On 12/30/2014 5:03 PM, Charles Sanders wrote: > Thanks for the suggestion. > > I did not do that originally because the documentation states: > This parameter is not required for the /select request handler. > > Which is what I am using. But I gave it a go, even though I'm not certain of > the shard names. Now I have a NPE. > > solr/collection1/select?q=kernel+p&rows=1&wt=json&indent=true&shards.qt=/ac&shards=shard1,shard2 >
If this is not SolrCloud, then the shards parameter must include most of the full base URL for each shard that you will be querying. You can only use a bare shard name if you're running SolrCloud. The shards.qt parameter that you have used means that when the shards are consulted, the /ac handler will be used rather than /select. Here's an example of a shards parameter that will combine results from three cores on two machines. When not running SolrCloud, this is how you do distributed searching: shards=idxa2.example.com:8981/solr/ai-inclive,idxa1.example.com:8981/solr/ai-0live,idxa2.example.com:8981/solr/ai-1live SolrCloud hides almost all of this complexity. Thanks, Shawn