Muokkaa

Jaa

Telemetry and troubleshooting

  • Artikkeli

Important

On 30 March 2025, Azure AI Vision Spatial Analysis will be retired. Please transition to Azure AI Video Indexer or another open-source solution before the specified date. We encourage you to make the switch sooner to gain the richer benefits of Azure AI Video Indexer. In addition to the familiar features you are using, here's a quick comparison between Azure AI Vision Spatial Analysis and Azure AI Video Indexer.

Feature Azure AI Vision Spatial Analysis Azure AI Video Indexer
Edge support Yes Yes
Object Detection People & Vehicle detection only Detects 1000+ objects
Audio/Speech Processing Not supported Supported (includes speech transcription, translation and summarization)
Supported >(includes speech transcription and sentiment analysis)
Event Detection & Tracking Supported (tracking people & vehicles, event detection) Not supported at the Edge yet. Is partially supported at the Cloud.
Azure Arc Support Not supported Native support
Focus Area Visual analysis with specialized tracking Comprehensive analysis of both audio and visual content

From now until 30 March 2025, you can continue to use Azure AI Vision Spatial Analysis or transition to Azure AI Video Indexer before the specified date. After 30 March 2025, the Spatial Analysis container will no longer be supported and will stop processing new streams.

Spatial Analysis includes a set of features to monitor the health of the system and help with diagnosing issues.

Enable visualizations

To enable a visualization of AI Insight events in a video frame, you need to use the .debug version of a Spatial Analysis operation on a desktop machine or Azure VM. The visualization isn't possible on Azure Stack Edge devices. There are four debug operations available.

If your device is a local desktop machine or Azure GPU VM (with remote desktop enabled), then you can switch to .debug version of any operation and visualize the output.

  1. Open the desktop either locally or by using a remote desktop client on the host computer running Spatial Analysis.

  2. In the terminal run xhost +

  3. Update the deployment manifest under the spaceanalytics module with the value of the DISPLAY environment variable. You can find its value by running echo $DISPLAY in the terminal on the host computer.

    "env": {        
    "DISPLAY": {
        "value": ":11"
        }
}

  4. Update the graph in the deployment manifest you want to run in debug mode. In the example below, we update the operationId to cognitiveservices.vision.spatialanalysis-personcrossingpolygon.debug. A new parameter VISUALIZER_NODE_CONFIG is required to enable the visualizer window. All operations are available in debug flavor. When using shared nodes, use the cognitiveservices.vision.spatialanalysis.debug operation and add VISUALIZER_NODE_CONFIG to the instance parameters.

    "zonecrossing": {
     "operationId" : "cognitiveservices.vision.spatialanalysis-personcrossingpolygon.debug",
     "version": 1,
     "enabled": true,
     "parameters": {
         "VIDEO_URL": "Replace http url here",
         "VIDEO_SOURCE_ID": "zonecrossingcamera",
         "VIDEO_IS_LIVE": false,
        "VIDEO_DECODE_GPU_INDEX": 0,
         "DETECTOR_NODE_CONFIG": "{ \"gpu_index\": 0 }",
        "CAMERACALIBRATOR_NODE_CONFIG": "{ \"gpu_index\": 0}",
        "VISUALIZER_NODE_CONFIG": "{ \"show_debug_video\": true }",
         "SPACEANALYTICS_CONFIG": "{\"zones\":[{\"name\":\"queue\",\"polygon\":[[0.3,0.3],[0.3,0.9],[0.6,0.9],[0.6,0.3],[0.3,0.3]], \"threshold\":35.0}]}"
     }
}

  5. Redeploy and you'll see the visualizer window on the host computer

  6. After the deployment has completed, you might have to copy the .Xauthority file from the host computer to the container and restart it. In the sample below, peopleanalytics is the name of the container on the host computer.

    sudo docker cp $XAUTHORITY peopleanalytics:/root/.Xauthority
sudo docker stop peopleanalytics
sudo docker start peopleanalytics
xhost +

Collect system health telemetry

Telegraf is an open source image that works with Spatial Analysis, and is available in the Microsoft Container Registry. It takes the following inputs and sends them to Azure Monitor. The Telegraf module can be built with desired custom inputs and outputs. The Telegraf module configuration in Spatial Analysis is part of the deployment manifest (linked above). This module is optional and can be removed from the manifest if you don't need it.

