On Thu, May 28, 2009 at 3:14 PM, Ashish P <ashish.ping...@gmail.com> wrote:
> > Hi Shalin, > I am trying to do the same point again. You suggested to use HTTP for > replication. > But for performance reasons I want to use embedded. What kind of performance are you looking for? We use http with very large indexes and very high traffic. Http helps us scale. > And because replication > is not available with embedded approach, I am trying this i.e. from two > different web applications (same network) using embedded solr giving same > data dir. just trying to index. > Are you using a shared network disk with the index on it? Even if you point multiple embedded solr servers to the same index, you should write with only one. Once you do a commit on the writer, you'd need to do a commit on each of the other embedded solrs as well to make them re-open the searcher. The problem with this approach is that you have a single point of failure (the shared disk) and performance with NFS is very bad (probably even worse than the overhead of http). -- Regards, Shalin Shekhar Mangar.
