cassandra
cassandra copied to clipboard
Fix the duration of speculative read delay
Speculative reads will now be triggered after sampleReadLatency past the replica reads are issued instead of past queryStartNanoTime. It's because sampleReadLatency measures the actual execution of read request whereas queryStartNanoTime may, after CASSANDRA-19215, be measured before wait in the NativeTransportRequests queue.
Patch by Jakub Zytka; reviewed by TBD for CASSANDRA-19330