We're updating the issue view to help you get more done. 

Simplify the Lucene writing backend by dropping the parallel execution of writer tasks

Description

This patch will drop the following configuration options:

  • thread_pool.size

  • buffer_queue.max

As both are hard to tune and of doubtful usefulness.

The MassIndexer doesn't use this pool anymore, and so there is no strong need anymore for parallelizing work at this very fine-grained level. It's also possible that the thread comunication here would make it all less efficient, making any benefit questionable.

If someone really experiences a throughput slow-down after this change, he should use sharding to make better use of the available cores by splitting the indexing workload upfront.

Environment

None

Status

Assignee

Sanne Grinovero

Reporter

Sanne Grinovero

Components

Fix versions

Priority

Major