Re: StateFun remote/embedded polyglot example

["Tzu-Li (Gordon) Tai" <tz...@apache.org>]

Hi,

On Mon, Jun 1, 2020 at 5:47 AM Omid Bakhshandeh <om...@gmail.com>
wrote:

> Hi,
>
> I'm very confused about StateFun 2.0 new architecture.
>
> Is it possible to have both remote and embedded functions in the same
> deployment?
>

Yes that is possible. Embedded functions simply run within the Flink
StateFun workers (which essentially are Flink TMs).
Remote functions run as standalone services independent of the Flink
StateFun cluster.


> Is there a tutorial that shows the deployment of the two types in the same
> Kubernetes cluster alongside with Flink(possible in Python and Java)?
>

You simply have to include all modules [1] (embedded and remote modules
included) when packaging your StateFun application image [2].
You can see the Dockerfile in [2] that demonstrates exactly that.
It copies an embedded module (a JAR containing the service files and Java
functions / ingresses / egresses) and a remote module (a YAML file defining
the endpoints of remote functions, ingresses / egresses) into a pre-defined
modules directory in the image.


>
> Also, is there a path towards KNative support and scale to zero option?
>

I've cc'ed Igal, who has been looking at KNative support.
As for scale-to-zero, that would already work for your remote function
deployments if you are deploying them with, for example, FaaS solutions
like KNative / AWS Lambda / GCP Cloud Functions, etc.
That is already supported with the new StateFun 2.0 architecture.


>
> Best,
> --
> -------------------------------------------
> Omid
>

Cheers,
Gordon

[1]
https://ci.apache.org/projects/flink/flink-statefun-docs-release-2.0/sdk/modules.html
[2]
https://ci.apache.org/projects/flink/flink-statefun-docs-release-2.0/deployment-and-operations/packaging.html