This closes #5001.

This closes #4748

This closes #5061

This closes #5042.

Update to the latest checkstyle version and fix the errors not
previously detected.

This closes #5044.

This close #4794.

This closes #5014.

Before, we had it in places that require it. This doesn't work when
running mvn javadoc:aggregate because this will only run for the root
pom and can then not find the "bundle" dependencies.

genjavadoc generated some Java code that was making Javadoc fail.

Disable it in most modules.

This closes #4972.

This closes #4971.

Another set of RAT exclusions to prevent errors on Windows.

Avro version is used multiple times (dependendies and plugins),
having a variable makes sure we use those in sync.

This closes #4947.

This also adds a new test that verifies that we correctly register
Avro Serializers when they are present and modifies an existing test to
verify that we correctly register dummy classes.

Excluding Avro in Hadoop will make basic classes like sequence file writers fail
to load, even if Avro is not used.

Shading Avro is not possible because it is part of the API for users that
use Avro types.

  - Rename the 'flink-shaded-curator-recipes' module to 'flink-shaded-curator',
    because it actually contains more curator code than just the recipes.

  - Move the exception handling logic of 'ZooKeeperAccess' directly into the
    ZooKeeperStateHandleStore

This closes #4927

Shading the ZooKeeper dependency makes sure that this specific version of
ZooKeeper is used by the Flink runtime module. The ZooKeeper version is
sensitive, because we depend on bug fixes in later ZooKeeper versions
for Flink's high availability.

This prevents situations where for example a set of added dependencies (for
example transtive dependencies of Hadoop) cause a different ZooKeeper version
to be in the classpath and be loaded.

This commit also removes the 'flink-shaded-curator' module, which was originally
created to shade guava within curator, but is now obsolete, because newer
versions of curator shade guava already.

This closes #4923.

The QS module is split into core and client. The core should
be put in the lib folder to enable queryable state, while the
client is the one that the user will program against. The
reason for the restructuring in mainly to remove the dependency
on the flink-runtime from the user's program.

* Do not throw exception when same metric is added twice
* Add possibility to configure port range
* Bump prometheus.version 0.0.21 -> 0.0.26
* Use simpleclient_httpserver instead of nanohttpd
* guard gauge report against null
* guard close() vs NPE

This closes #4586.

This closes #4867.

In accordance with the Apache Software License 2.0, the NOTICE
file is the dedicated/prefered place for any additional license
information that affects redistribution, because the NOTICE
file is required to be included in any redistribution.

The only project still having shaded dependencies that require
extra license statements is flink-table. Changing the licenses
to be in the NOTICE file also allows removing the maven-resource
plugin from the root pom.xml

jlnie is a command line shell used optionally for ZK's
command line tool.

We can do this now that we dropped support for Scala 2.10.

This closes #4807

Creates a separate for the queryable state and  moves the client
code outside the runtime. The Task Manager is now instantiating
the KvStateServer using reflection.

  - Set shorter heartbeats intervals. Default pause value of 60seconds is
    too large (tests would timeout before akka react)

  - Exclude netty dependency from zookeeper. Zookeeper was pulling in
    conflicting Netty version. Conflict was extremly subtle - TaskManager in
    Kafka tests was deadlocking in some rare corner cases.

This closes #4775

This changes the discovery mechanism of file from static class name configurations
to a service mechanism (META-INF/services).

As part of that, it factors HDFS and MapR FS implementations into separate modules.

With this change, users can add new filesystem implementations and make them available
by simply adding them to the class path.

This also makes them optional in flink-runtime, which is enabled by the
previous changes to only use Hadoop dependencies if they are available.

This also requires adding a few explicit dependencies in other modules
because they were using transitive dependencies of the Hadoop deps. The
most common dependency there is, ha!, commons-io.

This also adds an end-to-end test that verifies correct order for both
classes and resources.