[
https://issues.apache.org/jira/browse/FLINK-3294?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15356856#comment-15356856
]
Jens Kat commented on FLINK-3294:
---------------------------------
[~rmetzger] I've been looking into the Kafka09Fetcher and this one implements
from AbstractFetcher: public void
commitSpecificOffsetsToKafka(Map<KafkaTopicPartition, Long> offsets)
Using the consumer, however the 08Fetcher implements this using
zkHandler.writeOffsets(offsets)
It seems to me that the initial behavior already is in place? I'll look into it
to get this working.
> KafkaConsumer (0.8) commit offsets using SimpleConsumer.commitOffsets()
> -----------------------------------------------------------------------
>
> Key: FLINK-3294
> URL: https://issues.apache.org/jira/browse/FLINK-3294
> Project: Flink
> Issue Type: Bug
> Components: Kafka Connector
> Reporter: Robert Metzger
> Assignee: Robert Metzger
>
> Currently, the 0.8 consumer for Kafka is committing the offsets manually into
> Zookeeper so that users can track the lag using external tools.
> The 0.8 consumer has a pluggable design, and this component is easily
> pluggable.
> Since OffsetCommitRequest version=1 (supported in 0.8.2 or later), users can
> choose between two offset commit modes:
> a) Let the broker commit into ZK (this is what we are doing from the consumer
> b) Let the broker commit the offset into a special topic.
> By adding a different "OffsetHandler" backend, users can commit offsets from
> the brokers (reducing the total number of ZK connections) or into the
> broker's offset topic.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
