[
https://issues.apache.org/jira/browse/KAFKA-4822?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15890130#comment-15890130
]
Giri commented on KAFKA-4822:
-----------------------------
i understand that the sync can be achieved by the above workaround, but still
another thread will be spawned per producer to send the data in background.
i have a case where a predefined set of threads (40 to 80 depending on the
machine conf) receive data from a persistent medium and this data is sent to
kafka (1 producer per thread) and i have to commit the position of the
persistent medium to protect against restarts. I have achieved this in 0.8 sync
producer by storing the data from persistent medium in list and after some
content is cached i sent to kafka as batch and committed the position in the
persistent medium. But now i do not have explicit control over when the data
will be sent to kafka as data is completely handled by the new network thread
and also the batch full and new batch are not visible to user (lost at
KafkaProducer.doSend()).
> Kafka producer implementation without additional threads, similar to sync
> producer of 0.8.
> ------------------------------------------------------------------------------------------
>
> Key: KAFKA-4822
> URL: https://issues.apache.org/jira/browse/KAFKA-4822
> Project: Kafka
> Issue Type: New Feature
> Components: producer
> Affects Versions: 0.9.0.0, 0.9.0.1, 0.10.0.0, 0.10.0.1, 0.10.1.0, 0.10.1.1
> Reporter: Giri
> Priority: Minor
>
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
