v.22.1Experimental Feature
Parallel Reading from Multiple Replicas in Distributed Queries
Parallel reading from multiple replicas within a shard during distributed query without using sample key. To enable this, setallow_experimental_parallel_reading_from_replicas = 1andmax_parallel_replicasto any number. This closes #26748. #29279 (Nikita Mikhaylov).
Why it matters
This feature allows distributed queries to read data in parallel from several replicas of the same shard, improving query performance and resource utilization when sample keys are not available. It addresses the limitation of single-replica reading in such scenarios, enhancing efficiency and reducing latency.How to use it
Set the server configuration parameterallow_experimental_parallel_reading_from_replicas = 1 and assign a number greater than zero to max_parallel_replicas to specify how many replicas can be read in parallel.