Re: Batch stream Sink delay ?

2017-03-16 Thread Paul Smith
rd timestamp minus a safety margin. - If you use a periodic watermark assigner, the watermark function is periodically called. Best, Fabian 2017-03-16 22:54 GMT+01:00 Paul Smith mailto:psm...@aconex.com>>: I have managed to discover that my presumption of log4j log file being a _guara

Re: Batch stream Sink delay ?

2017-03-16 Thread Paul Smith
-examples/flink-examples-streaming/src/main/java/org/apache/flink/streaming/examples/wordcount/WordCount.java 2017-03-16 12:10 GMT+01:00 Paul Smith mailto:psm...@aconex.com>>: Thanks again for your reply. I've tried with both Parallel=1 through to 3. Same behavior. The log file is m

Re: Batch stream Sink delay ?

2017-03-16 Thread Paul Smith
me order and that all tasks are producing data. Hope this helps, Fabian 2017-03-15 23:54 GMT+01:00 Paul Smith mailto:psm...@aconex.com>>: Thanks Fabian, I’m pretty sure you are correct here. I can see in the Metric view that the currentLowWaterMark is set to MIN_VALUE by the looks of it, s

Re: Batch stream Sink delay ?

2017-03-15 Thread Paul Smith
ot be computed. When an file input is completely processed, Flink sends a Long.MAX_VALUE timestamp which might trigger the execution at the end of the job. I would try to debug the watermarks of your job. The web dashboard provides a few metrics for that. Best, Fabian 2017-03-14 2:47 GMT+01:00 Pa

Batch stream Sink delay ?

2017-03-13 Thread Paul Smith
there's some part of that 5 minutes that would be emitting chunks of summary data to the sink? But nothing turns up until the entire job is done. Maybe the data is just too small.. Maybe there's buffering going on somewhere in the chain. ? Any pointers would be appreciated in understanding the flow here. Cheers, Paul Smith