This has big implication on how users configure their memory. Previously, there 
was one buffer pool per state, now there is one buffer pool total. So it 
definitely breaks configurations (lowering overall memory usage, but impacting 
performance).

Is there a concrete reason to add this now, or maybe at some point when we 
revisit the overall RocksDB configuration model?

[ Full content available at: https://github.com/apache/flink/pull/6603 ]
This message was relayed via gitbox.apache.org for [email protected]

Reply via email to