uofsspaceteam/robocluster_manager

Manages process for the robocluster framework

robocluster_manager's Issues

Monitor process state

The process manager needs to be able to tell when a process is crashed, stuck in a loop, or operating normally, and decide what to do if abnormal behaviour is detected.

Updates to this repository are not being published to Slack

I have noticed that updates and issues from this repository were not being published on the Github and Github-Issues channels on Slack. I'm not sure if this is mandatory or not, but it would be nice to have the updates published on Slack.

Logging system

Instead of processes just printing to standard out, the process manager could agregate logging from all processes on the network and dump it to a common file.

USB device management

The process manager will also be in charge of handling USB devices.

• Detect when USB devices are plugged in or disconnected.
• "Loan" USB devices to processes.

Processes must talk to the process manager to find a USB serial device. Assuming the USB device is not already in use by another process, the process manager will give the requesting process a direct communication channel to the USB device. For now we well only allow devices to be accessed on the same computer, but in the future it may be interesting to allow processes on one computer to access USB devices on another computer.

Start and stop processes dynamically

There should be methods to start or stop all processes at once, or each processes individually.

Read process configuration from a file.

It would be nice to have a single file that identifies which process should be run on each mahcine.
I am thinking JSON would be a convenient format, as we will be using it for data serialization anyway.
Example format:

{ 
    "localhost" : {
        "sleeper" : {
            "cmd" : "python sleeper.py"
        },
        "crash": {
            "cmd" : "python crash.py"
        }
    },
   "DrivePy" : {
       "DriveProcess" : {
           "cmd" : "python DriveProcess.py",
           "some_setting": 5
       },
       "NavigationProcess" : {
           "cmd" : "./NavigationProcess"
       }
   },
   "ArmPy" : {
       "ArmProcess" : {
           "cmd" : "python ArmProcess.py"
       }
   }
}

The master process manager would read this file on start up and tell other process managers what processes to run.

Ensure robocluster has a kill switch

If something goes wrong during the operation of the rover and we need to shut everything down,
a kill command from the master node should ensure every process on the system is stopped immediately.

Share process state with other process managers

Each process manager will share the state of its own processes with all the other process managers.
This should also provide mechanisms for processes on one machine to talk directly to processes on another machine. Some sort of concensus algorithm should be used so that process managers on different machines agree with one another as to what the global state is.

Remote API

Currently in progess, we need to be able to start and stop process remotely from another machine.