Solutions to common issues for Azure IoT Edge

Caution

This article references CentOS, a Linux distribution that is nearing End Of Life (EOL) status. Please consider your use and planning accordingly.

Applies to: IoT Edge 1.4 checkmark IoT Edge 1.4

Important

IoT Edge 1.4 is the supported release. If you are on an earlier release, see Update IoT Edge.

Use this article to identify and resolve common issues when using IoT Edge solutions. If you need information on how to find logs and errors from your IoT Edge device, see Troubleshoot your IoT Edge device.

Provisioning and Deployment

IoT Edge module deploys successfully then disappears from device

Symptoms

After setting modules for an IoT Edge device, the modules are deployed successfully but after a few minutes they disappear from the device and from the device details in the Azure portal. Other modules than the ones defined might also appear on the device.

Cause

If an automatic deployment targets a device, it takes priority over manually setting the modules for a single device. The Set modules functionality in Azure portal or Create deployment for single device functionality in Visual Studio Code takes effect for a moment. You see the modules that you defined start on the device. Then the automatic deployment's priority starts and overwrites the device's desired properties.

Solution

Only use one type of deployment mechanism per device, either an automatic deployment or individual device deployments. If you have multiple automatic deployments targeting a device, you can change priority or target descriptions to make sure the correct one applies to a given device. You can also update the device twin to no longer match the target description of the automatic deployment.

For more information, see Understand IoT Edge automatic deployments for single devices or at scale.

IoT Edge runtime

IoT Edge agent stops after a minute

Symptoms

The edgeAgent module starts and runs successfully for about a minute, then stops. The logs indicate that the IoT Edge agent attempts to connect to IoT Hub over AMQP, and then attempts to connect using AMQP over WebSocket. When that fails, the IoT Edge agent exits.

Example edgeAgent logs:

2017-11-28 18:46:19 [INF] - Starting module management agent.
2017-11-28 18:46:19 [INF] - Version - 1.0.7516610 (03c94f85d0833a861a43c669842f0817924911d5)
2017-11-28 18:46:19 [INF] - Edge agent attempting to connect to IoT Hub via AMQP...
2017-11-28 18:46:49 [INF] - Edge agent attempting to connect to IoT Hub via AMQP over WebSocket...

Cause

A networking configuration on the host network is preventing the IoT Edge agent from reaching the network. The agent attempts to connect over AMQP (port 5671) first. If the connection fails, it tries WebSockets (port 443).

The IoT Edge runtime sets up a network for each of the modules to communicate on. On Linux, this network is a bridge network. On Windows, it uses NAT. This issue is more common on Windows devices using Windows containers that use the NAT network.

Solution

Ensure that there's a route to the internet for the IP addresses assigned to this bridge/NAT network. Sometimes a VPN configuration on the host overrides the IoT Edge network.

Edge Agent module reports 'empty config file' and no modules start on the device

Symptoms

  • The device has trouble starting modules defined in the deployment. Only the edgeAgent is running but and reports empty config file....

  • When you run sudo iotedge check on a device, it reports Container engine is not configured with DNS server setting, which may impact connectivity to IoT Hub. Please see https://aka.ms/iotedge-prod-checklist-dns for best practices.

Cause

By default, IoT Edge starts modules in their own isolated container network. The device may be having trouble with DNS name resolution within this private network.

Solution

Option 1: Set DNS server in container engine settings

Specify the DNS server for your environment in the container engine settings, which will apply to all container modules started by the engine. Create a file named daemon.json, then specify the DNS server to use. For example:

{
    "dns": ["1.1.1.1"]
}

This DNS server is set to a publicly accessible DNS service. However some networks, such as corporate networks, have their own DNS servers installed and won't allow access to public DNS servers. Therefore, if your edge device can't access a public DNS server, replace it with an accessible DNS server address.

Place daemon.json in the /etc/docker directory on your device.

If the location already contains a daemon.json file, add the dns key to it and save the file.

Restart the container engine for the updates to take effect.

sudo systemctl restart docker

Option 2: Set DNS server in IoT Edge deployment per module

You can set DNS server for each module's createOptions in the IoT Edge deployment. For example:

"createOptions": {
  "HostConfig": {
    "Dns": [
      "x.x.x.x"
    ]
  }
}

Warning

If you use this method and specify the wrong DNS address, edgeAgent loses connection with IoT Hub and can't receive new deployments to fix the issue. To resolve this issue, you can reinstall the IoT Edge runtime. Before you install a new instance of IoT Edge, be sure to remove any edgeAgent containers from the previous installation.

Be sure to set this configuration for the edgeAgent and edgeHub modules as well.

Edge Agent module with LTE connection reports 'empty edge agent config' and causes 'transient network error'

Symptoms

A device configured with LTE connection is having issues starting modules defined in the deployment. The edgeAgent isn't able to connect to the IoT Hub and reports empty edge agent config and transient network error occurred.

