Can wskdeploy orchestrates actions as a DAG?

[GitHubDiom]

Hi there,

I want to build a DAG job in openwhisk by using wskdeploy.

I know wskdeploy can create a sequence, but can wskdeploy create a DAG job?

suppose there are 4 DAG functions in a workflow, the relationship of the 4 functions is floowing:

A --> B ——>D
 ↘   ↗
   C

That is, function A sends the same outputs to function B and function C.

Note that functions B and function C can be executed in parallel (for different purposes).

function C receives the output of the function A and sends the output to function D.

Here, function D will not be executed until it receives both inputs from function B and function C.

I think it is a common demand in real-world workflow, could anyone tell me how to do this?

[mrutkows]

@GitHubDiom What you need is extra logic that permits aggregation (and waits for convergence of results). This is not in the domain of wskdeploy, but rather seems to need the features if the Composer tool/language/expressiveness which can be Conducted on the server side (see https://github.com/apache/openwhisk-composer).

These compositions would be expressed using the Composer operands/operations and not using wskdeploy (which is a declarative deployment tool for getting your models created and maintained (and setup with defaults params). The only composition wskdeploy really addresses are simple sequences.

[zacheusz]

this article discusses an extension to OpenWhisk with DAGular: Native DAG Execution in OpenWhisk but I couldn't find the source code