Re: Can you expect from snapshot logs to support huge refresh volumes ?!?

spendius_at_muchomail.com (Spendius) wrote in message news:<aba30b75.0312050701.18888fd2_at_posting.google.com>...
> Hi,
> I'm beginning to wonder if the mechanism of fast
> refreshes using the snapsh. logs MLOG$_* tables on
> the master side is supposed to be able to bear heavy
> loads and huge volumes of data to replicate. Here is
> why:
>
> Here a certain module of our app can generate 2 to 400 000
> of rows in a given snapshot on *EACH* site (we're dealing
> with one master site here, linked to 25 snapshot sites) =>
> every time this operation happens, ALL sites need 3 to 4 days
> to be suitably refreshed, we get phone calls from all over
> these sites of people complaining (the management complains as
> well) etc., in brief the replication queue becomes overloaded
> and the DBAs situation delicate ! It *looks like* this mechanism
> doesn't allow for following the application requirements and
> needs.

Three days is way, way too long. 400,000 rows isn't that great a volume in the grand scheme of things.

> We're on 8.1.7.4 on Sun/Solaris 8. The master DB runs on an
> Enterprise 450 with 2 CPUs @ 480MHz, the snapshots servers are
> Sun Fire 280. On snapshot sites refresh jobs occur every 10
> minutes.
>
> Any ideas (or questions, if you want more details about our
> environment) ?

Do you see activity on the remote sites for four days? Are all the remote sites on 8.1.7.4 as well? Is the network connection between the master and remote site potentially flaky? I seem to remember that there was a bug in 8.1.7 if there is a network hiccup, but I believe that was fixed in 8.1.7.4.

Justin
Distributed Database Consulting, Inc.
www.ddbcinc.com/askDDBC Received on Sat Dec 06 2003 - 00:48:43 CST