Cause

Some networks have packet overhead, which makes the default docker network MTU (1500) too high and causes packet fragmentation preventing access to external resources.

Solution

  1. Check the MTU setting for your docker network.

    docker network inspect <network name>

  2. Check the MTU setting for the physical network adaptor on your device.

    ip addr show eth0

Note

The MTU for the docker network cannot be higher than the MTU for your device. Contact your ISP for more information.

If you see a different MTU size for your docker network and the device, try the following workaround:

  1. Create a new network. For example,

    docker network create --opt com.docker.network.driver.mtu=1430 test-mtu

    In the example, the MTU setting for the device is 1430. Hence, the MTU for the Docker network is set to 1430.

  2. Stop and remove the Azure network.

    docker network rm azure-iot-edge

  3. Recreate the Azure network.

    docker network create --opt com.docker.network.driver.mtu=1430 azure-iot-edge

  4. Remove all containers and restart the aziot-edged service.

    sudo iotedge system stop && sudo docker rm -f $(docker ps -aq -f "label=net.azure-devices.edge.owner=Microsoft.Azure.Devices.Edge.Agent") && sudo iotedge config apply

IoT Edge agent can't access a module's image (403)

Symptoms

A container fails to run, and the edgeAgent logs report a 403 error.

Cause

The IoT Edge agent module doesn't have permissions to access a module's image.

Solution

Make sure that your container registry credentials are correct your device deployment manifest.

IoT Edge hub fails to start

Symptoms

The edgeHub module fails to start. You may see a message like one of the following errors in the logs:

One or more errors occurred.
(Docker API responded with status code=InternalServerError, response=
{\"message\":\"driver failed programming external connectivity on endpoint edgeHub (6a82e5e994bab5187939049684fb64efe07606d2bb8a4cc5655b2a9bad5f8c80):
Error starting userland proxy: Bind for 0.0.0.0:443 failed: port is already allocated\"}\n)

Or

info: edgelet_docker::runtime -- Starting module edgeHub...
warn: edgelet_utils::logging -- Could not start module edgeHub
warn: edgelet_utils::logging --     caused by: failed to create endpoint edgeHub on network nat: hnsCall failed in Win32:
        The process cannot access the file because it is being used by another process. (0x20)

Cause

Some other process on the host machine has bound a port that the edgeHub module is trying to bind. The IoT Edge hub maps ports 443, 5671, and 8883 for use in gateway scenarios. The module fails to start if another process has already bound one of those ports.

Solution

You can resolve this issue two ways:

If the IoT Edge device is functioning as a gateway device, then you need to find and stop the process that is using port 443, 5671, or 8883. An error for port 443 usually means that the other process is a web server.

If you don't need to use the IoT Edge device as a gateway, then you can remove the port bindings from edgeHub's module create options. You can change the create options in the Azure portal or directly in the deployment.json file.

In the Azure portal:

  1. Navigate to your IoT hub and select Devices under the Device management menu.

  2. Select the IoT Edge device that you want to update.

  3. Select Set Modules.

  4. Select Runtime Settings.

  5. In the Edge Hub module settings, delete everything from the Container Create Options text box.

  6. Select Apply to save your changes and create the deployment.

In the deployment.json file:

  1. Open the deployment.json file that you applied to your IoT Edge device.

  2. Find the edgeHub settings in the edgeAgent desired properties section:

      "edgeHub": {
          "restartPolicy": "always",
          "settings": {
             "image": "mcr.microsoft.com/azureiotedge-hub:1.4",
             "createOptions": "{\"HostConfig\":{\"PortBindings\":{\"443/tcp\":[{\"HostPort\":\"443\"}],\"5671/tcp\":[{\"HostPort\":\"5671\"}],\"8883/tcp\":[{\"HostPort\":\"8883\"}]}}}"
          },
          "status": "running",
          "type": "docker"
       }
    
  3. Remove the createOptions line, and the trailing comma at the end of the image line before it:

      "edgeHub": {
          "restartPolicy": "always",
          "settings": {
          "image": "mcr.microsoft.com/azureiotedge-hub:1.4",
          "status": "running",
          "type": "docker"
    }
    
  4. Select Create to apply it to your IoT Edge device again.

IoT Edge module fails to send a message to edgeHub with 404 error

Symptoms

A custom IoT Edge module fails to send a message to the IoT Edge hub with a 404 Module not found error. The IoT Edge runtime prints the following message to the logs:

Error: Time:Thu Jun  4 19:44:58 2018 File:/usr/sdk/src/c/provisioning_client/adapters/hsm_client_http_edge.c Func:on_edge_hsm_http_recv Line:364 executing HTTP request fails, status=404, response_buffer={"message":"Module not found"}u, 04 )

Cause

The IoT Edge runtime enforces process identification for all modules connecting to the edgeHub for security reasons. It verifies that all messages being sent by a module come from the main process ID of the module. If a message is being sent by a module from a different process ID than initially established, it will reject the message with a 404 error message.

