Starting ctlmgr on a different host

jfernand

Hi all,
We have an artiq_master and artiq_ctlmgr working in a "master" computer (192.168.1.8), to which we can connect using artiq_dashboard from other 2 "client" computers (.1.2 and .1.4). The master ctlmgr controls 2 local NDSPs (a camera and PMT).

Now we want to use another camera and PMT in one of the client computers (.1.4), same make and model, so the same NDSP sources work. We thought we could start another ctlmgr in the client computer using artiq_ctlmgr -s 192.168.1.8 but it doesn't work. It doesn't error out either, it just says DEBUG:ctlmgr(LaboratorioCont):asyncio:Using proactor: IocpProactor when artiq_ctlmgr is launched with -v -v -v.

We have checked that the controllers start in the client. We have even started an artiq_master in the client computer, and launched an artiq_ctlmgr too, which succesfully starts the "client" camera and PMT NDSPs. The command were just artiq_master and artiq_ctlmgr with device_db entries like

device_db.update({
    "ixon": {
        "type": "controller",
        "host": "192.168.1.8",
        "port": 3214,
        "command": "python aqctl_ixon.py -v -p {port}"
        # "command": "python aqctl_ixon.py -v -p {port} --simulation" # --bind + mast_addr
    },
})

device_db.update({
    "ixon3": {
        "type": "controller",
        "host": "::1",
        "port": 3214,
        "command": "python aqctl_ixon3.py -v -p {port}"
        # "command": "python aqctl_ixon.py -v -p {port} --simulation" # --bind + mast_addr
    },
})

If now we start the artiq_master in the "master" computer (and ctlmgr to launch the master-local camera NDSP), how could we start a client-local ctlmgr to launch the client-local camera NDSP? We tried artiq_ctlmgr -s 192.168.1.8 -v -v -v but we got DEBUG:ctlmgr(LaboratorioCont):asyncio:Using proactor: IocpProactor

Thanks for any hints

sb10q

jfernand DEBUG:ctlmgr(LaboratorioCont):asyncio:Using proactor: IocpProactor

This is a debug information message and it does not indicate anything wrong (other than you are using Windows, which is a lame excuse for an operating system).

Controller managers determine their IP addresses by looking at their end of the socket that they use to connect to the master, and then start all the controllers in the DDB that are assigned to this IP. This is probably what is not working out in your case.

fdomgon

sb10q Thank you very much for the prompt answer!

Sorry for the late feedback, @jfernand and I got it working with the following device_db.py lines:

device_db.update({
    "ixon3": {
        "type": "controller",
        "host": "192.168.1.4",                         # ::1 local master, 192.168.1.4 artiq-master
        "port": 3214,                                  # 3214 local master, N/A (0) artiq-master
        "command": "python aqctl_ixon3.py -v -p {port}"    # "python aqctl_ixon3.py -v -p {port}" local, N/A ("blah") artiq-master
        #"command": "python aqctl_ixon.py -v -p {port} --simulation" # --bind + mast_addr
    },
})

Now we face another problem. When we run artiq_ctlmgr in the client we can see the driver initialization mesages from the iXon3 EMCCD driver but the ping is not working. The ping works when artiq_master is run in the same machine but it doesn't when artiq_master is in a remote computer. We attach session log showing the connection to a local "dummy" artiq_master where the ping works and the error messages that appear when connecting to the remote "real" artiq_master.

We based our ping code on a tiny demonstration driver:

    async def ping(self):
        """Ping camera. The function 'GetCameraSerialNumber' is used. Also reads temp"""
        if not self._simulation:
            temp=c_long()
            self._error=self._lib.GetTemperature(byref(temp))
            if self._error == 20002:
                logger.warning(warning_text(self._error))
                return False
            else:
                logger.info('ping successful with model {} and temp {}C ({})'.format(self._serial,temp.value,ERROR_CODE[self._error]))
                return True
        else:
            logger.info(success_text(self._simulation))
            return True

Is there a verbose explanation of how to correctly program the ping function? Is async really required?

Thank you.

sb10q

What happens when you run the controller manually and call ping via sipyco_rpctool?

fdomgon

sb10q

Thank you for the hint! We run the controller manually using python aqctl_C8855_02_counter.py -v -p 3213 and called the ping from the remote computer using sipyco_rpctool 192.168.1.4 3213 call ping and got the error code: ConnectionRefusedError: [WinError 10061]

This led to the conclusion that the "--bind" parameter was required in the command entry of the device_db.py file located in the remote computer. Now it works both locally and remotely with the following setup:

device_db.py in the local computer:

device_db.update({
    "C8855_02": {
        "type": "controller",
        "host": "::1",                                        # ::1 local master, 192.168.1.4 artiq-master
        "port": 3213,                                         # 3213 local master, N/A (0) artiq-master
        "command": "python aqctl_C8855_02_counter.py -v -p {port}"      # "python aqctl_ixon3.py -v -p {port}" local, N/A ("blah") artiq-master
        ##"command": "python aqctl_C8855_02_counter.py -v -q -p {port} --bind " + mast_addr
    },
})

device_db.py in the remote computer:

device_db.update({
    "C8855_02": {
        "type": "controller",
        "host": "192.168.1.4",
        "port": 3213,
        "command": "python aqctl_C8855_02_counter.py -v -p {port} --bind 192.168.1.4"
        #"command": "python aqctl_C8855_02_counter.py -v -q -p {port} --bind " + mast_addr
    },
})

This way we can either run control manager connected to a local master or to a remote master and it works in both cases. We infer that it is the master who reads the device_db.py and sends the command to the control manager. Note the difference in the bind line when we run the control manager connected to the remote computer compared to the screenshot I attached in my previous post.