Re: Transactional replication: Deadlock issue

hi,

Let me explain the topology of the system and the need that prompt/
force me to adopt it:

Topology:

DB1--Transactional Replication-->DB2--Merge Replication(download to
subscirber, allow subscriber changes)-->DB3

There are two sets of users using the systems(System1 with DB1 and
System2 with DB3's as database).

There is a non-detailed version (System1) and a detailed
version(System2).

For the detailed version I have created user categories( for each user
category we have an instance of System2).
Basically the data flows from System1 to System2(not backwards) and as
there are multiple instances of System2 what I did was, created an
intermediate db server DB2 and set up merge replication between DB2
and DB3's to filter the data coming to DB3's as per the user category.

I think I have made myself clear so far, if not I can explain
more.Please ask.

Now to Paul's input:
If you just initialize the data betweek DB1 and DB2 without DB3 being
affected then won't there be non-convergence with DB3? If there are rows now
in DB2 and not in DB3 which the distribution agent updates, there will be an
error from the merge agent as it tries to apply the changes on the
subscriber. I'm assuming that there aren't relevant filters which will
prevent this.
At present DB1 and DB2 are out of sync and replication is not working
between them. Now as DB2 and DB3 are in sync and there is some data in
DB3 that I can't lose, what I can try is copy DB1 data to DB2 and
setup the subscription again without initializing DB2 after setting up
subscription.I dont think it will have any adverse effect on the merge
replication between DB2 and DB3.

I'm wondering whyt there are tables which are being republished in the merge
publication. Would it be possible to have the transactional publication
publish to DB3 and just have the merge publication apply to the remaining
tables.

As per my understandings of replication I dont think it is possible to
apply filters in transactional replication and as I have mentioned
earlier, I need to filter the data to DB3's based on the categories I
need merge replication.

Thanks,
Rakesh.



On Apr 14, 2:50 pm, Paul Ibison <Paul.Ibi...@xxxxxxxxxxxxx> wrote:
If you just initialize the data betweek DB1 and DB2 without DB3 being
affected then won't there be non-convergence with DB3? If there are rows now
in DB2 and not in DB3 which the distribution agent updates, there will be an
error from the merge agent as it tries to apply the changes on the
subscriber. I'm assuming that there aren't relevant filters which will
prevent this.
I'm wondering whyt there are tables which are being republished in the merge
publication. Would it be possible to have the transactional publication
publish to DB3 and just have the merge publication apply to the remaining
tables.
We probably need to understand the topology and schemas a bit more to form a
proper set of advice.
Cheers,
Paul Ibison

.

Relevant Pages