My understanding of the snapshot in merge replication is to provide a set
of schema and data so that the subscriber has a base point to start with.
We have a site that was running for a number of weeks quite happily, and
then we upgraded the software and re-generated the publications and
subscriptions which worked correctly.
The snapshot rebuilt itself at midnight and then today we find that the
subscribing sites databases have no data in them anymore, but there are no
deletes in the merge agent log!
Some questions
1. Should we be scheduling the snapshot to regenerate each night?
2. Where's the data for the snapshot held - there aren't any bcp files in
the snapshot directory.
3. Each article is marked with a pre-action of 'drop table', since the
schema changes at the publisher need to be pushed down to the
subscriber, is this likely to be causing the 'disappearing data' problem?
Regards
Paul Hatcher, MCSD
Principal Consultant
Graduate Associates Ltd