Hello! We have a large table in our Druid cluster that whenever it gets queried ends up eating all of the Druid resources. Is there any way to fence that particular datasource so that it doesn't end up eating all of the CPU? Our problem is that queries that normally execute in 10 ms end up taking > 10 seconds due to high resource usage on the large queries.

This doc provides some options: https://druid.apache.org/docs/latest/operations/mixed-workloads If you know who is making the query you could force all their requests to into a slower "lane", effectively giving those queries less resources to work with. They requests will take longer, but they'll have less of an impact on other requests. The other option is Historical tiering. Its a bit more complicated and 💰, but you could setup dedicated historicals that only host that datasource. That would mean requests against that datasource could not (mostly) effect queries to other data sources.

Thanks; will look into this