[ https://issues.apache.org/jira/browse/SPARK-17935?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15612663#comment-15612663
]
Cody Koeninger commented on SPARK-17935:
----------------------------------------
So the main thing to point out is that Kafka producers currently aren't idempotent, so this
sink can't be fault-tolerant.
Regarding the design doc, couple of comments
- KafkaSinkRDD Why is this necessary? Seems like KafkaSink should do basically the same
as existing ForeachSink class
- CachedKafkaProducer Why is this necessary? A singleton producer per JVM is generally what's
recommended by kafka docs.
> Add KafkaForeachWriter in external kafka-0.8.0 for structured streaming module
> ------------------------------------------------------------------------------
>
> Key: SPARK-17935
> URL: https://issues.apache.org/jira/browse/SPARK-17935
> Project: Spark
> Issue Type: Improvement
> Components: SQL, Streaming
> Affects Versions: 2.0.0
> Reporter: zhangxinyu
>
> Now spark already supports kafkaInputStream. It would be useful that we add `KafkaForeachWriter`
to output results to kafka in structured streaming module.
> `KafkaForeachWriter.scala` is put in external kafka-0.8.0.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscribe@spark.apache.org
For additional commands, e-mail: issues-help@spark.apache.org
|