Inputs:

  • Spatial Analysis Metrics
  • Disk Metrics
  • CPU Metrics
  • Docker Metrics
  • GPU Metrics

Outputs:

  • Azure Monitor

The supplied Spatial Analysis Telegraf module publishes all the telemetry data emitted by the Spatial Analysis container to Azure Monitor. See the Azure Monitor for information on adding Azure Monitor to your subscription.

After setting up Azure Monitor, you'll need to create credentials that enable the module to send telemetry. You can use the Azure portal to create a new Service Principal, or use the Azure CLI command below to create one.

Note

This command requires you to have Owner privileges on the subscription.

# Find your Azure IoT Hub resource ID by running this command. The resource ID  should start with something like 
# "/subscriptions/b60d6458-1234-4be4-9885-c7e73af9ced8/resourceGroups/..."
az iot hub list

# Create a Service Principal with `Monitoring Metrics Publisher` role in the IoTHub resource:
# Save the output from this command. The values will be used in the deployment manifest. The password won't be shown again so make sure to write it down
az ad sp create-for-rbac --role="Monitoring Metrics Publisher" --name "<principal name>" --scopes="<resource ID of IoT Hub>"

In the deployment manifest for your Azure Stack Edge device, desktop machine, or Azure VM with GPU, look for the Telegraf module, and replace the following values with the Service Principal information from the previous step and redeploy.

"Telegraf": { 
  "settings": {
  "image":   "mcr.microsoft.com/azure-cognitive-services/vision/spatial-analysis/Telegraf:1.0",
  "createOptions":   "{\"HostConfig\":{\"Runtime\":\"nvidia\",\"NetworkMode\":\"azure-iot-edge\",\"Memory\":33554432,\"Binds\":[\"/var/run/docker.sock:/var/run/docker.sock\"]}}"
},
"type": "docker",
"env": {
    "AZURE_TENANT_ID": {
        "value": "<Tenant Id>"
    },
    "AZURE_CLIENT_ID": {
        "value": "Application Id"
    },
    "AZURE_CLIENT_SECRET": {
        "value": "<Password>"
    },
    "region": {
        "value": "<Region>"
    },
    "resource_id": {
        "value": "/subscriptions/{subscriptionId}/resourceGroups/{resoureGroupName}/providers/Microsoft.Devices/IotHubs/{IotHub}"
    },
...

Once the Telegraf module is deployed, the reported metrics can be accessed either through the Azure Monitor service, or by selecting Monitoring in the IoT Hub on the Azure portal.

Azure Monitor telemetry report

System health events

Event Name Description
archon_exit Sent when a user changes the Spatial Analysis module status from running to stopped.
archon_error Sent when any of the processes inside the container crash. This is a critical error.
InputRate The rate at which the graph processes video input. Reported every five minutes.
OutputRate The rate at which the graph outputs AI insights. Reported every five minutes.
archon_allGraphsStarted Sent when all graphs have finished starting up.
archon_configchange Sent when a graph configuration has changed.
archon_graphCreationFailed Sent when the graph with the reported graphId fails to start.
archon_graphCreationSuccess Sent when the graph with the reported graphId starts successfully.
archon_graphCleanup Sent when the graph with the reported graphId cleans up and exits.
archon_graphHeartbeat Heartbeat sent every minute for every graph of a skill.
archon_apiKeyAuthFail Sent when the Vision resource key fails to authenticate the container for more than 24 hours, due to the following reasons: Out of Quota, Invalid, Offline.
VideoIngesterHeartbeat Sent every hour to indicate that video is streamed from the Video source, with the number of errors in that hour. Reported for each graph.
VideoIngesterState Reports Stopped or Started for video streaming. Reported for each graph.

