[
https://issues.apache.org/jira/browse/FLINK-11501?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16762965#comment-16762965
]
Lakshmi Rao commented on FLINK-11501:
-------------------------------------
Adding some screenshots of results from testing the Guava RateLimiter approach
on the Kafka consumer.
!RateLimiting-1.png|width=981,height=305!
!Ratelimiting-2.png|width=981,height=305!
> Add a ratelimiting feature to the FlinkKafkaConsumer
> ----------------------------------------------------
>
> Key: FLINK-11501
> URL: https://issues.apache.org/jira/browse/FLINK-11501
> Project: Flink
> Issue Type: Improvement
> Components: Kafka Connector
> Reporter: Lakshmi Rao
> Assignee: Lakshmi Rao
> Priority: Major
> Attachments: RateLimiting-1.png, Ratelimiting-2.png
>
>
> There are instances when a Flink job that reads from Kafka can read at a
> significantly high throughput (particularly while processing a backlog) and
> degrade the underlying Kafka cluster.
> While Kafka quotas are perhaps the best way to enforce this ratelimiting,
> there are cases where such a setup is not available or easily enabled. In
> such a scenario, ratelimiting on the FlinkKafkaConsumer is useful feature.
> The approach is essentially involves using Guava's
> [RateLimiter|https://google.github.io/guava/releases/19.0/api/docs/index.html?com/google/common/util/concurrent/RateLimiter.html]
> to ratelimit the bytes read from Kafka (in the
> [KafkaConsumerThread|https://github.com/apache/flink/blob/master/flink-connectors/flink-connector-kafka-0.9/src/main/java/org/apache/flink/streaming/connectors/kafka/internal/KafkaConsumerThread.java])
> More discussion here:
> [https://lists.apache.org/thread.html/8140b759ba83f33a22d809887fd2d711f5ffe7069c888eb9b1142272@%3Cdev.flink.apache.org%3E]
>
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
