patelh

For the time being, we've overcome this by caching the instance via backport of https://github.com/apache/spark/pull/24905/files to Spark 2.3.2

The threads are waiting on the same monitor. We haven't been able to root cause this yet. It is possible the JVM was in bad state due to heap corruption....

This was a non issue, do we want to close?

Last I checked Druid had a limit on the number of rows that can be sent from historical to broker and it didn't support pagination.

You can pass in your own value, see register method: https://github.com/yahoo/maha/blob/d6b48f24353d3a915c169286fbfc4fd80a1bd82b/core/src/main/scala/com/yahoo/maha/core/query/druid/DruidQueryGenerator.scala#L166

Generally, we like to keep this kind of functionality implemented across all query generators. It is not specific to one engine.

The current implementation assumes all items in the filter list are AND'd. We would want to keep the existing functionality intact. However, the user has the option to use logical...

@ryankwagner What you are proposing is the same, top level filter which is a logical operator. That's what I also said, just that the existing functionality of not having top...

Does the request have forceDimensionDriven set to true? Try to get debug info to see if isDimDriven or forceDimensionDriven are true on request model.

Also try running the request with forceDimensionDriven: "false" to see what the query looks like.