Pipeline containing script node alone does not work in standalone mode

zepman

I have the OAK-D Pro PoE and use a script node to send messages via MQTT. I'm finding that if the pipeline contains only the script node that does the MQTT publishing, it does not work in standalone mode (but it works fine in host-connected mode).

The following is an MRE.

import depthai as dai

# Change the IP to your MQTT broker
MQTT_BROKER = "192.168.0.5"
MQTT_BROKER_PORT = 1883
MQTT_TOPIC = "stat/camera1"

pipeline = dai.Pipeline()

script = pipeline.create(dai.node.Script)
script.setProcessor(dai.ProcessorType.LEON_CSS)

script_text = f"""
mqttc = Client()
node.warn('Connecting to MQTT broker...')
mqttc.connect("{MQTT_BROKER}", {MQTT_BROKER_PORT}, 60)
node.warn('Successfully connected to MQTT broker!')

mqttc.loop_start()
cnt = 0
while True:
    cnt+=1
    if cnt > 100000:
        node.warn("test_message")
        (ok, id) = mqttc.publish("{MQTT_TOPIC}", "test_message", qos=2)
        cnt = 0
"""

with open("paho-mqtt.py", "r") as f:
    paho_script = f.read()
    script.setScript(f"{paho_script}\n{script_text}")

The above pipeline runs fine when run in host-connected mode i.e., I see the message being published continuously, using MQTT Explorer. However, the same pipeline doesn't work in standalone mode (flashing completes successfully and the device restarts with a "clicking" sound); I don't see any message published.

Now, if I modify the script to serve the message via HTTP instead of MQTT, like below, then the pipeline works fine in both host-connected and standalone modes.

script.setScript("""
    from socketserver import ThreadingMixIn
    from http.server import BaseHTTPRequestHandler, HTTPServer

    class ThreadingSimpleServer(ThreadingMixIn, HTTPServer):
        pass

    class HTTPHandler(BaseHTTPRequestHandler):
        def do_GET(self):
            if self.path == '/':
                self.send_response(200)
                self.end_headers()
                self.wfile.write(b'<h1>test_message</h1>')

    with ThreadingSimpleServer(("", 8080), HTTPHandler) as httpd:
        httpd.serve_forever()
""")

I don't quite understand what might be the issue here. Any help would be appreciated!

erik

Hi @zepman ,
The issue you're experiencing might be related to the fact that the standalone mode currently lacks a DNS resolver. This means that you need to use IP addresses instead of domain names. In your MQTT script, ensure that you're using the IP address of your MQTT broker and not a domain name.

zepman

Thanks erik. I did read the bit about the lack of a DNS resolver in the docs. Unfortunately, that's not the issue here. I am using the IP address of the MQTT broker exactly as shown in the MRE. It only works in host-connected mode but not in standalone mode.

jakaskerl

Hi @zepman
Can you enable logging when in standalone mode; you can utilize the http to send messages over to the host.

Thanks,
Jaka

zepman

Thanks jakaskerl. So I modified the MRE - specified a log path and added another script node that serves via HTTP. This is as below. Is this what you meant?

import depthai as dai
import random

# Change the IP to your MQTT broker
MQTT_BROKER = "192.168.0.5"
MQTT_BROKER_PORT = 1883
MQTT_TOPIC = "stat/camera1"

pipeline = dai.Pipeline()
cfg = dai.BoardConfig()
cfg.emmc = False
cfg.logPath = f'/depthai_log_file{random.randrange(0, 100)}.txt'
config = dai.Device.Config()
config.board = cfg
pipeline.setBoardConfig(cfg)

