[ARTEMIS-3848] High cpu usage on ReadWrite locks

• Type: Bug
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 2.22.0
• Fix Version/s: 2.23.0
• Assignee: Clebert Suconic
• Reporter: Clebert Suconic
• Created: 25/May/22 13:20
• Updated: 08/Jun/22 03:07
• Resolved: 07/Jun/22 13:34

Description

our HandlerBase is setting a boolean value if it's in handler.

I thought by setting it to null I was clearing the ThreadLocal variable, however that's just creating a null entry.

What makes it worst is the fact that threadLocal was non static. as a result we are getting a lot of entries on the ThreadLocal HashMap. as connections come and go, that will generate a leak on the thread.

That will cause threads on the following stack trace to consume a lot of CPU:

at java.lang.ThreadLocal$ThreadLocalMap.expungeStaleEntry(ThreadLocal.java:638)
at java.lang.ThreadLocal$ThreadLocalMap.remove(ThreadLocal.java:520)
at java.lang.ThreadLocal$ThreadLocalMap.access$200(ThreadLocal.java:319)
at java.lang.ThreadLocal.remove(ThreadLocal.java:242)
at java.util.concurrent.locks.ReentrantReadWriteLock$Sync.tryReleaseShared(ReentrantReadWriteLock.java:426)
at java.util.concurrent.locks.AbstractQueuedSynchronizer.releaseShared(AbstractQueuedSynchronizer.java:1341)
at java.util.concurrent.locks.ReentrantReadWriteLock$ReadLock.unlock(ReentrantReadWriteLock.java:881)

Issue Links

• links to: GitHub Pull Request #4098