For such queries, a lot of records have the same score. My assumption is that the first results in the list come from the first core that has answered. I could count 4 different orders for BB and 6 for green which tends to prove this (now we have both clusters productive).
Possible solutions :
add some arbitrary parameters to the ranking (for example based on record id) to avoid having two documents with the same score
it might be possible to configure solr a bit differently to avoid such cases
The empty query or the "new acquisitions" lists deliver different results when reloaded :
For such queries, a lot of records have the same score. My assumption is that the first results in the list come from the first core that has answered. I could count 4 different orders for BB and 6 for green which tends to prove this (now we have both clusters productive).
Possible solutions :
@guenterh : any idea ?
This needs to be investigated a bit futher.