ModuleNotFoundError: No module named 'fkie_multimaster_msgs.grpc.file_pb2_grpc'

[simchanu29]

I can't manage to launch node_manager.
Is there a specific way to run it or unlisted dependencies ?

Context :

OS : Ubuntu 20.04
ROS : noetic

Step to reproduce :

1. sudo apt install ros-noetic-fkie-multimaster
2. roscore &
3. rosrun fkie_node_manager node_manager

Error

Traceback (most recent call last):
  File "/opt/ros/noetic/lib/fkie_node_manager/node_manager", line 5, in <module>
    import fkie_node_manager
  File "/opt/ros/noetic/lib/python3/dist-packages/fkie_node_manager/__init__.py", line 45, in <module>
    from fkie_node_manager_daemon import host as nmdhost
  File "/opt/ros/noetic/lib/python3/dist-packages/fkie_node_manager_daemon/__init__.py", line 45, in <module>
    from .server import GrpcServer
  File "/opt/ros/noetic/lib/python3/dist-packages/fkie_node_manager_daemon/server.py", line 41, in <module>
    import fkie_multimaster_msgs.grpc.file_pb2_grpc as fgrpc
ModuleNotFoundError: No module named 'fkie_multimaster_msgs.grpc.file_pb2_grpc'

Troubleshooting

I already tried to fix this issue with :

• installing sudo apt install python3-grpcio python3-grpc-tools
• updating grpcio to 1.33.2
• checking files in /opt/ros/noetic/share/fkie_multimaster_msgs/grpc. I get :

-rw-r--r-- 1 root root  6292 août   3 10:39 file.proto
-rw-r--r-- 1 root root 11837 août   3 10:39 launch.proto
-rw-r--r-- 1 root root  3451 août   3 10:39 monitor.proto
-rw-r--r-- 1 root root  2610 août   3 10:39 screen.proto
-rw-r--r-- 1 root root  2131 août   3 10:39 settings.proto
-rw-r--r-- 1 root root  1879 août   3 10:39 version.proto

• searching on internet for similar issues.
• sourcing /opt/ros/noetic/setup.bash

With no success

I haven't tried yet to use the master branch of this package. I will probably but this is mainly to report the deb on the official repo has an issue.

[atiderko]

This is indeed a problem as the auto-generated files were not included in the Debian package.
I will try to fix it.

In the meantime you have to use the noetic-devel branch to be able to use node manager!

Thank you for reporting!

[tkazik]

Thx for writing that down @simchanu29, I just ran into the same issue.

@atiderko, I was able to build the noetic-devel branch without any warnings.
However, when trying to run the node_manager, I get the following error:
AttributeError: module 'google.protobuf.descriptor' has no attribute '_internal_create_key'

My setup:

• python3-grpcio and python3-grpc-tools are installed
• protoc --version returns: libprotoc 3.6.1

Any ideas why it is failing? Am I missing something? Thx!

[atiderko]

@tkazik : did you tried this solution on stackoverflow.com?

[tkazik]

Thx for the quick reply! Yes, I tried that yesterday without success.

• Using pip install --upgrade protobuf the version gets bumped to 3.14.0
• Now, I upgraded the official protobuf package (which is at version 3.6.1 for focal) using this method.
• Now both pip show protobuf and protoc --version show 3.14.0 and node_manager spawns 🎉

The only thing I now encouter about every second time is the following error:

...
[INFO] [1613065385.776784]: start node manager daemon for http://P1G3:11311/
[WARN] [1613065385.827447]: Error while list_path from grpc://p1g3:12321: <_InactiveRpcError of RPC that terminated with:
	status = StatusCode.UNAVAILABLE
	details = "failed to connect to all addresses"
	debug_error_string = "{"created":"@1613065385.823932252","description":"Failed to pick subchannel","file":"src/core/ext/filters/client_channel/client_channel.cc","file_line":4142,"referenced_errors":[{"created":"@1613065385.823930622","description":"failed to connect to all addresses","file":"src/core/ext/filters/client_channel/lb_policy/pick_first/pick_first.cc","file_line":397,"grpc_status":14}]}"
...

In the 'Log' of the 'Node Manager'-window it says:
[WARN] 11.02.2021 19:04:38.613067 (/node_manager:main_window.py:MainWindow.on_nmd_err:2314): Error while list_path from grpc://p1g3:12321:

However, I don't know whether that is related to protobuf or something else. Have you seen that before?

[atiderko]

did you removed the build files after you updated protobuf and protoc?

Node Manager tries to connect to the node manager daemon which should be started automatically. On errors you see these warnings.

Try rosrun fkie_node_manager_daemon node_manager_daemon to see if there is all ok.

[tkazik]

Yes, I cleaned the build files after updating: Still the same error.
Now, running rosrun fkie_node_manager_daemon node_manager_daemon before node_manager, I observe the following:

1. rosrun fkie_node_manager_daemon node_manager_daemon prints out the following:

...
[INFO] [1613078424.309408]: Create monitor servicer
/usr/lib/python3/dist-packages/psutil/_pslinux.py:1222: RuntimeWarning: ignoring OSError(6, 'No such device or address') for file '/sys/class/hwmon/hwmon6/temp8_input'
  warnings.warn("ignoring %r for file %r" % (err, path),
[INFO] [1613078424.425969]: Create launch manger servicer
...

2. node_manager does not show any warnings anymore (which is good I guess).

Does this help you? Or make sense? Thx for your awesome support!

[atiderko]

this looks good!

The warning only affects the sensor information, which is read out or not.

If you see launch files in launch dock of the node manager than all should be ok.

[tkazik]

Ah ok...guess that is good news then :)
Yes, I see launch files in the dock (=> so I am just ignoring the [WARN] output on the console for now).
Interesting fact: If I close node_manager without shutting down the master and restart the node_manager, I don't get any warning at all. Strange...

[tkazik]

A maybe related follow-up on this one: I now have troubles discovering the robot on the latest noetic-devel branch and get the following errors:

Do you have an idea what that temporary name resolution failure might be? Thx!

PS: I can ssh into the robot without issues from the terminal.

[atiderko]

It looks like paramiko and ssh resolve names following different paths. paramiko is the library I use to handle SSH connections.

Unfortunately I cannot answer what causes temporary failure name resolution.

Did you try to use the IP address instead of robot?

[tsafs]

For the record

We stumbled across the same problem, installing multimaster_fkie via apt. We instead manually installed the package via catkin_make, as @atiderko suggested, which resolved the problem for our side. We are running the multimaster inside a docker image using Ubuntu 20.04 on an Ubuntu 20.04 host.

[atiderko]

@sebastianFast thank you for reporting. Currently I still did not found a solution to this problem. I don't know why the generated code will not be included into the package on Ubuntu 20.04...

[Sr4l]

@atiderko my guess is the missing dependency of "python3-grpc-tools". I don't know how rosdep get the required package names, but as i can see with a fresh ubuntu20.04 and noetic the package "python3-grpc-tools" is not going to be installed and building of the node_manager fails.

[jk-ethz]

@atiderko my guess is the missing dependency of "python3-grpc-tools". I don't know how rosdep get the required package names, but as i can see with a fresh ubuntu20.04 and noetic the package "python3-grpc-tools" is not going to be installed and building of the node_manager fails.

Made no difference for me. Still not working.