Hi,
We've been using bucardo in its various incarnations for years now, and love it. We're actively changing our database connectivity structure and wanted to check for sanity before we get too crazy.
We have many clients, all of whom currently have a separate database on our servers. Each client has their own mobile server which is periodically connected to the internet but spends most its life powered down. We currently have each of those mobile servers set up to be a unique bucardo server, so when it powers up it starts bucardo and syncs all the fresh data from that client's database.
We are planning to consolidate all the separate databases on our central servers to a single database. So now, there is a single central database, and many client databases that are normally not connected but when they are should get a copy of all fresh data. While powered on, each should replicate live.
Configuration A: Each mobile server continues to run its own bucardo process and syncs data with the main central server whenever the mobile version is powered on. The issue here is that the central server would need to somehow track deltas for each mobile database. I'd be OK with having a different bucardo schema for each client if that was the only way (e.g. instead of bucardo.*, bucardo_clientx.*). That seems cumbersome to me.
Configuration B: bucardo runs on our server cluster and handles the replication down to each mobile database whenever the database comes online. Lots of special processing happens when a client comes online anyway (like software auto-updated via git), so adding something to kick the main server would be easy. However, I've been skimming this list and it seems that we'd have an issue since all mobile databases are never online at the same time and from my cursory reading of this mailing list over the years, bucardo needs all client master databases online simultaneously to replicate to any of them. Did I misread, or is there some way to cope with this?
Note that I'm only concerned with the bucardo configuration, the primary key issue is one that we solved years ago (all databases are set up with a unique code that is prefixed to each primary key in the database, so we can even tell which database originated a specific piece of data).
Any thoughts? Chris. _______________________________________________ Bucardo-general mailing list [email protected] https://mail.endcrypt.com/mailman/listinfo/bucardo-general
