Hello,

I noticed a possible concurrency issue related to DefaultReactiveExecutor
and MultiCastReactiveTask in which random warnings appear causing the kafka
consumers which initiate the routes to die after a poll timeout.

This issue started manifesting with camel 4.14 and onwards(also camel 4.15
is affected).

Setup:

We have a kafka consumer route which consumes xml messages that are routed
to a splitter that performs additional transformations on the parts and
then pushes those messages to an outgoing topic.

Split definition:

Custom splitter
Custom aggregator
shareUnitOfWork()
stopOnException()
parallelProcessing()
executorService(Executors.newWorkStealingPool())

Warnings:

Error executing reactive work due to null. This exception is ignored.
Error executing reactive work due to Index 29 out of bounds for length 20.
This exception is ignored.

Apparently, the route stalls (split definition is inside a
try-catch-finally, but the finally never triggers, as that is the spot
where we perform the manual commit).

Also, the warn is not descriptive enough, as the exception stacktrace is
not logged.

Issue is happening only for a couple of topics with relatively low
throughput(we have other ones with a lot higher throughput).

Thanks,

Branco

Reply via email to