You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Technically it would be possible to chain sub-workflows as step of a bigger workflow.
This has to be tested though in the context of weaver.
There is a good chance that data-source resolution (where to dispatch the step process) fails since weaver looks for a remote ADES to run the process. Since in that case the step would be a workflow, it needs to be executed by an EMS.
There is also a change that weaver auto-resolves the default data-source to 'self' and it could then transparently work, but it is not a guarantee.
We should probably consider some package type check that enforce resolution of a sub-workflow to 'self' to avoid unexpected problems.
It is also important to resolve to 'self' instead of "any remote EMS" because the sub-workflow step will need to find all sub-sub-processes to call them. Dispatching a sub-workflow step to another EMS would require to recursively dispatch every underlying process application, which rapidly becomes unmanageable. Also, there is technically no advantage to deploy to a remote EMS since its job should be only to monitor remote jobs, which should be low in terms of resource requirements.
The text was updated successfully, but these errors were encountered:
Technically it would be possible to chain sub-workflows as step of a bigger workflow.
This has to be tested though in the context of weaver.
There is a good chance that data-source resolution (where to dispatch the step process) fails since weaver looks for a remote ADES to run the process. Since in that case the step would be a workflow, it needs to be executed by an EMS.
There is also a change that weaver auto-resolves the default data-source to 'self' and it could then transparently work, but it is not a guarantee.
We should probably consider some package type check that enforce resolution of a sub-workflow to 'self' to avoid unexpected problems.
It is also important to resolve to 'self' instead of "any remote EMS" because the sub-workflow step will need to find all sub-sub-processes to call them. Dispatching a sub-workflow step to another EMS would require to recursively dispatch every underlying process application, which rapidly becomes unmanageable. Also, there is technically no advantage to deploy to a remote EMS since its job should be only to monitor remote jobs, which should be low in terms of resource requirements.
The text was updated successfully, but these errors were encountered: