#Impact client 1.11 upgrade#
#Impact client 1.11 code#

The examples of Dockerfiles and docker image build.sh scripts have been removed from the Flink Github repository. Previous setups that use the start-cluster.sh and stop-cluster.sh scripts need to rename that file. Replaced slave file name with workers # FLINK-18307 #įor Standalone Setups, the file with the worker nodes is no longer called slaves but workers. The benefit is that this change drastically reduces pressure on the JVM’s metaspace. This changes Flink’s recovery behaviour slightly so that it will not reload static fields.

#Impact client 1.11 code#

The user code class loader is being reused by the TaskExecutor as long as there is at least a single slot allocated for the respective job. The only valid option for jobmanager.scheduler is the default value ng.īind user code class loader to lifetime of a slot Hence, setting jobmanager.scheduler: legacy will no longer work and fail with an IllegalArgumentException. Removal of LegacyScheduler # FLINK-15629 #įlink no longer supports the legacy scheduler. There is no need to download manually jar files for flink-csv and flink-json formats as they are now bundled in the lib folder. flink-csv and flink-json are bundled in lib folder Users need to provide Hadoop dependencies through the HADOOP_CLASSPATH environment variable (recommended) or via lib/ folder.Īlso, the include-hadoop Maven profile has been removed. See also comments on related PR.įlink project does not provide any updated “flink-shaded-hadoop-*” jars. The DataSet API that were using sinks like print(), count() or collect() will now throw an exception whileīefore the output was simply never printed. This implies that, for instance, jobs based on The same behavior as submitting a job through the CLI in detached mode. With FLINK-16657 the web submission logic changes and it exposes Web Submission behaves the same as detached mode.

This allows for more light-weight application submission. The user can now submit applications and choose to execute their main() method on the cluster rather than the client.

#Impact client 1.11 upgrade#

These notes carefully if you are planning to upgrade your Flink version to 1.11. Or dependencies, that changed between Flink 1.10 and Flink 1.11. These release notes discuss important aspects, such as configuration, behavior,

Hadoop MapReduce compatibility with Flink.

Conversions between Table and DataStream.

Conversions between PyFlink Table and Pandas DataFrame.