[ https://issues.apache.org/jira/browse/SOLR-6822?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Greg Harris updated SOLR-6822: ------------------------------ Description: If you have in your setup _ root _ a different type than your uniqueKey, when you update blocks with the same id, you start getting duplicates of the entire block. So if you have parent->child->grandchild and update the block with the same or different children you will have both blocks in their entirety still in the index found with "star":"star" This can further create wierdness when doing calls from different shards as a call to the shard itself will give back all results... a call to a different shard with shards=dupeshard will take out the duplicates being sent for rows=x and affects the numFound and total docs returned if there are dupes in the rows being returned. So it is not consistent. was: If you have in your setup _ root _ a different type than your uniqueKey, when you update blocks with the same id, you start getting duplicates of the entire block. So if you have parent->child->grandchild and update the block with the same or different children you will have both blocks in their entirety still in the index found with "star":"star" This can further create wierdness when doing calls from different shards as a call to the shard itself will give back all results... a call to a different shard with shards=dupeshard will take out the duplicates being sent for rows=x and affects the numFound and total docs returned if there are dupes in the resultset. So it is not consistent. > Block Join duplicates when _root_ is different type than uniqueKey > ------------------------------------------------------------------ > > Key: SOLR-6822 > URL: https://issues.apache.org/jira/browse/SOLR-6822 > Project: Solr > Issue Type: Bug > Affects Versions: 4.10 > Reporter: Greg Harris > > If you have in your setup _ root _ a different type than your uniqueKey, when > you update blocks with the same id, you start getting duplicates of the > entire block. So if you have parent->child->grandchild and update the block > with the same or different children you will have both blocks in their > entirety still in the index found with "star":"star" > This can further create wierdness when doing calls from different shards as a > call to the shard itself will give back all results... a call to a different > shard with shards=dupeshard will take out the duplicates being sent for > rows=x and affects the numFound and total docs returned if there are dupes in > the rows being returned. So it is not consistent. -- This message was sent by Atlassian JIRA (v6.3.4#6332) --------------------------------------------------------------------- To unsubscribe, e-mail: dev-unsubscr...@lucene.apache.org For additional commands, e-mail: dev-h...@lucene.apache.org