Github user tumativ commented on a diff in the pull request:
https://github.com/apache/zookeeper/pull/689#discussion_r231718078
--- Diff:
zookeeper-server/src/main/java/org/apache/zookeeper/server/watch/WatcherCleaner.java
---
@@ -50,6 +50,8 @@
private volatile boolean stopped = false;
private final Object cleanEvent = new Object();
+ private final Object produserAndConsumerLock = new Object();
--- End diff --
The producer of the dead watcher is trying to get the lock on
totalDeadWatchers when totalDeadWatchers reaches the maximum. The expectation
is the consumer needs to reduce this count .totalDeadWatchers is AtomicInteger
and getting a lock on AtomicInteger is a not a good idea as it has their own
concurrency control mechanisms. It will basically confuse why the lock is
required on totalDeadWatchers at producer level when we are expecting consumer
should reduce this count and it has their own concurrency control mechanisms
---
