Re: How to achieve co-location constraints in Flink 1.9.1

[Arvid Heise]

Hi Si-li,

slot sharing is indeed the way that Flink performs co-location. It's
actually enabled by default. It should work as expected if upstream and
downstream operators have the same parallelism.

In certain cases, two operators can be even chained into one task where no
serialization/network traffic happens.

Btw, I also responded to your initial mail to point out an alternative
solution.

On Thu, Nov 19, 2020 at 12:52 PM Si-li Liu  wrote:

> Hi
>
> Flink only have slotSharingGroup API on DataStream class, I can't find any
> public API to achieve co-location constraints. Could anyone provide me an
> example?
>
> Another question is that if I use slotSharing group, Flink will schedule
> two sub tasks to same slot is possible. I think such schedule will always
> success because two tasks run in one slot just decrease the resource usage.
> Could anyone provide me an example when slotSharing group schedule fail ?
>
> Thanks
>
> --
> Best regards
>
> Sili Liu
>


-- 

Arvid Heise | Senior Java Developer



Follow us @VervericaData

--

Join Flink Forward  - The Apache Flink
Conference

Stream Processing | Event Driven | Real Time

--

Ververica GmbH | Invalidenstrasse 115, 10115 Berlin, Germany

--
Ververica GmbH
Registered at Amtsgericht Charlottenburg: HRB 158244 B
Managing Directors: Timothy Alexander Steinert, Yip Park Tung Jason, Ji
(Toni) Cheng

How to achieve co-location constraints in Flink 1.9.1

[Si-li Liu]

Hi

Flink only have slotSharingGroup API on DataStream class, I can't find any
public API to achieve co-location constraints. Could anyone provide me an
example?

Another question is that if I use slotSharing group, Flink will schedule
two sub tasks to same slot is possible. I think such schedule will always
success because two tasks run in one slot just decrease the resource usage.
Could anyone provide me an example when slotSharing group schedule fail ?

Thanks

-- 
Best regards

Sili Liu