[mapguide-internals] Berkeley DBXML replication

[Tom Fukushima]

There were a couple of emails about this topic...
Jason started with:

I believe that the site server/support server used MapGuide's internal TCP/IP to communicate between the various services.  BDB replication would allow you to run a repository "farm", with a single master and multiple slaves.  Writes all go to one repository, and reads get balanced to the slaves.  Apparently it allows the client software to elect a new master from the slaves if the primary is down too.  This would nicely get around the current problems with load balancing and sessions / server affinity.  I didn't get a chance to look at it in depth, but it seemed to be a good fit.



Zac followed up with:

Jason, your comments about the berkley replication stuff was quite

interesting, but was

that in fact what the old site server / support server stuff was based on?


Some follow up would be good so I'll reply to this best as I can. I've been told that the replication support was investigated a few years ago but the API was quite complex, so we probably just didn't have the resources to do anything here.  Replication was not used for Site/support servers.  The one issue that I see with replication is that MGOS doesn't store everything in the repository.  I would think that everything stored in the DataFiles section would be problematic to replicate; but I'm just guessing here.

Tom