Troubleshooting an IoT Edge Device

You can use iotedge command line tool to check the status and logs of the running modules. For example:

  • iotedge list: Reports a list of running modules. You can further check for errors with iotedge logs edgeAgent. If iotedge gets stuck, you can try restarting it with iotedge restart edgeAgent
  • iotedge logs <module-name>
  • iotedge restart <module-name> to restart a specific module

Collect log files with the diagnostics container

Spatial Analysis generates Docker debugging logs that you can use to diagnose runtime issues, or include in support tickets. The Spatial Analysis diagnostics module is available in the Microsoft Container Registry for you to download. In the manifest deployment file for your Azure Stack Edge Device, desktop machine, or Azure VM with GPU look for the diagnostics module.

In the "env" section, add the following configuration:

"diagnostics": {  
  "settings": {
  "image":   "mcr.microsoft.com/azure-cognitive-services/vision/spatial-analysis/diagnostics:1.0",
  "createOptions":   "{\"HostConfig\":{\"Mounts\":[{\"Target\":\"/usr/bin/docker\",\"Source\":\"/home/data/docker\",\"Type\":\"bind\"},{\"Target\":\"/var/run\",\"Source\":\"/run\",\"Type\":\"bind\"}],\"LogConfig\":{\"Config\":{\"max-size\":\"500m\"}}}}"
  }

To optimize logs uploaded to a remote endpoint, such as Azure Blob Storage, we recommend maintaining a small file size. See the example below for the recommended Docker logs configuration.

{
    "HostConfig": {
        "LogConfig": {
            "Config": {
                "max-size": "500m",
                "max-file": "1000"
            }
        }
    }
}

Configure the log level

Log level configuration allows you to control the verbosity of the generated logs. Supported log levels are: none, verbose, info, warning, and error. The default log verbose level for both nodes and platform is info.

Log levels can be modified globally by setting the ARCHON_LOG_LEVEL environment variable to one of the allowed values. It can also be set through the IoT Edge Module Twin document either globally, for all deployed skills, or for every specific skill by setting the values for platformLogLevel and nodesLogLevel as shown below.

{
    "version": 1,
    "properties": {
        "desired": {
            "globalSettings": {
                "platformLogLevel": "verbose"
            },
            "graphs": {
                "samplegraph": {
                    "nodesLogLevel": "verbose",
                    "platformLogLevel": "verbose"
                }
            }
        }
    }
}

Collecting Logs

Note

The diagnostics module does not affect the logging content, it is only assists in collecting, filtering, and uploading existing logs. You must have Docker API version 1.40 or higher to use this module.

The sample deployment manifest file for your Azure Stack Edge device, desktop machine, or Azure VM with GPU includes a module named diagnostics that collects and uploads logs. This module is disabled by default and should be enabled through the IoT Edge module configuration when you need to access logs.

The diagnostics collection is on-demand and controlled via an IoT Edge direct method, and can send logs to an Azure Blob Storage.

Configure diagnostics upload targets

From the IoT Edge portal, select your device and then the diagnostics module. In the sample Deployment manifest file for your Azure Stack Edge device, desktop machines, or Azure VM with GPU look for the Environment Variables section for diagnostics, named env, and add the following information:

Configure Upload to Azure Blob Storage

  1. Create your own Azure Blob Storage account, if you haven't already.
  2. Get the Connection String for your storage account from the Azure portal. It is located in Access Keys.
  3. Spatial Analysis logs are automatically uploaded into a Blob Storage container named rtcvlogs with the following file name format: {CONTAINER_NAME}/{START_TIME}-{END_TIME}-{QUERY_TIME}.log.
"env":{
    "IOTEDGE_WORKLOADURI":"fd://iotedge.socket",
    "AZURE_STORAGE_CONNECTION_STRING":"XXXXXX",   //from the Azure Blob Storage account
    "ARCHON_LOG_LEVEL":"info"
}

Uploading Spatial Analysis logs

Logs are uploaded on-demand with the getRTCVLogs IoT Edge method, in the diagnostics module.

  1. Go to your IoT Hub portal page, select Edge Devices, then select your device and your diagnostics module.
  2. Go to the details page of the module and select the direct method tab.
  3. Type getRTCVLogs on Method Name, and a json format string in payload. You can enter {}, which is an empty payload.
  4. Set the connection and method timeouts, and select Invoke Method.
  5. Select your target container, and build a payload json string using the parameters described in the Logging syntax section. Select Invoke Method to perform the request.

