I am completely confused about what is going on in my SolrCloud setup. That said, you see I am new to the topic.
My current setup for testing is: I have created a SolrCloud on a single server (I do not have more for testing in the moment). I use the embedded Zookeeper. I have three shards on three nodes with two replica in each one. Every node has a different IP (8984,7574,7575).
I set up a collection with
curl 'http://localhost:8984/solr/admin/collections?action=CREATE&name=mycollection&numShards=3&replicationFactor=2&collection.configName=myconfigset&router.name=compositeId'
When I index some Documents with
curl 'http://localhost:8984/solr/mycollection/update/csv?update.chain=all-into-one&commit=true&separator=%09&encapsulator=%00&header=true&trim=true&f.itau_name3.split=true&f.itau_name3.separator=;&f.itau_lnx3.split=true&f.itau_lnx3.separator=;&f.itau_af3.split=true&f.itau_af3.separator=;&f.itau_init3.split=true&f.itau_init3.separator=;' --data-binary @file.txt -H 'Content-type:text/plain; charset=utf-8'
the routing works ... sometimes. It seems to depend on the prefix/compositeID I use:
When I use r1!<uniquekey>, r2!<uniquekey>, r3!<uniquekey> a query : with _route_=r1! returns documents with r1 and r2 prefix. _route_=r3! seems to be correctly routed. In the very same setup (after clearing the index and reindexing) the same with year1!, year2!, year3! as prefixes works. A query with q=*:*&_route_=year1! etc. only returns results of the respective route-prefix.
Can somebody help me to find out what's going on here? Are there any restrictions on what how to construct the route-prefix?
you did not show your full query, but, I suspect this might be happening:
Does this make sense?