Re: Excessive redo generation on Master site in a very frequent master/snapshot replication

swim4life wrote:
> Hi all,
> I have the following problem:
> in a Master snapshot replication environment, which starts each 5 secs,
> regardless the number of DML (which is quite small), a huge quantity of
> redo logs are produced during a day.
>
> Even if the frequency of replication is very high, only few hundred of
> rows are replicated per hour, nonetheless, at the master site the
> redolog entries produced are really a lot of megas.
>
> Mining the logs, I see that each time the snapshot job starts, even
> though there are no rows to be propagated, an update on the MLOG$ table
> occurs, which, of course, produces the related entries in the current
> online redolog.
>
> Unfortunately, since the changes occurs randomly and must be propagated
> immediately, I need to take the snapshot frequence so high.
>
> As you know, is there a way to avoid this excessive generation of
> logs?
>
> About the environment:
>
> -- Windows 2000 Server
> -- Oracle EE rel.8.1.7.4
>
> Thanks in advance for any suggestion.
>
> lottini
>

Your biggest problems are:

1. de-supported version of Oracle RDBMS and (AFAIK) de-supported version of Windows server
2. Advanced Replication is considered a legacy product, you should plan to migrate to Streams-based replication as of version 9i or later.

Beyond that, what do you gauge as "really a lot of megas" or "huge quantity of redo logs" for your redo generation rate? How do you know that the replication process is the cause of your high redo rate?

-Mark Bole Received on Mon Aug 29 2005 - 21:40:33 CDT