On Wed, Jul 4, 2018 at 10:06 AM Thomas Weise <t...@apache.org> wrote:
> * The gradle task to run job server takes really long to start the server > (> 1 minute on my macbook). Every restart appears to occur the same > penalty, regardless of whether there are changes or not. Following can be > used to just start the server: > > java -jar > ./runners/flink/job-server/build/install/beam-runners-flink_2.11-job-server-shadow/lib/beam-runners-flink_2.11-job-server-2.6.0-SNAPSHOT-shaded.jar > "--job-host=localhost:8099" "--artifacts-dir=/tmp/flink-artifacts" > > Upon further inspection, the gradle task does the right thing. Once the server is built, a control-C and restart will take only a few seconds. I was able to get the basic streaming translation work for a simple example pipeline [1] on top of changes in [2] Note that currently we are lacking the capability for the runner to retrieve the pipeline options (they are dropped by the Python client). IMO that is something that needs to be addressed for Flink MVP, please see related discussion on the PR. [1] https://gist.github.com/tweise/5b4158fd040a263ddd6bdb885ffc0f3f [2] https://github.com/apache/beam/pull/5888
