Hi,
We have problems performing SQL PULL merge replication over the web. Most of the time it works fine, but periodically we experience issues as described below. Our application uses SQL Express for subscribers and we have our own C# RMO code for the agent. We have eliminated our application as the source of the problem by setting up a full SQL Server as a subscriber and using the built in agent. It experiences the same problems.
1. Unable to download a snapshot.
After reinitializing a subscriber, the snapshot will partially download and will then suddenly stop with the error:
The connection with the server was terminated abnormally
There is no obvious timeout issue, the error typically occurs as soon as one of the .bcp files is about to be downloaded. It's often the same .bcp file that has the problem, but not always so there is no consistency.
The only solution is to continually retry until the download eventually succeeds. Recreating the snapshot before retrying seems to improve the chances of a successful sync.
2. Unable to perform a normal sync (i.e. an update, rather than a full snapshot)
An attempt to perform a sync results in the following information being logged:
The upload message to be sent to Publisher 'SERVER2' is being generated
The merge process is using Exchange ID 'F82F1A56-A81A-4786-A469-7A76259F9ED8' for this web synchronization session.
No data needed to be merged.
Request message generated, now making it ready for upload.
Upload request size is 1895 bytes.
At this point the sync simply stops doing anything. There are no errors, the agent still appears to be running, but nothing else happens.
We typically experience this problem once every few weeks. Reinitializing a subscriber does not help, the same problem occurs. The only solution is to recreate the snapshot and then to reinitialize the subscriber. This then requires the subscriber to perform a full sync which takes around 45 minutes for our dataset.
I am aware through online searches that others have experienced similar issues with SQL replication over web. Are there any practical solutions to these issues, or is SQL replication over web simply broken?