Note

Invoking the getRTCVLogs method with an empty payload will return a list of all containers deployed on the device. The method name is case sensitive. You will get a 501 error if an incorrect method name is given.

Invoking the getRTCVLogs method getRTCVLogs Direct method page

Logging syntax

The below table lists the parameters you can use when querying logs.

Keyword Description Default Value
StartTime Desired logs start time, in milliseconds UTC. -1, the start of the container's runtime. When [-1.-1] is used as a time range, the API returns logs from the last one hour.
EndTime Desired logs end time, in milliseconds UTC. -1, the current time. When [-1.-1] time range is used, the API returns logs from the last one hour.
ContainerId Target container for fetching logs. null, when there is no container ID. The API returns all available containers information with IDs.
DoPost Perform the upload operation. When this is set to false, it performs the requested operation and returns the upload size without performing the upload. When set to true, it initiates the asynchronous upload of the selected logs false, do not upload.
Throttle Indicate how many lines of logs to upload per batch 1000, Use this parameter to adjust post speed.
Filters Filters logs to be uploaded null, filters can be specified as key value pairs based on the Spatial Analysis logs structure: [UTC, LocalTime, LOGLEVEL,PID, CLASS, DATA]. For example: {"TimeFilter":[-1,1573255761112]}, {"TimeFilter":[-1,1573255761112]}, {"CLASS":["myNode"]

The following table lists the attributes in the query response.

Keyword Description
DoPost Either true or false. Indicates if logs have been uploaded or not. When you choose not to upload logs, the API returns information synchronously. When you choose to upload logs, the API returns 200, if the request is valid, and starts uploading logs asynchronously.
TimeFilter Time filter applied to the logs.
ValueFilters Keywords filters applied to the logs.
TimeStamp Method execution start time.
ContainerId Target container ID.
FetchCounter Total number of log lines.
FetchSizeInByte Total amount of log data in bytes.
MatchCounter Valid number of log lines.
MatchSizeInByte Valid amount of log data in bytes.
FilterCount Total number of log lines after applying filter.
FilterSizeInByte Total amount of log data in bytes after applying filter.
FetchLogsDurationInMiliSec Fetch operation duration.
PaseLogsDurationInMiliSec Filter operation duration.
PostLogsDurationInMiliSec Post operation duration.

Example request

{
    "StartTime": -1,
    "EndTime": -1,
    "ContainerId": "5fa17e4d8056e8d16a5a998318716a77becc01b36fde25b3de9fde98a64bf29b",
    "DoPost": false,
    "Filters": null
}

Example response

{
    "status": 200,
    "payload": {
        "DoPost": false,
        "TimeFilter": [-1, 1581310339411],
        "ValueFilters": {},
        "Metas": {
            "TimeStamp": "2020-02-10T04:52:19.4365389+00:00",
            "ContainerId": "5fa17e4d8056e8d16a5a998318716a77becc01b36fde25b3de9fde98a64bf29b",
            "FetchCounter": 61,
            "FetchSizeInByte": 20470,
            "MatchCounter": 61,
            "MatchSizeInByte": 20470,
            "FilterCount": 61,
            "FilterSizeInByte": 20470,
            "FetchLogsDurationInMiliSec": 0,
            "PaseLogsDurationInMiliSec": 0,
            "PostLogsDurationInMiliSec": 0
        }
    }
}

Check fetch log's lines, times, and sizes, if those settings look good replace DoPost to true and that will push the logs with same filters to destinations.

You can export logs from the Azure Blob Storage when troubleshooting issues.

Troubleshooting the Azure Stack Edge device

The following section is provided for help with debugging and verification of the status of your Azure Stack Edge device.

Access the Kubernetes API Endpoint. 

  1. In the local UI of your device, go to the Devices page.
  2. Under Device endpoints, copy the Kubernetes API service endpoint. This endpoint is a string in the following format: https://compute..[device-IP-address].
  3. Save the endpoint string. You'll use this later when configuring kubectl to access the Kubernetes cluster.

