NPE when Clean.apply called with null as first argument

[zybreak]

Our Cassandra 5.0.2 cluster is struggling when consuming large amounts of embeddings (>200M) and at some point (a node dies with the following stack trace:

java.lang.NullPointerException: Cannot invoke "java.lang.Boolean.booleanValue()" because "res" is null
	at org.apache.cassandra.utils.memory.MemtableCleanerThread$Clean.apply(MemtableCleanerThread.java:97)
	at org.apache.cassandra.utils.concurrent.ListenerList$CallbackBiConsumerListener.run(ListenerList.java:244)
	at org.apache.cassandra.concurrent.ImmediateExecutor.execute(ImmediateExecutor.java:140)
	at org.apache.cassandra.utils.concurrent.ListenerList.safeExecute(ListenerList.java:166)
	at org.apache.cassandra.utils.concurrent.ListenerList.notifyListener(ListenerList.java:157)
	at org.apache.cassandra.utils.concurrent.ListenerList$CallbackBiConsumerListener.notifySelf(ListenerList.java:250)
	at org.apache.cassandra.utils.concurrent.ListenerList.lambda$notifyExclusive$0(ListenerList.java:124)
	at org.apache.cassandra.utils.concurrent.IntrusiveStack.forEach(IntrusiveStack.java:195)
	at org.apache.cassandra.utils.concurrent.ListenerList.notifyExclusive(ListenerList.java:124)
	at org.apache.cassandra.utils.concurrent.ListenerList.notify(ListenerList.java:96)
	at org.apache.cassandra.utils.concurrent.AsyncFuture.trySet(AsyncFuture.java:104)
	at org.apache.cassandra.utils.concurrent.AbstractFuture.tryFailure(AbstractFuture.java:148)
	at org.apache.cassandra.utils.concurrent.AsyncPromise.tryFailure(AsyncPromise.java:139)
	at org.apache.cassandra.db.memtable.AbstractAllocatorMemtable.lambda$flushLargestMemtable$0(AbstractAllocatorMemtable.java:306)
	at org.apache.cassandra.concurrent.ImmediateExecutor.execute(ImmediateExecutor.java:140)
	at org.apache.cassandra.utils.concurrent.ListenerList.safeExecute(ListenerList.java:166)
	at org.apache.cassandra.utils.concurrent.ListenerList.notifyListener(ListenerList.java:157)
	at org.apache.cassandra.utils.concurrent.ListenerList$RunnableWithExecutor.notifySelf(ListenerList.java:345)
	at org.apache.cassandra.utils.concurrent.ListenerList.lambda$notifyExclusive$0(ListenerList.java:124)
	at org.apache.cassandra.utils.concurrent.IntrusiveStack.forEach(IntrusiveStack.java:195)
	at org.apache.cassandra.utils.concurrent.ListenerList.notifyExclusive(ListenerList.java:124)
	at org.apache.cassandra.utils.concurrent.ListenerList.notify(ListenerList.java:96)
	at org.apache.cassandra.utils.concurrent.AsyncFuture.trySet(AsyncFuture.java:104)
	at org.apache.cassandra.utils.concurrent.AbstractFuture.tryFailure(AbstractFuture.java:148)
	at org.apache.cassandra.concurrent.FutureTask.tryFailure(FutureTask.java:87)
	at org.apache.cassandra.concurrent.FutureTask.run(FutureTask.java:75)
	at java.base/java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1136)
	at java.base/java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:635)
	at io.netty.util.concurrent.FastThreadLocalRunnable.run(FastThreadLocalRunnable.java:30)
	at java.base/java.lang.Thread.run(Thread.java:840)

Which results from the Clean.apply method being called with null for the first argument, which seems to be intentional given the code in ListenerList.java:244, and since the apply function in MemtableCleanerThread.java:97 accepts a Boolean and not a bool primitive, we should handle it being NULL, otherwise the parameter should be of a primitive type to disallow NULL.

When the res parameter is NULL, i treat it as if it is false which seems correct from the context but this needs to be reviewed.

Addresses the immediate issue in: CASSANDRA-20141

[bbotella]

Can you please create a Jira ticket for this?

[zybreak]

@bbotella Yes, just got the account activated so here is a issue connected to this pull request https://issues.apache.org/jira/browse/CASSANDRA-20141