Solution

As of version 1.0.7, all module processes are authorized to connect. For more information, see the 1.0.7 release changelog.

If upgrading to 1.0.7 isn't possible, complete the following steps. Make sure that the same process ID is always used by the custom IoT Edge module to send messages to the edgeHub. For instance, make sure to ENTRYPOINT instead of CMD command in your Docker file. The CMD command leads to one process ID for the module and another process ID for the bash command running the main program, but ENTRYPOINT leads to a single process ID.

Stability issues on smaller devices

Symptoms

You may experience stability problems on resource constrained devices like the Raspberry Pi, especially when used as a gateway. Symptoms include out of memory exceptions in the IoT Edge hub module, downstream devices failing to connect, or the device failing to send telemetry messages after a few hours.

Cause

The IoT Edge hub, which is part of the IoT Edge runtime, is optimized for performance by default and attempts to allocate large chunks of memory. This optimization isn't ideal for constrained edge devices and can cause stability problems.

Solution

For the IoT Edge hub, set an environment variable OptimizeForPerformance to false. There are two ways to set environment variables:

In the Azure portal:

  1. In your IoT Hub, select your IoT Edge device and from the device details page and select Set Modules > Runtime Settings.

  2. Create an environment variable for the IoT Edge hub module called OptimizeForPerformance with type True/False that is set to False.

    Screenshot that shows where to add the OptimizeForPerformance environment variable in the Azure portal.

  3. Select Apply to save changes, then select Review + create.

    The environment variable is now in the edgeHub property of the deployment manifest:

       "edgeHub": {
          "env": {
                "OptimizeForPerformance": {
                   "value": false
                }
          },
          "restartPolicy": "always",
          "settings": {
                "image": "mcr.microsoft.com/azureiotedge-hub:1.4",
                "createOptions": "{\"HostConfig\":{\"PortBindings\":{\"443/tcp\":[{\"HostPort\":\"443\"}],\"5671/tcp\":[{\"HostPort\":\"5671\"}],\"8883/tcp\":[{\"HostPort\":\"8883\"}]}}}"
          },
          "status": "running",
          "type": "docker"
       }
    
  4. Select Create to save your changes and deploy the module.

Security daemon couldn't start successfully

Symptoms

The security daemon fails to start and module containers aren't created. The edgeAgent, edgeHub and other custom modules aren't started by IoT Edge service. In aziot-edged logs, you see this error:

  • The daemon could not start up successfully: Could not start management service
  • caused by: An error occurred for path /var/run/iotedge/mgmt.sock
  • caused by: Permission denied (os error 13)

Cause

For all Linux distros except CentOS 7, IoT Edge's default configuration is to use systemd socket activation. A permission error happens if you change the configuration file to not use socket activation but leave the URLs as /var/run/iotedge/*.sock, since the iotedge user can't write to /var/run/iotedge meaning it can't unlock and mount the sockets itself.

Solution

You don't need to disable socket activation on a distribution where socket activation is supported. However, if you prefer to not use socket activation at all, put the sockets in /var/lib/iotedge/.

  1. Run systemctl disable iotedge.socket iotedge.mgmt.socket to disable the socket units so that systemd doesn't start them unnecessarily
  2. Change the iotedge config to use /var/lib/iotedge/*.sock in both connect and listen sections
  3. If you already have modules, they have the old /var/run/iotedge/*.sock mounts, so docker rm -f them.

Networking

IoT Edge security daemon fails with an invalid hostname

Symptoms

Attempting to check the IoT Edge security manager logs fails and prints the following message:

Error parsing user input data: invalid hostname. Hostname cannot be empty or greater than 64 characters

Cause

The IoT Edge runtime can only support hostnames that are shorter than 64 characters. Physical machines usually don't have long hostnames, but the issue is more common on a virtual machine. The automatically generated hostnames for Windows virtual machines hosted in Azure, in particular, tend to be long.

Solution

When you see this error, you can resolve it by configuring the DNS name of your virtual machine, and then setting the DNS name as the hostname in the setup command.

  1. In the Azure portal, navigate to the overview page of your virtual machine.

  2. Open the configuration panel by selecting Not configured (if your virtual machine is new) under DNS name, or select your existing DNS name. If your virtual machine already has a DNS name configured, you don't need to configure a new one.

    Screenshot of how to open the configuration panel of your DNS name.