Connect to PowerShell interface

Remotely, connect from a Windows client. After the Kubernetes cluster is created, you can manage the applications via this cluster. You'll need to connect to the PowerShell interface of the device. Depending on the operating system of client, the procedures to remotely connect to the device may be different. The following steps are for a Windows client running PowerShell.

Tip

  • Before you begin, make sure that your Windows client is running Windows PowerShell 5.0 or later.
  • PowerShell is also available on Linux.

  1. Run a Windows PowerShell session as an Administrator.

    Make sure that the Windows Remote Management service is running on your client. At the command prompt, type winrm quickconfig.

  2. Assign a variable for the device IP address. For example, $ip = "<device-ip-address>".

  3. Use the following command to add the IP address of your device to the client's trusted hosts list.

    Set-Item WSMan:\localhost\Client\TrustedHosts $ip -Concatenate -Force

  4. Start a Windows PowerShell session on the device.

    Enter-PSSession -ComputerName $ip -Credential $ip\EdgeUser -ConfigurationName Minishell

  5. Provide the password when prompted. Use the same password that is used to sign into the local web interface. The default local web interface password is Password1.

Access the Kubernetes cluster

After the Kubernetes cluster is created, you can use the kubectl command line tool to access the cluster.

  1. Create a new namespace.

    New-HcsKubernetesNamespace -Namespace

  2. Create a user and get a config file. This command outputs configuration information for the Kubernetes cluster. Copy this information and save it in a file named config. Do not save the file a file extension.

    New-HcsKubernetesUser -UserName

  3. Add the config file to the .kube folder in your user profile on the local machine.

  4. Associate the namespace with the user you created.

    Grant-HcsKubernetesNamespaceAccess -Namespace -UserName

  5. Install kubectl on your Windows client using the following command:

    curl https://storage.googleapis.com/kubernetesrelease/release/v1.15.2/bin/windows/amd64/kubectl.exe -O kubectl.exe

  6. Add a DNS entry to the hosts file on your system.

    1. Run Notepad as administrator and open the hosts file located at C:\windows\system32\drivers\etc\hosts.
    2. Create an entry in the hosts file with the device IP address and DNS domain you got from the Device page in the local UI. The endpoint you should use will look similar to: https://compute.asedevice.microsoftdatabox.com/10.100.10.10.

  7. Verify you can connect to the Kubernetes pods.

    kubectl get pods -n "iotedge"

To get container logs, run the following command:

kubectl logs <pod-name> -n <namespace> --all-containers

Useful commands

Command Description
Get-HcsKubernetesUserConfig -AseUser Generates a Kubernetes configuration file. When using the command, copy the information into a file named config. Don't save the file with a file extension.
Get-HcsApplianceInfo Returns information about your device.
Enable-HcsSupportAccess Generates access credentials to start a support session.

How to file a support ticket for Spatial Analysis

If you need more support in finding a solution to a problem you're having with the Spatial Analysis container, follow these steps to fill out and submit a support ticket. Our team will get back to you with further guidance.

Fill out basic information

Create a new support ticket at the New support request page. Follow the prompts to fill in the following parameters:

Support basics

  1. Set Issue Type to be Technical.
  2. Select the subscription that you're utilizing to deploy the Spatial Analysis container.
  3. Select My services and select Azure AI services as the service.
  4. Select the resource that you're utilizing to deploy the Spatial Analysis container.
  5. Write a brief description detailing the problem you're facing.
  6. Select Spatial Analysis as your problem type.
  7. Select the appropriate subtype from the drop-down.
  8. Select Next: Solutions to move on to the next page.

The next stage will offer recommended solutions for the problem type that you selected. These solutions will solve the most common problems, but if it isn't useful for your solution, select Next: Details to go to the next step.

Details

On this page, add some extra details about the problem you've been facing. Be sure to include as much detail as possible, as this will help our engineers better narrow down the issue. Include your preferred contact method and the severity of the issue so we can contact you appropriately, and select Next: Review + create to move to the next step.

Review and create

Review the details of your support request to ensure everything is accurate and represents the problem effectively. Once you're ready, select Create to send the ticket to our team! You'll receive an email confirmation once your ticket is received, and our team will work to get back to you as soon as possible. You can view the status of your ticket in the Azure portal.

Next steps