Oracle FAQ Your Portal to the Oracle Knowledge Grid
HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US
 

Home -> Community -> Usenet -> c.d.o.server -> Re: Redundant Replicated Systems???

Re: Redundant Replicated Systems???

From: Brian Douglass <briand_at_tps-eoltp.com>
Date: 2000/05/16
Message-ID: <39221F66.4483FC60@tps-eoltp.com>#1/1

Job wrote:

> slhuffst_at_my-deja.com wrote:
> >
> > I have been posed a question to which I have been unable to locate an answer.
> > Maybe someone can help me out.
> >
> > I have been asked to identify a method to configure a system with the
> > following qualities:
> >
> > 1) 2 pairs of replicated databases ( System 1 ( Server A & B replicated),
> > System 2 (Server C& D replicated) ( System 1B replicated to System2D) 2) The
> > ability to take one System offline - modify the Stucture of the Database -
> > while keeping the second system up and running.
> >
> > For instance:
> > System1 - taken offline
> > System 2 - continues transaction processing
> >
> > System 1 - modify database structure
> > System 1 - bring online (somehow)
> >
> > System 2 - propogate data mods to System 1
> > System 1 - progagate db structure changes to System2
> >
> > The order of this is not important - as long as the system is never
> > completely down.
> >
> > Has anyone ever seen anything like this? Any information would be appreciated.
>
> You're funny.
>
> Have you been able to allow system2 to continue operating
> without removing the replication group (Sys 1B & Sys 2D)?
>
> If so, I'd be interested in how! At present I can only
> see removing that group. Removing System1 repgroup.
> Make db structure changes. Add repgroup system1 back.
> Add repgroup (Sys 1B & Sys 2D) back. After getting
> status back to normal, remove (Sys 1B & Sys 2D) again!
> make db changes and then restore repgroup and coalesce.
>
> This is my understanding, I'd like to know it's wrong.
> Anyone?

I have been playing with Oracle 8.1.6 Replication for exactly the scenario originally proposed.

Oracle Replication has a large set of tools to accomplish a variety of scenarios. Oracle Replication provides a MultiMaster configuration which allows 1A to replicate to 2A, and transactions on 2A to be replicated back to 1A (I'll ignore B for now since it is the same process. Having 1A go to 1B is also allowed). This is on the data transaction level. Updates can be synchronous (slow) or asynchronous (delayed). However, I could only get the asych replications to process every 30 seconds, even though it was configured for every second. Unfortunately for me, I need something under 2 seconds.

Anyway, DDL statements at 1A processes are replicated to 2A synchronously. So there is no real need to stop 1A and mod the database. Both databases can be changed simultaneously. Solves the original problem nicely.

I found replication in 8.1.6 to be a very robust in its implementation. Unfortunately for me, my requirements went beyond its abilities. But if your requirements are in the tens of seconds to minutes range, it should work well.

WARNING: If you are trying to build a Highly Available system with geographic failover (System 1 in NY, System 2 in SF), research your requirements and the capabilities of your tools closely. I thought for sure Oracle Replication alone was going to solve my problems. Only after 2 weeks of intensive testing did I conclude that it was too coarse for my problem, after which I embarked down a different path.

Good luck

--
Brian Douglass
Transaction Processing Solutions, Inc.
8555 W. Sahara
Suite 112
Las Vegas, NV  89117
Voice: 702-254-5485
Fax:   702-254-9449
e-mail: briand_at_tps-eoltp.com
Received on Tue May 16 2000 - 00:00:00 CDT

Original text of this message

HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US