Best practice for closing LogConfig

[gelanchez]

Hi,

I’m working with the FPV example and am adding functionality to close the connection and restart it. However, I’m encountering some errors (e.g. OSError and BrokenPipeError) that seem to originate from the logging framework. An additional problem comes after the second or third reconnection where stop receiving log data and have to restart the Crazyflie, although the images are correctly processed. I wonder if I should close better the framework while the CF is still connected after the stop command.

Here’s the approach I’m currently using:

def stop(self):
    if self.state == State.CONNECTED:
        self.logconfig.data_received_cb.remove_callback(self.pos_data)
        self.cf.log.reset()
        self.logconfig.stop() # TODO needed?
        self.logconfig.delete() # TODO needed?
        self.cf.close_link()
    if self.image_downloader_thread:
        self.image_downloader_thread.stop()
    self.update_state(State.DISCONNECTED)

Many thanks,
Jose

[gemenerik]

Hi Jose,

Given the specific issues you're encountering, it might be helpful to simplify the scenario to isolate the problem more effectively. I suggest creating a standalone script that focuses solely on connecting to the Crazyflie, starting the log, stopping it, and then reconnecting. This approach can make it easier to reproduce the issue and help us understand what might be going wrong without the complexity of the full application.

[gelanchez]

Hi, @ataffanel, thanks and sorry for my late reply, I'm having some issues with the flashing procedure (I've opened another discussion). Yes, my connect function is that, the only difference is the addition of the callbacks, which doesn't matter here. This is the full code:

import logging
import time

import cflib.crtp
from cflib.crazyflie import Crazyflie
from cflib.crazyflie.log import LogConfig
from cflib.utils import uri_helper

logging.basicConfig(level=logging.INFO)

class LoggingExample:
    def __init__(self):
        self.uri = uri_helper.uri_from_env(default='tcp://192.168.4.1:5000')
        self.is_connected = False

    def connect(self):
        if not self.is_connected:
            self._cf = Crazyflie(ro_cache=None, rw_cache='cache')
            self._cf.connected.add_callback(self._connected)
            self._cf.disconnected.add_callback(self._disconnected)
            self._cf.open_link(self.uri)
            if not self._cf.link:
                print(f'Could not connect to {self.uri}')
                return
            if not hasattr(self._cf.link, 'cpx'):
                print('Not connecting with WiFi')
                self._cf.close_link()

    def _connected(self, link_uri):
        self.is_connected = True
        print(f'Connected to {link_uri}')
        self.logconfig = LogConfig(name='Position', period_in_ms=500)
        self.logconfig.add_variable('stabilizer.roll')

        try:
            self._cf.log.add_config(self.logconfig)
            if self.logconfig.valid:
                self.logconfig.data_received_cb.add_callback(self.log_data)
                self.logconfig.error_cb.add_callback(self.log_error)
                self.logconfig.start()
            else:
                print('One or more of the variables in the configuration was not found in log TOC')
        except KeyError as e:
            print(f'Could not start log configuration, {e} not found in TOC')
        except AttributeError:
            print('Could not add Position log config, bad configuration')

    def stop(self):
        if self.is_connected:
            self.logconfig.data_received_cb.remove_callback(self.log_data)
            self.logconfig.error_cb.remove_callback(self.log_error)
            self.logconfig.stop() # TODO needed?
            self.logconfig.delete() # TODO needed?
            self._cf.log.reset()
            self._cf.close_link()

    def log_error(self, logconf, msg):
        print('Error when logging %s: %s' % (logconf.name, msg))

    def log_data(self, timestamp, data, logconf):
        print(f'[{timestamp}][{logconf.name}]: ', end='')
        for name, value in data.items():
            print(f'{name}: {value:3.3f} ', end='')
        print()

    def _disconnected(self, link_uri):
        print('Disconnected from %s' % link_uri)
        self.is_connected = False
        del self._cf


if __name__ == '__main__':
    cflib.crtp.init_drivers()
    connections = 0
    le = LoggingExample()
    try:
        while connections < 3: # Try and connect 3 times
            time.sleep(10)
            if not le.is_connected:
                le.connect()
            else:
                connections += 1
                time.sleep(10)
                le.stop()    
    except Exception as e:    
        print(e)

I'm able connect to the Crazyflie the three times, but I only receive the roll information during the first one. I wonder if there is something I need to send to the CF or close on the client side, or if it has to do with the firmware I've flashed. The only way to get the information back is by restarting the Crazyflie.

[gemenerik]

I tried your attached code a couple of times, and it has worked for me when using self._cf.fully_connected.add_callback(self._connected), however inconsistent. I suspect it has more to do with connecting over wifi & cpx than it being a problem with the logconfig.

[gelanchez]

Hi, @gemenerik. I've tried with the fully_connected method and the result is the same. I agree this looks more a problem of the WiFi & CPX than the logconfig, since the only way of seeing the logs again is by rebooting the crazyflie. I'm using the wifi_img_streamer example there. Is there anything we can change in the Gap8 code to make it work? Nothing in the client?

Thanks,
Jose

[ToveRumar]

Hi @gelanchez !

After looking into the wifi streaming part of the AI deck code,
it seems like we are never closing the socket, which will result in the issue you are seeing.

We are also not taking care of other issues that could occur during sending, even though this is uncommon to bump into other issues it could happen. I think this will cause complete stop in the flow of data.

Im creating an issue here bitcraze/aideck-esp-firmware#24.

If you need to fix this ASAP, I would recommend trying to fix this issue by calling close() on the socket after explicit disconnect or after getting error on receive or send. This allows a new socket to take its place if you re.connect again from the client side.

I hope this can help you to move forward with this!

[gelanchez]

Hi, @ToveRumar!

Thanks for confirming that it was an issue with the socket handling. I'll subscribe to that issue to see how it's fixed. Shouldn't you open another one in the aideck-gap8-examples repository?

I'm only handling the connection in the gap8 code as it's done in the wifi_img_streamer, and I'm not sure if I can do something with it to help fixing it, as I'm now dealing directly with sockets:

void rx_task(void *parameters)
{
    (void)parameters; // Inform the compiler parameters are not used to avoid warnings
    char *taskname = pcTaskGetName(NULL);
    cpxPrintToConsole(LOG_TO_CRTP, "Task %s created\n", taskname);

    while (true)
    {
        // TODO client alive, multiple clients?
        cpxReceivePacketBlocking(CPX_F_WIFI_CTRL, &g_rxPacket);

        WiFiCTRLPacket_t *wifiCtrl = (WiFiCTRLPacket_t*) g_rxPacket.data; // Pointers to the same memory location

        switch (wifiCtrl->cmd)
        {
            case WIFI_CTRL_STATUS_WIFI_CONNECTED: // Not used in access point (I think)
                cpxPrintToConsole(LOG_TO_CRTP, "WiFi connected (%u.%u.%u.%u)\n",
                                wifiCtrl->data[0], wifiCtrl->data[1],
                                wifiCtrl->data[2], wifiCtrl->data[3]);
                g_wifiConnected = true;                
                break;
            case WIFI_CTRL_STATUS_CLIENT_CONNECTED:
                cpxPrintToConsole(LOG_TO_CRTP, "WiFi client connection status: %u\n", wifiCtrl->data[0]);
                g_wifiClientConnected = true;
                break;
            default:
                break;
        }
    }
}

I might need to wait for the fix or an example of better handling.

Thanks for all,
Jose