Workflows: tool version is replaced (randomly?) by older version

[bernt-matthias]

Describe the bug

In a workflow (galaxyproject/iwc#321) on my instance the tool version of the map_value tool is sometimes reduced to 0.1.1 (instead of 0.2.0 as defined in the workflow).

This happened a few times when opening the editor (but not always).

And also when I executed planemo workflow_test_init --from_invocation 475029a28781f288 --galaxy_url https://galaxy.intranet.ufz.de --galaxy_user_key XXX

which showed me bioblend WARNING: GET: error 500: b'{"err_msg": "Uncaught exception in exposed API method:", "err_code": 0}', 9 attempts left. In the server logs was:

galaxy.workflow.modules DEBUG 2024-01-11 22:52:15,629 [pN:main.4,p:629336,tN:WSGI_1] <span title="tool id 'toolshed.g2.bx.psu.edu/repos/iuc/map_param_value/map_param_value/0.2.0'">Using version '0.1.1' instead of version '0.2.0' specified in this workflow.
galaxy.web.framework.decorators ERROR 2024-01-11 22:52:15,687 [pN:main.4,p:629336,tN:WSGI_1] Uncaught exception in exposed API method:
Traceback (most recent call last):
  File "/gpfs1/data/galaxy_server/galaxy/lib/galaxy/web/framework/decorators.py", line 337, in decorator
    rval = func(self, trans, *args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/gpfs1/data/galaxy_server/galaxy/lib/galaxy/webapps/galaxy/api/workflows.py", line 385, in workflow_dict
    ret_dict = self.workflow_contents_manager.workflow_to_dict(
               ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/gpfs1/data/galaxy_server/galaxy/lib/galaxy/managers/workflows.py", line 843, in workflow_to_dict
    wf_dict = self._workflow_to_dict_export(trans, stored, workflow=workflow)
              ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/gpfs1/data/galaxy_server/galaxy/lib/galaxy/managers/workflows.py", line 1386, in _workflow_to_dict_export
    tool_state = module.get_export_state()
                 ^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/gpfs1/data/galaxy_server/galaxy/lib/galaxy/workflow/modules.py", line 320, in get_export_state
    return self.get_state(nested=True)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/gpfs1/data/galaxy_server/galaxy/lib/galaxy/workflow/modules.py", line 315, in get_state
    return self.state.encode(Bunch(inputs=inputs), self.trans.app, nested=nested)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/gpfs1/data/galaxy_server/galaxy/lib/galaxy/tools/__init__.py", line 678, in encode
    value = params_to_strings(tool.inputs, self.inputs, app, nested=nested)
            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/gpfs1/data/galaxy_server/galaxy/lib/galaxy/tools/parameters/__init__.py", line 266, in params_to_strings
    value = params[key].value_to_basic(value, app, use_security=use_security)
            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/gpfs1/data/galaxy_server/galaxy/lib/galaxy/tools/parameters/grouping.py", line 753, in value_to_basic
    current_case = rval["__current_case__"] = self.get_current_case(value[self.test_param.name])
                                              ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/gpfs1/data/galaxy_server/galaxy/lib/galaxy/tools/parameters/grouping.py", line 746, in get_current_case
    raise ValueError("No case matched value:", self.name, str_value)
ValueError: ('No case matched value:', 'input_param_type', 'data')

Galaxy Version and/or server at which you observed the bug
Galaxy Version: 23.1
Commit: 0e3a0ca

[mvdbeek]

Unclear how that is a bug ? If you don't have the tool installed what else are we supposed to do ?

[bernt-matthias]

Both versions are installed. But might be related to #17272 (comment)

On my instance I do not do a restart Galaxy after a tool installation. Maybe some "components" (handlers?) still have an old tool set.

Will try a restart.

[bgruening]

I also see that from time to time when new tools get installed and web handlers have not been restarted.

[bernt-matthias]

Windows style: Restart "solved" it. Question is, if this is expected.

[mvdbeek]

If everything is set up correctly and your processes are listening to the internal task queue, no. I've yet to see a reproduction. Whatever the problem, it's not related to workflows.

[mvdbeek]

and you can check that the task queue listening works with e.g. curl -X 'PUT' $GALAXY_URL -H "x-api-key: $API_KEY and checking that every single process has reloaded the toolbox.