Remote module interface validation error

[Klafyvel]

Hi, I am using a remote hardware module accessed from an interfuse. The remote module implements the ProcessControlInterface, and when the interfuse that connects to it is on the same machine everything works fine. However, when I try to use it on the remote I get the following error:

RuntimeError: Module "<qudi.hardware.laser.matisse.MatisseCommander(0x1ddd8cc1ef0) at 0x000001DDD9B328C0>" connected to connector "matisse" does not implement interface "ProcessControlInterface". It implements {'module', 'object'}. 
Traceback (most recent call last): 
 
  File "C:\Users\Molecola-II-Lab66\.conda\envs\qudi-env\lib\site-packages\qudi\core\modulemanager.py", line 201, in activate_module 
    self._modules[module_name].activate() 
 
  File "C:\Users\Molecola-II-Lab66\.conda\envs\qudi-env\lib\site-packages\qudi\core\modulemanager.py", line 505, in activate 
    self._connect() 
 
  File "C:\Users\Molecola-II-Lab66\.conda\envs\qudi-env\lib\site-packages\qudi\core\modulemanager.py", line 732, in _connect 
    self._instance.connect_modules(module_connections) 
 
  File "C:\Users\Molecola-II-Lab66\.conda\envs\qudi-env\lib\site-packages\qudi\core\module.py", line 408, in connect_modules 
    conn.connect(target) 
 
  File "C:\Users\Molecola-II-Lab66\.conda\envs\qudi-env\lib\site-packages\qudi\core\connector.py", line 90, in connect 
    raise RuntimeError(

For reference, here is the relevant configuration on the server, with a dummy example of how it works locally:

global: 
    # list of modules to load when starting 
    startup_modules: [] 
 
    # Module server configuration for accessing qudi GUI/logic/hardware modules from remote clients 
    remote_modules_server: 
        address: "an address" 
        port: a port 
    # Server port for serving the active qudi module namespace locally (localhost). 
    # Used by e.g. the Qudi jupyter kernel. 
    namespace_server_port: 18861 
 
    # If this flag is set (True), all arguments passed to qudi module APIs from remote 
    # (jupyter notebook, qudi console, remote modules) will be wrapped and passed "per value" 
    # (serialized and de-serialized). This is avoiding a lot of inconveniences with using numpy in 
    # remote clients. 
    # If you do not want to use this workaround and know what you are doing, you can disable this 
    # feature by setting this flag to False. 
    force_remote_calls_by_value: True 
 
    # Qss stylesheet for controlling the appearance of the GUIs. 
    # Absolute path or relative to qudi.artwork.styles 
    stylesheet: 'qdark.qss' 
 
    # Default root directory for measurement data storage. All eventual data sub-directories should 
    # be contained within this directory. This is not enforced, just convention. 
    # The fallback directory is <user home>/qudi/Data/ 
    default_data_dir: "C:/Users/NanoP/qudi/data" 
 
    # Save data to daily data sub-directories by default 
    daily_data_dirs: True 
gui:
    matisse_dummy: 
        module.Class: "excitation_scanner.excitation_gui.ScanningExcitationGui" 
        connect: 
            excitation_logic: matisse_excitation_logic 
logic: 
    matisse_excitation_logic: 
        module.Class: "scanning_excitation_spectroscopy.ScanningExcitationLogic" 
        connect: 
            scanner: matisse_scanner
hardware: 
    matisse_scanner: 
        module.Class: "interfuse.remote_matisse_scanner.RemoteMatisseScanner" 
        connect: 
            input: sampler_dummy 
            matisse: matisse_laser 
            matisse_sw: matisse_laser 
        options: 
            input_channel: "APD counts" 
    sampler_dummy: 
        module.Class: "dummy.finite_sampling_input_dummy.FiniteSamplingInputDummy" 
        options: 
            simulation_mode: LORENTZ 
    matisse_laser: 
        module.Class: "laser.matisse.MatisseCommander" 
        allow_remote: True 
        options: 
            address: "127.0.0.1" 
            port: 30000 

And here is the relevant configuration on the client:

global:
    # list of modules to load when starting
    startup_modules: []

    # Module server configuration for accessing qudi GUI/logic/hardware modules from remote clients
    remote_modules_server:
        address: 'localhost'
        port: 12345

    # Server port for serving the active qudi module namespace locally (localhost).
    # Used by e.g. the Qudi jupyter kernel.
    namespace_server_port: 18861

    # If this flag is set (True), all arguments passed to qudi module APIs from remote
    # (jupyter notebook, qudi console, remote modules) will be wrapped and passed "per value"
    # (serialized and de-serialized). This is avoiding a lot of inconveniences with using numpy in
    # remote clients.
    # If you do not want to use this workaround and know what you are doing, you can disable this
    # feature by setting this flag to False.
    force_remote_calls_by_value: True

    # Qss stylesheet for controlling the appearance of the GUIs.
    # Absolute path or relative to qudi.artwork.styles
    stylesheet: 'qdark.qss'

    # Default root directory for measurement data storage. All eventual data sub-directories should
    # be contained within this directory. This is not enforced, just convention.
    # The fallback directory is <user home>/qudi/Data/
    default_data_dir: "I:/Other computers/Spectro2021/Lab66_spectrometer_2021/DataLab37"

    # Save data to daily data sub-directories by default
    daily_data_dirs: True
gui:
    matisse_excitation_scanner:         
        module.Class: "excitation_scanner.excitation_gui.ScanningExcitationGui" 
            connect:
                excitation_logic: matisse_excitation_logic 
logic:
    matisse_excitation_logic:
        module.Class: "scanning_excitation_spectroscopy.ScanningExcitationLogic" 
        connect:
            scanner: matisse_scanner 
hardware:
    ni_finite_sampling_input:
        module.Class: 'ni_x_series.ni_x_series_finite_sampling_input.NIXSeriesFiniteSamplingInput'
        options:
            device_name: 'Dev1'
            digital_channel_units:  # optional
                pfi1: 'c/s'
                pfi2: 'c/s'
            # external_sample_clock_source: 'PFI0'  # optional
            # external_sample_clock_frequency: 1000  # optional
            adc_voltage_range: [-10, 10]  # optional, default [-10, 10]
            max_channel_samples_buffer: 10000000  # optional, default 10000000
            read_write_timeout: 10  # optional, default 10
            sample_clock_output: '/Dev1/PFI0'  # optiona
l            trigger_edge: RISING  # optional
    matisse_laser:
        native_module_name: "matisse_laser"
        address: "address of the server"
        port: port of te server 
    matisse_scanner:
         module.Class: "interfuse.remote_matisse_scanner.RemoteMatisseScanner" 
        connect:
            input: ni_finite_sampling_input 
            matisse: matisse_laser
            matisse_sw: matisse_laser 
        options:
            input_channel: "pfi1"

[timoML]

This is most likely caused by incompatible versions of rpyc on server or client.
We recently changed our setup requirements to "rpyc>=5.0.1, <6" for this reason.

Which versions are installed in your environments?

[Klafyvel]

Ah! that could be the issue. I have 6.0.0 on the client and 6.0.1 on ther server. Can I simply python -m pip install --force-reinstall -v "rpyc>=5.0.1, <6" on both?

[timoML]

We usually recommend to pull the latest qudi-core and run the install script again python -m pip install -e . in the qudi-core directory (assuming dev mode install). Yours will most likely work, too.

[Klafyvel]

Indeed, the command line worked. Thank you very much for your super fast answer. I will try to upgrade qudi as soon as possible to get that fixed.