Details
-
Improvement
-
Status: Resolved
-
Low
-
Resolution: Fixed
-
None
-
None
Description
When GC begins taking longer and longer, it is often a sign that the JVM is going OOM, but it is not clear why. Often, users swamp the MDP by either writing at CL.ZERO or issuing many read requests while the machine is block on disk IO. To more easily diagnose this, we should log the threadpool stats when we see a GC take too long.