Google Cloud IoT Python MQTT内存不足错误排查求助

Hey there, let's break down this frustrating issue you're facing. It's odd that your Node.js sample works flawlessly across both projects, but Python only fails in Project B with that vague "out of memory" disconnect error. Let's walk through targeted troubleshooting steps to get to the root cause:

1. Check IoT Core and Project Resource Quotas in Project B #

Even though Node.js connects fine, Python's client library might handle connections or resource allocation differently.

• Head to the GCP Console for Project B, navigate to IAM & Admin > Quotas, and search for IoT Core-related limits (like "Device connections per registry" or "Messages per second"). Make sure you aren't hitting or approaching these quotas—exceeding them can trigger unexpected disconnects that manifest as memory errors.
• Also verify overall project resource limits: if you're running the Python code on a VM or Cloud Function, check if the instance's allocated memory is too low (e.g., a 256MB Cloud Function might struggle with Python's TLS overhead where Node.js doesn't).

2. Validate Dependency Version Consistency #

Version mismatches in Python dependencies are a common culprit for environment-specific bugs.

• In both Project A and B's runtime environments, run pip list and compare versions of core packages: google-cloud-iot, paho-mqtt, google-auth, and cryptography. If Project B has older (or newer) versions, match them to Project A's working set—for example, if Project A uses paho-mqtt==1.6.1, install that exact version in Project B with pip install paho-mqtt==1.6.1.

3. Double-Check Device Authentication and Permissions #

While Node.js might be more forgiving, Python's client can be stricter about key formatting and permissions.

• Confirm the device key (RS256/ES256) in Project B is identical to Project A's—check the key file path, file permissions (ensure it's readable only by the running user with chmod 600 your-key-file.pem), and key expiration date.
• Verify the service account associated with the Python code (or the default gcloud auth) has the necessary IoT Core permissions in Project B: cloudiot:devices.connect and cloudiot:devices.publish. Even if Node.js works, Python might be using a different auth flow that lacks these permissions.

4. Enable Debug Logging for the Python Client #

The truncated error message isn't enough—enable verbose logging to see exactly what's happening before the disconnect:
Add this at the top of your Python script:

import logging
logging.basicConfig(level=logging.DEBUG)

This will print detailed MQTT handshake, authentication, and message flow logs. Look for repeated retries, TLS errors, or unexpected resource allocation that could lead to memory exhaustion.

5. Rule Out Network Environment Differences #

Project B's network might have restrictions that affect Python's connection handling but not Node.js:

• Check if Project B's runtime uses a proxy server—Python's MQTT client doesn't auto-detect proxies like Node.js might. If a proxy is required, configure it in your Python code using the proxy_set parameter for the Paho client.
• Adjust MQTT connection parameters to reduce resource strain: try setting keepalive=60 (lower than default if needed) and reconnect_delay_max=30 to limit excessive reconnection attempts that could eat up memory.

6. Test a Minimal Python Implementation #

The full sample code from Git might include loops or resource-heavy logic that's causing memory leaks in Project B's environment. Create a stripped-down test script:

from google.cloud import iot_v1
from paho.mqtt import client as mqtt_client
import ssl

# Replace with your Project B details
PROJECT_ID = "your-project-b-id"
REGISTRY_ID = "my-registry"
DEVICE_ID = "my-device"
PRIVATE_KEY_FILE = "rsa_private.pem"

def on_connect(client, userdata, flags, rc):
    print(f"Connected with result code {rc}")
    client.subscribe(f"/devices/{DEVICE_ID}/config")

def on_message(client, userdata, msg):
    print(f"Received message: {msg.payload.decode()}")

client = mqtt_client.Client(client_id=f"projects/{PROJECT_ID}/locations/us-central1/registries/{REGISTRY_ID}/devices/{DEVICE_ID}")
client.username_pw_set(username="unused", password=None)
client.tls_set(ca_certs="roots.pem", certfile=None, keyfile=PRIVATE_KEY_FILE, cert_reqs=ssl.CERT_REQUIRED, tls_version=ssl.PROTOCOL_TLSv1_2, ciphers=None)
client.on_connect = on_connect
client.on_message = on_message

client.connect(f"{PROJECT_ID}.mqtt.googleapis.com", 8883, 60)
client.loop_start()

# Send a single test message
client.publish(f"/devices/{DEVICE_ID}/events", b"Test message from minimal Python script")

If this minimal script works, the issue is in the original sample's extra logic (e.g., infinite loops without proper garbage collection). If it still fails, the problem is tied to Project B's configuration or runtime environment.

内容的提问来源于stack exchange，提问作者Jun Ying