scriptServer = pipeline.create(dai.node.Script)
scriptServer.setProcessor(dai.ProcessorType.LEON_CSS)
scriptServer.setScript("""
import http.server
import socketserver
import socket
import fcntl
import struct
import os
import time
def get_ip_address(ifname):
    s = socket.socket(socket.AF_INET, socket.SOCK_DGRAM)
    return socket.inet_ntoa(fcntl.ioctl(
        s.fileno(),
        -1071617759,  # SIOCGIFADDR
        struct.pack('256s', ifname[:15].encode())
    )[20:24])
# Note: `chdir` here will prevent unmount, this should be improved!
time.sleep(1)
os.chdir('/')
PORT = 80
Handler = http.server.SimpleHTTPRequestHandler
with socketserver.TCPServer(("", PORT), Handler) as httpd:
    ip = get_ip_address('re0')
    node.warn(f'===== HTTP file server accessible at: http://{ip}')
    httpd.serve_forever()
""")

script = pipeline.create(dai.node.Script)
script.setProcessor(dai.ProcessorType.LEON_CSS)

script_text = f"""
mqttc = Client()
node.warn('Connecting to MQTT broker...')
mqttc.connect("{MQTT_BROKER}", {MQTT_BROKER_PORT}, 60)
node.warn('Successfully connected to MQTT broker!')

mqttc.loop_start()
cnt = 0
while True:
    cnt+=1
    if cnt > 100000:
        # node.warn("test_message")
        (ok, id) = mqttc.publish("{MQTT_TOPIC}", "test_message", qos=2)
        cnt = 0
"""

In any case, even the above pipeline works only in host-connected mode but not in standalone mode (site can't be reached when I access the file server on a browser).

jakaskerl

Hi @zepman
Directly send the log messages via http. No need to save them unless you have emmc and want to do it.

Thanks,
Jaka

zepman

Thanks jakaskerl. My knowledge of communication/networking stuff is very limited, but I cobbled something together based on one of the examples in the docs. Not sure if this is the right way to do it? Anyway, it still only works in host-connected mode but not in standalone mode.

import depthai as dai

# Change the IP to your MQTT broker
MQTT_BROKER = "192.168.0.5"
MQTT_BROKER_PORT = 1883
MQTT_TOPIC = "stat/camera1"

pipeline = dai.Pipeline()

script = pipeline.create(dai.node.Script)
script.setProcessor(dai.ProcessorType.LEON_CSS)

script_text = f"""

def send_log(log_message):
    b = Buffer(len(log_message))
    b.setData(log_message.encode('utf-8'))
    node.io['out'].send(b)

mqttc = Client()
send_log('Connecting to MQTT broker...')
mqttc.connect("{MQTT_BROKER}", {MQTT_BROKER_PORT}, 60)
send_log('Successfully connected to MQTT broker!')

mqttc.loop_start()
cnt = 0
while True:
    cnt+=1
    if cnt > 100000:
        send_log('Publishing test message...')
        (ok, id) = mqttc.publish("{MQTT_TOPIC}", "test_message", qos=2)
        if (ok == 0):
            send_log('Published!')

        cnt = 0
"""

with open("utils/paho-mqtt.py", "r") as f:
    paho_script = f.read()
    script.setScript(f"{paho_script}\n{script_text}")

scriptServer = pipeline.create(dai.node.Script)
scriptServer.setProcessor(dai.ProcessorType.LEON_CSS)
scriptServer.setScript("""
    from socketserver import ThreadingMixIn
    from http.server import BaseHTTPRequestHandler, HTTPServer

    class ThreadingSimpleServer(ThreadingMixIn, HTTPServer):
        pass

    class HTTPHandler(BaseHTTPRequestHandler):
        def do_GET(self):
            if self.path == '/':
                try:
                    self.send_response(200)
                    self.end_headers()
                    while True:
                        in_msg = node.io['msg'].get()
                        self.wfile.write(in_msg.getData())
                        self.end_headers()

                except Exception as ex:
                    node.warn(str(ex))

    with ThreadingSimpleServer(("", 8080), HTTPHandler) as httpd:
        httpd.serve_forever()
""")

script.outputs['out'].link(scriptServer.inputs['msg'])

jakaskerl

Hi @zepman
I think I figured out the issue. The way you are running your MQTT server likely only allows local connections.

