough time explaining the question. I'll post question when I can clarify it. Thanks
Previous Topic: new DB using DBCA
Next Topic: ASM diskgroup
Goto Forum:
  


Current Time: Thu Dec 08 02:20:37 CST 2016

Total time taken to generate the page: 0.06124 seconds