  3. Provide a value for DNS name label if you don't have one already and select Save.

  4. Copy the new DNS name, which should be in the format:
    <DNSnamelabel>.<vmlocation>.cloudapp.azure.com.

  5. On the IoT Edge device, open the config file.

    sudo nano /etc/aziot/config.toml
    
  6. Replace the value of hostname with your DNS name.

  7. Save and close the file, then apply the changes to IoT Edge.

    sudo iotedge config apply
    

IoT Edge module reports connectivity errors

Symptoms

IoT Edge modules that connect directly to cloud services, including the runtime modules, stop working as expected and return errors around connection or networking failures.

Cause

Containers rely on IP packet forwarding in order to connect to the internet so that they can communicate with cloud services. IP packet forwarding is enabled by default in Docker, but if it gets disabled then any modules that connect to cloud services won't work as expected. For more information, see Understand container communication in the Docker documentation.

Solution

Use the following steps to enable IP packet forwarding.

  1. Open the sysctl.conf file.

    sudo nano /etc/sysctl.conf
    
  2. Add the following line to the file.

    net.ipv4.ip_forward=1
    
  3. Save and close the file.

  4. Restart the network service and docker service to apply the changes.

IoT Edge behind a gateway can't perform HTTP requests and start edgeAgent module

Symptoms

The IoT Edge runtime is active with a valid configuration file, but it can't start the edgeAgent module. The command iotedge list returns an empty list. The IoT Edge runtime reports Could not perform HTTP request in the logs.

Cause

IoT Edge devices behind a gateway get their module images from the parent IoT Edge device specified in the parent_hostname field of the config file. The Could not perform HTTP request error means that the downstream device isn't able to reach its parent device via HTTP.

Solution

Make sure the parent IoT Edge device can receive incoming requests from the downstream IoT Edge device. Open network traffic on ports 443 and 6617 for requests coming from the downstream device.

IoT Edge behind a gateway can't perform HTTP requests and start edgeAgent module

Symptoms

The IoT Edge daemon is active with a valid configuration file, but it can't start the edgeAgent module. The command iotedge list returns an empty list. The IoT Edge daemon logs report Could not perform HTTP request.

Cause

IoT Edge devices behind a gateway get their module images from the parent IoT Edge device specified in the parent_hostname field of the config file. The Could not perform HTTP request error means that the downstream device isn't able to reach its parent device via HTTP.

Solution

Make sure the parent IoT Edge device can receive incoming requests from the downstream IoT Edge device. Open network traffic on ports 443 and 6617 for requests coming from the downstream device.

IoT Edge behind a gateway can't connect when migrating from one IoT hub to another

Symptoms

When attempting to migrate a hierarchy of IoT Edge devices from one IoT hub to another, the top level parent IoT Edge device can connect to IoT Hub, but downstream IoT Edge devices can't. The logs report Unable to authenticate client downstream-device/$edgeAgent with module credentials.

Cause

The credentials for the downstream devices weren't updated properly when the migration to the new IoT hub happened. Because of this, edgeAgent and edgeHub modules were set to have authentication type of none (default if not set explicitly). During connection, the modules on the downstream devices use old credentials, causing the authentication to fail.

Solution

When migrating to the new IoT hub (assuming not using DPS), follow these steps in order:

  1. Follow this guide to export and then import device identities from the old IoT hub to the new one
  2. Reconfigure all IoT Edge deployments and configurations in the new IoT hub
  3. Reconfigure all parent-child device relationships in the new IoT hub
  4. Update each device to point to the new IoT hub hostname (iothub_hostname under [provisioning] in config.toml)
  5. If you chose to exclude authentication keys during the device export, reconfigure each device with the new keys given by the new IoT hub (device_id_pk under [provisioning.authentication] in config.toml)
  6. Restart the top-level parent Edge device first, make sure it's up and running
  7. Restart each device in hierarchy level by level from top to the bottom

IoT Edge has low message throughput when geographically distant from IoT Hub

Symptoms

Azure IoT Edge devices that are geographically distant from Azure IoT Hub have a lower than expected message throughput.

Cause

High latency between the device and IoT Hub can cause a lower than expected message throughput. IoT Edge uses a default message batch size of 10. This limits the number of messages that are sent in a single batch, which increases the number of round trips between the device and IoT Hub.

Solution

Try increasing the IoT Edge Hub MaxUpstreamBatchSize environment variable. This allows more messages to be sent in a single batch, which reduces the number of round trips between the device and IoT Hub.

To set Azure Edge Hub environment variables in the Azure portal:

  1. Navigate to your IoT Hub and select Devices under the Device management menu.
  2. Select the IoT Edge device that you want to update.
  3. Select Set Modules.
  4. Select Runtime Settings.
  5. In the Edge Hub module settings tab, add the MaxUpstreamBatchSize environment variable as type Number with a value of 20.
  6. Select Apply.

Next steps

Do you think that you found a bug in the IoT Edge platform? Submit an issue so that we can continue to improve.

If you have more questions, create a Support request for help.