Issue is not easily reproducible - sometimes it is hard to capture this behavior, and sometimes it seems to happen on every run. For me it seems it is easiest to spot it with jppf driver and no nodes (so that job never really gets executed), but it can be reproduced also with both driver and node.

client.awaitWorkingConnectionPools(5000)
job = new JPPFJob("Job1")
job.setBlocking(false)
job.add(new JPPFAnnotatedTask(...))
client.submitJob(job)
job.cancel(true)
job.awaitResults()

When running such code, race condition between submitting job to driver and job cancel request makes awaitResult() wait for job to execute (what never happens when there is no node available) rather than returning immediately.

Please take a look at attached client and driver log files. Some debugging pointed out that probably job was canceled while being dispatched; client side cancellation flag was set, but job was still delivered for execution and cancel request was never sent to driver.