Sergio Bossa

I think you're referring to the max file size, which only refers to the max size of a _single_ journal file; a journal is made up of several files, so...

> Think: "Journal.WriteType.MANUAL" and if the batch size has been filled, an error is thrown. So you want to only flush by calling Journal#sync ?

Proposed solution: make defrag automatically happen after a given number of updates, to make it deterministic and provide users a way to configure it based on the desired I/O pressure...

The new defrag behaviour is controlled via the new "defrag.op-limit" configuration option, which is going to replace the current "defrag.limit" option (now no-op). The new "defrag.op-limit" default value will be...

The problem is, if that configuration isn't set, and you have more threads than configured hosts, exceeding threads will be timed-out by the Astyanax client pool with PoolTimeoutException, making your...

Any further comments? Did you have the chance to actually try it?

@niksajakovljevic This is ready for review: the first commit has been rebased to the latest `main-ucu`, the second commit is some new stuff needed by CNDB.

> we should properly mark constructors for testing Done it based on your previous comments but forgot to push :man_facepalming:, here you are: https://github.com/datastax/cassandra/pull/1148/commits/d4b6a3d3f255f87aa02f197e45de484086934b35