[
https://issues.apache.org/jira/browse/FLINK-24037?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17406398#comment-17406398
]
Robert Metzger commented on FLINK-24037:
----------------------------------------
> Supporting patters bears the risk of accidentally enabling more plugins after
> an upgrade.
We could fail the entrypoint script, if a pattern matches more than one plugin.
there should anyways be only one plugin jar file per plugin directory.
> Considering that for docker we ship all of flink-dist anyway, do we even need
> optional plugins for that use-case in the first place?
You mean we do with the filesystems what we do already with the metric
reporters, and set them up as plugins in docker?
This could backfire if people have the hadoop s3 plugin, and now get a new
Flink docker image with the hadoop and presto s3 filesystem implementations,
because then s3 will default to the presto implementation.
> Allow wildcards in ENABLE_BUILT_IN_PLUGINS
> ------------------------------------------
>
> Key: FLINK-24037
> URL: https://issues.apache.org/jira/browse/FLINK-24037
> Project: Flink
> Issue Type: Improvement
> Components: flink-docker
> Reporter: Robert Metzger
> Priority: Major
>
> As a user of Flink, I would like to be able to specify a certain default
> plugin, (such as the S3 presto FS) without having to specific the Flink
> version again.
> The Flink version is already specified by the Docker container I'm using.
> If one is using generic deployment scripts, I don't want to put the Flink
> version in two locations.
> Suggested solutions:
> a) Allow wildcards in ENABLE_BUILT_IN_PLUGINS
> b) remove the version string from the jars in the distribution
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
