optimal deployment model for Flink Streaming programs

[Rico Bergmann]

Hi!

We want to build an infrastructure for automated deployment of Flink
Streaming programs to a dedicated environment. This includes automated
tests (unit and integration) via Jenkins and in case of a successful
build&test the program should be deployed to the execution environment.

Since streaming programs run infinitely, the problem is to switch from
the running program to the newly deployed. The CLI has some features
that would make it possible (list, cancel). Is there another way of
somehow restarting a streaming program? Do you have a suggested way for
the deployment (automated!)?

Regards
Rico B.

[Stephan Ewen]

I think what you outline is the right way to go for the time being.

The "Client" class is being reworked to bet more of these deployment/controlling methods, to make it easier to deploy/cancel/restart jobs programatically, but that will probably take a few more weeks to converge.

Stephan

On Thu, Jul 30, 2015 at 8:59 AM, Dipl.-Inf. Rico Bergmann <[hidden email]> wrote:

Hi!

We want to build an infrastructure for automated deployment of Flink Streaming programs to a dedicated environment. This includes automated tests (unit and integration) via Jenkins and in case of a successful build&test the program should be deployed to the execution environment.

Since streaming programs run infinitely, the problem is to switch from the running program to the newly deployed. The CLI has some features that would make it possible (list, cancel). Is there another way of somehow restarting a streaming program? Do you have a suggested way for the deployment (automated!)?

Regards
Rico B.