I'm a little confused. The "aggregator" node is just the Solr node that happens to receive the call from the client. There is no fixed aggregator node unless you're doing something custom.
So any replica can act as the aggregator, and in fact each and every node acts as an aggregator at some point if there's any decent load balancing going on. If you're accessing Solr from a SolrJ client, the internal LBHttpSolrClient is sending queries to random nodes on the system. So it doesn't seem like there's any point to trying to load the cache only on the searchers since all the Solr nodes in the collection will be searchers at some point. This is assuming SolrCloud, which I infer from your comment that this is in a distributed environment. Or I'm totally confused ;) Best, Erick On Fri, Mar 25, 2016 at 9:21 PM, Jitka <jitk...@gmail.com> wrote: > Hello. My company uses Solr-4.10 in a distributed environment. I have > written a SearchComponent which contains a cache which is loaded at > start-up. The cache is only used on the searchers, never on the > aggregators. Is there some way I can signal that the cache should be loaded > only if the Solr instance in question is a searcher, not if it is an > aggregator? At present the cache is loaded in the inform() function; is > that the wrong place for it? > > Thanks in advance for your help. > Jitka > > > > -- > View this message in context: > http://lucene.472066.n3.nabble.com/Can-I-take-Solr-role-aggregator-searcher-into-consideration-at-start-up-tp4266197.html > Sent from the Solr - User mailing list archive at Nabble.com.