I used mosquitto and set allow_anonymous to true.

full config file:

# Place this configuration in /usr/local/etc/mosquitto/mosquitto.conf or the relevant config path for your installation.

# Configures the port Mosquitto will listen on. The default MQTT port is 1883.
port 1883

# By default, Mosquitto listens on all network interfaces (0.0.0.0).
# If you want to specify this explicitly, uncomment the following line.
bind_address 0.0.0.0

# Optional: Specify which IP addresses or ranges are allowed to connect.
# By default, all addresses are allowed.
allow_anonymous true

# Optional: Configure listener settings to listen on multiple ports or bind to specific addresses.
# Example for TLS/SSL:
# listener 8883
# cafile /path/to/ca.crt
# certfile /path/to/server.crt
# keyfile /path/to/server.key

# Enable logging for debugging purposes.
log_type all
log_dest file /usr/local/var/log/mosquitto/mosquitto.log

# Optional: Set the persistence file path to store messages and client information.
persistence true
persistence_location /usr/local/var/lib/mosquitto/

full oak file:

import depthai as dai

# Change the IP to your MQTT broker
MQTT_BROKER = "10.12.118.151"
MQTT_BROKER_PORT = 1883
MQTT_TOPIC = "test/topic"

pipeline = dai.Pipeline()

script = pipeline.create(dai.node.Script)
script.setProcessor(dai.ProcessorType.LEON_CSS)

script_text = f"""

import time
time.sleep(10)


mqttc = Client()
node.warn('Connecting to MQTT broker...')
mqttc.connect("{MQTT_BROKER}", {MQTT_BROKER_PORT}, 60)
node.warn('Successfully connected to MQTT broker!')

mqttc.loop_start()
cnt = 0
while True:
    cnt+=1
    if cnt > 100000:
        node.warn("test_message")
        (ok, id) = mqttc.publish("{MQTT_TOPIC}", "test_message", qos=2)
        cnt = 0
"""

with open("paho-mqtt.py", "r") as f:
    paho_script = f.read()
    script.setScript(f"{paho_script}\n{script_text}")

(f, bl) = dai.DeviceBootloader.getFirstAvailableDevice()
bootloader = dai.DeviceBootloader(bl)
progress = lambda p : print(f'Flashing progress: {p*100:.1f}%')
(r, errmsg) = bootloader.flash(progress, pipeline)
if r: print("Flash OK")
else: print("Flash ERROR:", errmsg)

Thanks,
Jaka

zepman

jakaskerl The allow_anonymous flag in my mosquitto config file has been true all along. So that wasn't the problem.

However, thanks to your full oak Python code, I have narrowed it down a bit. Your code actually works for me without an issue in standalone mode too. The difference when compared to my original MRE is the delay at the start of the script introduced by the line time.sleep(10).

So, without that delay in the script, the pipeline doesn't work in standalone mode but with that delay, it works fine. Could this be due to some networking-related initialisation in the device that needs to be allowed time to complete before the script node is executed?

erik

@zepman I believe so, so I'd leave it in there.

zepman

erik Yes, I will leave that delay in there as it allows me to get past this issue. However, I am curious if the exact details as to why the delay is necessary are already known?

jakaskerl

zepman
I put it in to make sure the network stack is fully initialized before any attempt at connection is made, since I had suspicion that might be the cause. Haven't tested without it though. Good thing I forgot to remove it then 😉. Looks like the bootup sequence doesn't require network to be fully setup before executing the script.

Thanks,
Jaka

erik

@zepman I'm not sure.

zepman

jakaskerl

Yeah, that makes sense. I had also observed that the pipeline in my original MRE worked fine if I simply added a neural network node to the pipeline (didn't even need to link up the nodes). Looks like that prolonged the bootup long enough to circumvent the issue,