[
https://issues.apache.org/jira/browse/FLINK-13324?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16892533#comment-16892533
]
Yang Wang commented on FLINK-13324:
-----------------------------------
[~Tison]
Sorry for late.
# The configuration element <optional>true</optional> is to avoid transitive
dependencies. Such as, flink-s3-fs-base -> flink-hadoop-fs ->
flink-shaded-hadoop-2, because of optional, flink-shaded-hadoop-2 is not
included in flink-s3-fs-base.
# I think flink-yarn depends on flink-shaded-hadoop-2 is expected, we need the
classes to submit flink job to Yarn cluster. Of course you could set
flink-shaded-hadoop-2 in flink-yarn as optional. However, by now all the module
that depends on flink-yarn has excluded flink-shaded-hadoop-2. It is not
necessary.
AFAI, the dependencies of hadoop are all shaded in flink-shaded-hadoop-2 jar.
So we do not need to worry about class conflicts.
> flink-shaded-hadoop could exclude curator dependency
> ----------------------------------------------------
>
> Key: FLINK-13324
> URL: https://issues.apache.org/jira/browse/FLINK-13324
> Project: Flink
> Issue Type: Improvement
> Components: BuildSystem / Shaded
> Reporter: TisonKun
> Priority: Minor
> Attachments: patch.diff
>
>
> Flink's high-availability functionality depends on curator, which means flink
> has its own dependent curator version(now is "2.12.0"). However, there are
> several transitive dependencies referring to
> curator-(client|framework|recipe)-2.7.1 from flink-shaded-hadoop(2). We might
> exclude these dependencies for a single curator version used in flink.
> cc [~Zentol]
--
This message was sent by Atlassian JIRA
(v7.6.14#76016)
