> In our case they get created exclusively during repairs. Compactionstats > showed a huge number of sstable build compactions
Do you have an indication that at least the disk space is in fact consistent with the amount of data being streamed between the nodes? I think you had 90 -> ~ 450 gig with RF=3, right? Still sounds like a lot assuming repairs are not running concurrently (and compactions are able to run after a repair before the next repair of a neighbor starts). -- / Peter Schuller (@scode on twitter)