andyew on Fri, 27 Nov 2015 04:13:36

Hope you folks can help.

We have an application that uses SQL Ce in a merge replication scenario. It's been incredibly stable for many years. However we've discovered an oddity that no one seems to have experienced.

The scenario is this.

The application is installed an SDF is created and initialized and first sync'ed by USER A.(awilliams)

USER A logs out of the PC.

USER B logs into the PC.

USER B enters some data into the SDF and SYNCS.(suser)

USER B logs out

USER A logs into the PC enters some data into the SDF and SYNCs (awilliams)

What we see in replication monitor is this.

Don't get too excited about the path being a user specific\permission path. The SUSER was given full access. We have repeated this test many times using generic paths with the same result in Replication Monitor. It's just this is the only image I had available.

The highlighted text shows a new REPLNICK for each sync attempted. I am suggesting that, on a per user login and sync basis, our implementation is reinitializing the SDF.

This is undesirable as with each re-initialization the identity range is discarded. We're also having trouble managing reporting on success as the identity of the successful sync keeps changing.

Related tests where the intermediate user (SUSER) adds records but doesn't sync show the original creator and subsequent synchronizer (AWILLIAMS) DOES NOT reinitialize the subscription.

Also note reinitialization is non-destrcutive and all captured records from both users are passed up to the server in both testing scenarios.

I have also tested this performance using Erik Ejlskov Jensen's project Using both scenario's above I get the same performance with regard to the results in Replication Monitor.

Our questions are 

Can other's confirm this same performance?

Is this functionality by design?

Can we set up our sync parameters so, regardless of logged in user, the same SDF and initialization is used?




ErikEJ on Fri, 27 Nov 2015 08:02:07

Do you use SQL security or Integrated security?

andyew on Fri, 27 Nov 2015 08:50:55

We use SQL security in all cases.

andyew on Mon, 07 Dec 2015 00:40:28

If anyone even has vague ideas I'd be really interested to hear?