It looks like SnapPuller.java doesn't allow for the possibility of the slave having a later index version than the master. It only checks whether the versions are equal.
It's easy enough to add that check and prevent the index fetch when the slave has a later version (in fact I'm running it in a sandbox right now). But I wonder what other problems it might create in a production environment (or what problems I am overlooking). Does anyone have any thoughts on this? thanks, Tim On Thu, May 13, 2010 at 2:37 PM, Tim Heckman <theck...@gmail.com> wrote: > Does bi-directional replication work in solr 1.4? In other words, if I > wanted to have 2 servers that are both master and slave. Call them A > and B. I would configure things so that normally, A runs a DIH > periodically to rebuild the index, and then B pulls the updated index > from A. The idea here is that if A goes down, B could run the data > imports, and then A would pick up the up-to-date index from B when it > comes back up. > > Based on the note about Repeaters on the wiki, it looks like it might > be possible, but there aren't enough details for me to know for sure. > > thanks, > Tim >
