[
https://issues.apache.org/jira/browse/FLUME-2921?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15374648#comment-15374648
]
Mike Percy commented on FLUME-2921:
-----------------------------------
bq. I think we should avoid a custom classloader
It would be a lot of work to implement, and so far we have been able to get
away without doing it. I agree that avoiding it as long as possible is
desirable.
bq. If a user wants to run both ES and HDFS sinks in the same agent, he would
have to shade one of them.
I don't think that is possible. The client JARs for those systems would have to
shade the dependency and ship it. Once there is a com.google.guava import in a
compiled JAR (i.e. the HDFS or ES client JAR), we have to classload classes
that match that package name. If one requires Guava 11, and the other requires
Guava 18, then it simply isn't possible to run them at the same time in the
same classloader. However, the workaround is pretty simple: Run multiple agents.
bq. I think that the best thing to do right now is to block this issue until we
clean our public API.
I agree with you. It's a very sticky problem at this point.
> Support Elasticsearch 2.0+
> --------------------------
>
> Key: FLUME-2921
> URL: https://issues.apache.org/jira/browse/FLUME-2921
> Project: Flume
> Issue Type: Improvement
> Components: Sinks+Sources
> Reporter: Lior Zeno
> Assignee: Lior Zeno
> Fix For: v1.7.0
>
> Attachments: FLUME-2921-0.patch, FLUME-2921-1.patch
>
>
> Elasticsearch sink supports an ancient version of ES. We should make the sink
> work with newer versions of Elasticsearch.
> I attached a patch for that. Please note that this involves upgrading netty
> and guava.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
