Slow compactions due to big collection

[onmstester onmstester <on...@zoho.com>]

Hi, 



I'm using cassandra's set, rate of records for a few of partitions went beyond expactations and now there are 100s of thousands of items (each a text with about 60 chars) in the set of those partitions, all of a sudden on nodes that host such big sets, compactions became so slow and there are increasing number of pending compactions.

I fixed my app to prevent such cases in future but in regards with already persisted big sets:



Is there any configuration that could fix compaction throughput?

I'm using version 3.11.2, do upgrading to v4 helps solving this problem?



Best Regards