Are you asking how the elements from the source are distributed across
the subtasks of the next operation?
That differs a bit across operations; in this case (a map/sink after a
source) AFAIK they are distributed in a round-robin manner.
On 07/10/2019 11:26, Dominik Wosiński wrote:
Thanks a
Thanks a lot for Your answer here Chesnay. I have one more question,
since the `fromElements` is creating the stream with parallelism 1, and I
can see that the env created for my local machine has a default parallelism
== 12. So I assume that the entries from the stream are propagated for the
first
I've filed FLINK-14335 for fixing the java example.
On 07/10/2019 11:08, Chesnay Schepler wrote:
Are you referring to "ExampleIntegrationTest"? If so, then I'm afraid
this test is slightly misleading since the order isn't guaranteed in
this case.
1) As long as the parallelism of the sink is 1
Are you referring to "ExampleIntegrationTest"? If so, then I'm afraid
this test is slightly misleading since the order isn't guaranteed in
this case.
1) As long as the parallelism of the sink is 1 the elements should
arrive in order.
2) The order is maintained if parallelism=1 since elements
Hello,
I have a question, since I am observing quite weird behavior. In the
documentation[1] the example of FlinkMiniCluster usage, shows that we can
expect the results to appear in the same order as they were injected to the
stream by use of *fromElements(). *I mean that Java version of the code i
