This is not the case in my situation.
I am using an older version where it doesn't have the subscriber model, and
processing messages can occur concurrently in multiple threads.

If I use the KafkaConsumer with multiple threads and partitions -
kafkaConsumer.run(5) //5 thread count -
How can I commit an offset for a specific thread and partition in
kafka-clients 0.9.0.1?

Thanks



Regards,
Amir Zuker

On Tue, Aug 16, 2016 at 1:36 PM, Sudev A C <sudev...@goibibo.com> wrote:

> Hi,
>
> Message object consists of partition, topic, offset and message.
> https://kafka.apache.org/090/javadoc/index.html?org/apache/
> kafka/clients/consumer/KafkaConsumer.html
> You many use this to get current offsets for topic-partition combination.
>
> Thanks
> Sudev
>
> On Tue, Aug 16, 2016 at 2:20 PM, Amir Zuker <a.zu...@gmail.com> wrote:
>
> > Hi everyone,
> >
> > I have a question regarding the 'KafkaConsumer' and its API in regards to
> > committing offsets. (kafka-clients 0.9.0.1)
> >
> > *The scenario -*
> > I am working with auto commit set to disabled because I want to
> implement a
> > retry mechanism and eventually transfer the message to another topic that
> > contains the poison messages.
> > Since I want it to be reliable, I am not using the auto commit and I wish
> > to take control on when that should happen
> >
> > *The implementation detail -*
> > My class that extends 'Runnable' and is created by the KafkaConsumer
> needs
> > to commit the offset once it is done with handling the topic message.
> > However, the API for committing messages is located on the KafkaConsumer
> > with no relation to partition or thread.
> >
> > *The problem -*
> > If I understand correctly, I can use the same KafkaConsumer instance with
> > multiple threads against multiple partitions.
> > If that is the case, how can I commit the offset specific to my
> 'Runnable'
> > instance that just processed a single message without affecting other
> > threads and partitions?
> >
> > Thanks in advance,
> > Amir Zuker
> >
>
>
>
> --
> Thanks
> Sudev A C
> Data Team
>

Reply via email to