You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@flink.apache.org by Mazen Ezzeddine <ma...@etu.unice.fr> on 2020/09/16 06:50:10 UTC

KeyedCoProcessFunction, processElement1, processElement2, onTimer timeout

Hey all, 

I am using the KeyedCoProcessFunction class in Flink DataStream APIs to
implement a timeout like use case. The scenario is as follows: I have an
input kafka topic and an output Kafka topic, a service reads from the input
topic processes it (for variable amount of time) and then publishes the
response in the output kafka topic.

Now to implement the timeout (must be using Flink datastream APIs), I have a
FlinkKafkaConsumer that reads from the kafka input topic, and another
FlinkKafkaConsumer that reads from the kafka output topic (once processed
and published by the external service). I am connecting the two streams, and
using the processElement1 I am registering a timer and waiting either that
the onTimer method be fired (a timeout is declared), or the processElement2
is fired before and hence I delete the timer and do not declare a timeout.

 In the situation described above can the scenario of reading an element
from the output topic (processElement2 is fired) happen before reading from
the input topic (processElement1 is fired) knowing that the time taken to
process the element by the external service might take seconds before
publishing it to the output topic, is it possible? is that how by design
Flink works, are there any way to force Flink connected streams to operate
based first comes first served. 

In such case what is the best case to implement the timeout functionality as
described above strictly using the Flink DataStream APIs, Any hint please?

Thank you so much.



--
Sent from: http://apache-flink-mailing-list-archive.1008284.n3.nabble.com/

Re: KeyedCoProcessFunction, processElement1, processElement2, onTimer timeout

Posted by Mazen Ezzeddine <ma...@etu.unice.fr>.
OK,  thanks so much David very helpful.

Sorry for any inconvenience.



--
Sent from: http://apache-flink-mailing-list-archive.1008284.n3.nabble.com/