flink-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Christophe Jolif <cjo...@gmail.com>
Subject Re: Kafka and parallelism
Date Mon, 05 Feb 2018 10:42:16 GMT
Thanks. It helps indeed.

I guess the last point it does not explicitly answer is "does just creating
a kafka consumer reading from multiple partition set the parallelism to the
number of partitions". But reading between the lines I think this answer is
clearly no. You have to set your parallelism yourself and then it will
round robin between them.

Thanks again,

On Mon, Feb 5, 2018 at 9:52 AM, Tzu-Li (Gordon) Tai <tzulitai@apache.org>

> Hi Christophe,
> You can set the parallelism of the FlinkKafkaConsumer independently of the
> total number of Kafka partitions (across all subscribed streams, including
> newly created streams that match a subscribed pattern).
> The consumer deterministically assigns each partition to a single consumer
> subtask, in a round-robin fashion.
> E.g. if the parallelism of your FlinkKafkaConsumer is 2, and there is 6
> partitions, each consumer subtask will be assigned 3 partitions.
> As for topic pattern subscription, FlinkKafkaConsumers starting from
> version 1.4.0 support this feature. You can take a look at [1] on how to do
> that.
> Hope this helps!
> Cheers,
> Gordon
> [1] https://ci.apache.org/projects/flink/flink-docs-
> release-1.4/dev/connectors/kafka.html#kafka-consumers-
> topic-and-partition-discovery
> On 3 February 2018 at 6:53:47 PM, Christophe Jolif (cjolif@gmail.com)
> wrote:
> Hi,
> If I'm sourcing from a KafkaConsumer do I have to explicitly set the Flink
> job parallelism to the number of partions or will it adjust automatically
> accordingly? In other word if I don't call setParallelism will get 1 or the
> number of partitions?
> The reason I'm asking is that I'm listening to a topic pattern not a
> single topic and the number of actual topic (and so partitions) behind the
> pattern can change so it is not possible to know ahead ot time how many
> partitions I will get.
> Thanks!
> --
> Christophe

View raw message