Details
-
Improvement
-
Status: Resolved
-
Normal
-
Resolution: Won't Fix
-
None
-
None
-
None
Description
Currently the rpc max threads defaults to unlimited. The dangers of doing that in certain circumstances were discussed on CASSANDRA-8116 and a warning was added to the yaml.
I've worked with two clusters in the last couple of weeks that has had the heap full of rpc threads for this reason. I'm wondering if we should set a more reasonable default so that people don't hurt themselves unnecessarily.
Is there a reason to not default this to perhaps 2048, what's commented out in the yaml? We could have a yaml comment about it as well.