[
https://issues.apache.org/jira/browse/HADOOP-11656?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14347217#comment-14347217
]
Jason Lowe commented on HADOOP-11656:
-------------------------------------
bq. There are plenty of ways we can make the transition easier for downstream
folks. I've already mentioned giving upgrade docs that include maven pom
changes needed to get the same set of dependencies. As you mention, we could
also include some option toggle that says "I want to see the framework
libraries." I happen to think this is a bad idea because it leads straight back
to where we are now. In any case, either of these mitigations require
downstream projects to change what they are doing, which sounds incompatible to
me.
I think the idea here is to flip the defaults around. The easiest transition
for existing downstream folks is to opt in, rather than opt out, of classpath
isolation. We can debate whether that's custom classloaders, OSGi packaging,
or what-not when it's turned on. But if not turned on by default then it is
backwards compatible, to the extent that we support backwards compatibility
today. Clients/jobs that ran before continue to run on the new version. Those
that want/need the isolation can ask for it, and we can iterate the isolation
feature without necessarily breaking the existing users that aren't asking for
it because it didn't exist back then and would break their old workflow if it
suddenly does. At some point in the future we can (and probably want to)
switch the defaults so clients/apps get classpath isolation by default. I
totally agree that decision necessarily breaks backwards compatibility.
IMHO the smoothest transition for major features, this or otherwise, is to
develop the feature if possible as opt in, rather than opt out, until it is
mature, stable, and the community agrees it should be active by default. Some
features are such that they inherently cannot be turned off, but if possible
it'd be great to develop and mature them as options that people can try out
until they become stable to ease transitions and avoid unnecessary breakage at
an early stage.
> Classpath isolation for downstream clients
> ------------------------------------------
>
> Key: HADOOP-11656
> URL: https://issues.apache.org/jira/browse/HADOOP-11656
> Project: Hadoop Common
> Issue Type: New Feature
> Reporter: Sean Busbey
> Assignee: Sean Busbey
> Labels: classloading, classpath, dependencies
>
> Currently, Hadoop exposes downstream clients to a variety of third party
> libraries. As our code base grows and matures we increase the set of
> libraries we rely on. At the same time, as our user base grows we increase
> the likelihood that some downstream project will run into a conflict while
> attempting to use a different version of some library we depend on. This has
> already happened with i.e. Guava several times for HBase, Accumulo, and Spark
> (and I'm sure others).
> While YARN-286 and MAPREDUCE-1700 provided an initial effort, they default to
> off and they don't do anything to help dependency conflicts on the driver
> side or for folks talking to HDFS directly. This should serve as an umbrella
> for changes needed to do things thoroughly on the next major version.
> We should ensure that downstream clients
> 1) can depend on a client artifact for each of HDFS, YARN, and MapReduce that
> doesn't pull in any third party dependencies
> 2) only see our public API classes (or as close to this as feasible) when
> executing user provided code, whether client side in a launcher/driver or on
> the cluster in a container or within MR.
> This provides us with a double benefit: users get less grief when they want
> to run substantially ahead or behind the versions we need and the project is
> freer to change our own dependency versions because they'll no longer be in
> our compatibility promises.
> Project specific task jiras to follow after I get some justifying use cases
> written in the comments.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
