Hey all -
We have a number of scripts that folks run on an ad-hoc basis through ArcMap and/or Pro tools/toolboxes. People and scripts will soon be entirely migrated to Pro and Python3. Those scripts almost always utilize additional Python packages outside the standard install. In the past, we’ve walked each person who needs to run a script through changing their PATH variable, cloning their Python environment (for Pro at least), installing packages through the command line, etc. It works, but is not exactly user-friendly.
We recently implemented Portal and I think I can publish those scripts as web tools that will actually run on the server (even though they would be accessed via Pro on the user’s machine). That would mean I would only need to configure the server with the needed packages and scripts would run seamlessly for users, no matter what user was working from what machine running what script. The process looks pretty straightforward: https://enterprise.arcgis.com/en/server/10.8/publish-services/windows/deploying-custom-python-packag...
However, I do have a question:
The documentation says to “Repeat on each machine in your ArcGIS Server site.” We have 5 servers in our Enterprise deployment for publishing services to (a mix of general use and image servers). If I’m only going to publish web tools to one server, and I'm the only person publishing web tools, do I really need to implement this on all the servers? Or can I just modify the single server I'll be publishing web tool to? Do all the server Python installations need to be in sync for some reason? Or is it just recommended so that web tools can be published to any server? What's the reasoning behind that instructions from Esri?
We're currently on Enterprise 10.8.1 (potentially moving to 11 this winter, but that's very TBD).