thanks Mark and Tomas. Tomas, you mean doing soft commits to all the slave nodes right? If so, that is what I'm planning to do with the update processor commented above.
2011/10/21 Tomás Fernández Löbbe <tomasflo...@gmail.com>: > I was thinking in this, would it make sense to keep the master / slave > architecture, adding documents to the master and the slaves, do soft commits > (only) to the slaves and hard commits to the master? That way you wouldn't > be doing any merges on slaves. Would that make sense? > > On Fri, Oct 21, 2011 at 5:43 PM, Mark Miller <markrmil...@gmail.com> wrote: > >> Yeah - a distributed update processor like the one Yonik wrote will do fine >> in simple situations. >> >> On Oct 17, 2011, at 7:33 PM, Esteban Donato wrote: >> >> > thanks Yonik. Any idea of when this should be completed? In the >> > meantime I think I will have to add docs to every replica, possibly >> > implementing an update processor. Something similar to SOLR-2355? >> > >> > On Fri, Oct 14, 2011 at 7:31 PM, Yonik Seeley >> > <yo...@lucidimagination.com> wrote: >> >> On Fri, Oct 14, 2011 at 5:49 PM, Esteban Donato >> >> <esteban.don...@gmail.com> wrote: >> >>> I found soft commits very useful for NRT search requirements. >> >>> However I couldn't figure out how replication works with this feature. >> >>> I mean, if I have N replicas of an index for load balancing purposes, >> >>> when I soft commit a doc in one of this nodes, is there any way that >> >>> those "in-memory" docs get replicated to the rest of replicas? >> >> >> >> Nope. Index replication isn't really that compatible with NRT. >> >> But the new distributed indexing features we're working on will be! >> >> The parent issue for this effort is SOLR-2358. >> >> >> >> -Yonik >> >> http://www.lucene-eurocon.com - The Lucene/Solr User Conference >> >> >> >> - Mark Miller >> lucidimagination.com >> 2011.lucene-eurocon.org | Oct 17-20 | Barcelona >> >> >> >> >> >> >> >> >> >> >> >
