Each node needs a -Dsolr.solr.home pointing to a solr.xml, but the
configuration-subfolder does not need to be there. It only needs to be
there for the node you start with -Dbootstrap_confdir (to have it load
the config into ZK). The next time you start this Solr you do not need
to provide -Dbootstrap_confdir, since config is already loaded into ZK
(well unless you run your ZK embedded in the Solr - in this case I
believe all ZK state is removed when you close the Solr, but that is
also just for "playing")
In general, IMHO, using a Solr node to load a configuration during
startup is only for "playing". You ought to load configs into ZK as a
separate operation from starting Solrs (and creating collections for
that matter). Also see recent mail-list dialog "Submit schema definition
using curl via SOLR"
Regards, Per Steffensen
On 1/23/13 11:12 PM, Walter Underwood wrote:
I can get one Solr 4.1 instance up with the config bootstrapped into Zookeeper.
In zk I see two configs, two collections, and I can run the DIH on the first
node.
I can get the other two nodes to start and sync if I give them a
-Dsolr.solr.home pointing to a directory with a solr.xml and subdirectories
with configuration for each collection. If I don't do that, they look for
solr/solr.xml, then fail. But what is the point of putting configs in Zookeeper
if each host needs a copy anyway?
The wiki does not have an example of how to start a cluster with multiple
collections.
Am I missing something here?
wunder
--
Walter Underwood
wun...@wunderwood.org
