Re: Custom TimestampExtractor and FlinkKafkaConsumer082

Hi Aljoscha, I have put together a gist [1] with two classes, a short processing pipeline, which shows the behavior and a data generator to write records into Kafka. I hope I remembered everything we discussed correctly. So basically in the example it works with "TimestampExtractor1" only for par

Re: Watermarks as "process completion" flags

Hi Fabian Defining a special flag for record seems like a checkpoint barrier. I think I will end up re-implementing checkpointing myself. I found the discussion in flink-dev: mail-archives.apache.org/mod_mbox/flink-dev/201511.mbox/…

Re: Cleanup of OperatorStates?

Hi, The sessionid is present in the measurements. It can also be seen as a form of 'browser id'. Most websites use either a 'long lived random value in a cookie' or a 'application session id' for this. So with the id of the browser in hand I have the need to group all events into "periods of acti

Re: Working with the Windowing functionality

In my specific case I can be more accurate about when the next time is when it would make sense to check. Then for all next events I only record a little bit of state about the event stream (i.e. first and last event time). I then only place a new timer in the timer handler and this way I can limit