[SDK] tool for copying CWL file + input collections
|Velocity based estimate||-|
arv-copy can copy workflow records (+ dependencies). However, our current workflow is oriented around CWL files stored in a git repo + input documents with keep references. There is a gap in the tooling to make it easy to copy all the dependencies (reference collections, docker images, input collections) from one cluster to another starting from a git repo of CWL files and input documents.
Suggest an mode for arvados-cwl-runner which specifies one or more "source" Arvados sites. During workflow initialization, if a collection or docker image is not available on the site on which the workflow will run, copy it from one of the "source sites". Arvados-cwl-runner already has similar behavior to upload files and Docker images from the local file system.