Dispatcher logs ArrayIndexOutOfBoundsExceptions

[JackSilk]

Describe the bug
The logs for the Kafka broker dispatcher show a lot of ArrayIndexOutOfBoundsExceptions.

Apr 14, 2021 1:41:08 PM io.vertx.core.impl.ContextImpl
SEVERE: Unhandled exception
java.lang.ArrayIndexOutOfBoundsException: Index 1 out of bounds for length 1
at dev.knative.eventing.kafka.broker.dispatcher.consumer.impl.UnorderedOffsetManager$OffsetTracker.recordNewOffset(UnorderedOffsetManager.java:207)
at dev.knative.eventing.kafka.broker.dispatcher.consumer.impl.UnorderedOffsetManager.recordDiscarded(UnorderedOffsetManager.java:109)
at dev.knative.eventing.kafka.broker.dispatcher.RecordDispatcher.onFilterNotMatching(RecordDispatcher.java:153)
at dev.knative.eventing.kafka.broker.dispatcher.RecordDispatcher.lambda$onRecordReceived$1(RecordDispatcher.java:137)
at io.vertx.core.impl.future.FutureBase.emit(FutureBase.java:96)
at io.vertx.core.impl.future.SucceededFuture.onSuccess(SucceededFuture.java:61)
at dev.knative.eventing.kafka.broker.dispatcher.RecordDispatcher.onRecordReceived(RecordDispatcher.java:132)
at dev.knative.eventing.kafka.broker.dispatcher.RecordDispatcher.apply(RecordDispatcher.java:116)
at dev.knative.eventing.kafka.broker.dispatcher.RecordDispatcher.handle(RecordDispatcher.java:126)
at dev.knative.eventing.kafka.broker.dispatcher.RecordDispatcher.handle(RecordDispatcher.java:40)
at io.vertx.kafka.client.consumer.impl.KafkaConsumerImpl.lambda$handler$1(KafkaConsumerImpl.java:81)
at io.vertx.kafka.client.consumer.impl.KafkaReadStreamImpl.lambda$null$10(KafkaReadStreamImpl.java:250)
at io.vertx.core.impl.EventLoopContext.emit(EventLoopContext.java:52)
at io.vertx.core.impl.DuplicatedContext.emit(DuplicatedContext.java:194)
at io.vertx.core.impl.AbstractContext.emit(AbstractContext.java:49)
at io.vertx.kafka.client.consumer.impl.KafkaReadStreamImpl.lambda$tracedHandler$11(KafkaReadStreamImpl.java:247)
at io.vertx.kafka.client.consumer.impl.KafkaReadStreamImpl.run(KafkaReadStreamImpl.java:237)
at io.vertx.kafka.client.consumer.impl.KafkaReadStreamImpl.lambda$schedule$8(KafkaReadStreamImpl.java:192)
at io.vertx.core.impl.AbstractContext.dispatch(AbstractContext.java:96)
at io.vertx.core.impl.AbstractContext.dispatch(AbstractContext.java:59)
at io.vertx.core.impl.EventLoopContext.lambda$runOnContext$0(EventLoopContext.java:40)
at io.netty.util.concurrent.AbstractEventExecutor.safeExecute(AbstractEventExecutor.java:164)
at io.netty.util.concurrent.SingleThreadEventExecutor.runAllTasks(SingleThreadEventExecutor.java:472)
at io.netty.channel.nio.NioEventLoop.run(NioEventLoop.java:500)
at io.netty.util.concurrent.SingleThreadEventExecutor$4.run(SingleThreadEventExecutor.java:989)
at io.netty.util.internal.ThreadExecutorMap$2.run(ThreadExecutorMap.java:74)
at io.netty.util.concurrent.FastThreadLocalRunnable.run(FastThreadLocalRunnable.java:30)
at java.base/java.lang.Thread.run(Unknown Source)

Expected behavior
No ArrayIndexOutOfBoundsExceptions

To Reproduce
Steps to reproduce the behavior.
Put a large-ish volume of events (1000s) through a broker + trigger, particularly if some of those

Knative release version
0.22

Additional context
UnorderedOffsetManager seems to try and make the array large enough to avoid this index out of bounds exception. I think it is just an off-by-one error - perhaps the check should be less-than-or-equal?
private void checkAcksArraySize(int blockIndex) { if (this.uncommitted.length < blockIndex) { // Let's make sure we create enough room for more unordered records this.uncommitted = Arrays.copyOf(this.uncommitted, (blockIndex + 1) * 2); } }

[slinkydeveloper]

Yep I think that's the issue, it's strange that unit tests didn't caught it...

[slinkydeveloper]

/assign

[slinkydeveloper]

We should get a new 0.22 release that fixes this bug, thanks for the help!

[JackSilk]

Thank you for fixing it :)