On Dec 7, 2010, at 12:43 AM, Matt Adams wrote: > Hi folks, > > I have a scenario where it would be really great to leverage BigCouch at the > data center level and regular CouchDB at the mobile client level. > > Are there any hurdles to doing this? The intent is that replication will > either involve many smaller databases (most mobile clients will only need to > replicate a few of these) or a few very large databases in which case a > client will likely only replicate a very small subset of one of these > databases (using filters and so on). > > I would appreciate if anyone with a better knowledge of both projects would > be able to chime in with some reasons "why this would not work" or draw my > attention to caveats. > > > Thanks in advance, > > Matt
Hi Matt, you'll want to be using 1.0.1 for this on the CouchDB side, as it contains some fixes for dealing with the weird sequences transmitted by BigCouch. Other than that, there's one bug I'm aware of that can cause the replicator to stop saving checkpoints when BigCouch is used as a source. The bug stems from the fact that the replicator sorts sequences, and the BigCouch sequences are strings, so when the replicator sees a sequence like "9999-..." it thinks it has reached the end of time :P The replication will continue, but it won't save any checkpoints, so the next time it runs it will scan through a bunch of revisions again (but it will save checkpoints then). Best, Adam
