I don’t exactly remember the mechanism but roughly the main reason behind is, lets say you wanted the top 10 results. By default solr considers say ~15 top documents from each shard/replica as candidates for top 10. What if the 16th result from a couple shards had many documents and would have superseeded the 15th result if it was considered? This results in an incorrect top 10 list. And what if something in that top 10 list was on the 16th place in some replica? Then it won’t be counted towards the end count and result in an incorrect count.
When you search about overrequest and overrefine parameters there are resources which explained that much better than me 😃 those parameters are there to alleviate this problem. Such as: https://medium.com/@sarkaramrit2/overrequest-and-refine-json-facet-bucket-counts-in-solr-1acfa77cd90c -ufuk yilmaz Sent from Mail for Windows From: Modassar Ather Sent: Wednesday, August 23, 2023 2:01 PM To: users@solr.apache.org Subject: Re: Effect of facet.limit on facet count value. Thanks for your response. The Solr cluster is deployed on 48 shards and all the documents of a group is on one shard only. Can you please help me understand the reason behind this behaviour of faceting? Thanks, Modassar On Wed, 23 Aug 2023 at 3:55 PM, ufuk yılmaz <uyil...@vivaldi.net.invalid> wrote: > I’d suggest to take a look at the overrequest and overrefine parameters, > especially if running SolrCloud and collection in question has multiple > shards > > Sent from Mail for Windows > > From: Modassar Ather > Sent: Wednesday, August 23, 2023 11:36 AM > To: users@solr.apache.org > Subject: Effect of facet.limit on facet count value. > > Hi, > > With an increase and decrease in the value of facet.limit the facet counts > are changing. This is happening in grouped faceting. > > I noticed when the facet limit is set to -1 or a higher number the grouped > facet counts are correct. > > Kindly help me understand and resolve this issue to get the correct grouped > facet count. > > Thanks, > Modassar > >
