Hi, Currently both secondary index build and compactions are run by compaction manager within the same thread pool (C* 2.0.x). I'm currently experiencing a situation where repair's stream sessions get stuck waiting because they don't get scheduled to build secondary indexes (and finalise the stream session) because there are already many compaction tasks pending to be finished (the node is recently bootstrapped.).
Is this the expected behaviour? How is it possible to run repair on a column family with secondary indexes while one of the nodes involved in repair have loads of pending compaction tasks? Thanks, Omid
