Hello,
I found possible deadlock in AbstractCommitLogSegmentManager. The root
cause is incorrect use of LockSupport.park/unpark pair. Unpark should be
invoked only if caller is sure that thread was parked in appropriate place.
Otherwice permission given by calling unpark can be consumed by other
Hello,
Could someone provide feedback (and possible merge) my PR [1]
for CASSANDRA-13444?
I rewrote cpu-critical part of Cassandra. Merging this PR will reduce cpu
and GC consumtion produced by Compaction threads.
You can find more details in issue description [2].
Thanks.
Feodor Bobin
[1]