Re: [Katta-developer] Katta "under the hood" documentation?

[Ted D. <ted...@gm...>]

On Mon, Sep 26, 2011 at 1:20 AM, Johannes Zillmann <jzi...@go...
> wrote:

> 2 roundtrips are involved querying a distributed katta lucene index. First
> one get the document frequency for the query-terms for every shard. Second
> roundtrip contains the query, sending the document-frequencies along for
> proper ranking.
>

This guarantees identical ranking.  It is also possible to presume random
distribution and use only one round-trip.


>
> > - Split Brain syndrome (e.g. I link up 2 LANS that were previously
> disconnected and ran 2 separate groups)?
>
> There is no mechanism for automatically 'merging' two groups of katta-nodes
> with 2 master to a single group with one master if thats what you're asking!
>

Split brain in the sense of one cluster that split due to network fault does
not really exist.  Zookeeper avoids that.  If you separated the clusters
yourself, then you can add the nodes to the cluster and they will integrate
in pretty sensibly, deleting redundant indexes (if the indexes are
compatible).