Troubleshooting AWS IoT Analytics

See the following section to troubleshoot errors and find and possible solutions to resolve issues with AWS IoT Analytics.

How do I know if my messages are getting into AWS IoT Analytics?

Check if the rule to inject data into the channel through the rules-engine is configured correctly.

aws iot get-topic-rule --rule-name your-rule-name

The response should look like the following.

{
    "ruleArn": "arn:aws:iot:us-west-2:your-account-id:rule/your-rule-name",
    "rule": {
        "awsIotSqlVersion": "2016-03-23",
        "sql": "SELECT * FROM 'iot/your-rule-name'",
        "ruleDisabled": false,
        "actions": [
            {
                "iotAnalytics": {
                    "channelArn": "arn:aws:iotanalytics:region:your_account_id:channel/your-channel-name"
                }
            }
        ],
        "ruleName": "your-rule-name"
    }
}

Make sure the region and channel name used in the rule are correct. To ensure your data is reaching the rules engine and the rule is being executed correctly, you might want to add a new target to store incoming messages in the Amazon S3 bucket temporarily.

Why is my pipeline losing messages? How do I fix it?

• An activity has received an invalid JSON input:

  All activities, except Lambda activities, specially require a valid JSON string as input. If the JSON received by an activity is invalid, then the message is dropped and does not make its way into the data store. Make sure you are ingesting valid JSON messages into the service. In case of binary input, make sure the first activity in your pipeline is a Lambda activity that converts the binary data to valid JSON before passing it to the next activity or storing it in the data store. For more information, see Lambda function example 2.

• A Lambda function invoked by a Lambda activity has insufficient permissions:

  Make sure that each Lambda function in a Lambda activity has permission to be invoked from the AWS IoT Analytics service. You can use the following AWS CLI command to grant permission.

  aws lambda add-permission --function-name <name> --region <region> --statement-id <id> --principal iotanalytics.amazonaws.com --action lambda:InvokeFunction

• A filter or removeAttribute activity is incorrectly defined:

  Make sure the definitions if any filter or removeAttribute activities are correct. If you filter out a message or remove all attributes from a message, that message is not added to the data store.

Why is there no data in my data store?

• There is a delay between data ingestion and data availability:

  It might take several minutes after data is ingested into a channel before that data is available in the data store. The time varies based on the number of pipeline activities and the definition of any custom Lambda activities in your pipeline.

• Messages are being filtered out in your pipeline:

  Make sure you are not dropping messages in the pipeline. (See the previous question and response.)

• Your dataset query is incorrect:

  Make sure the query that generates the dataset from the data store is correct. Remove any unnecessary filters from the query to ensure your data reaches your data store.

Why does my dataset just show __dt?

• This column is added by the service automatically and contains the approximate ingestion time of the data. It may be used to optimize your queries. If your dataset contains nothing but this, see the previous question and response.

How do I code an event driven by the dataset completion?

• You must set up polling based on the describe-dataset command to check if the status of the dataset with a particular timestamp is SUCCEEDED.

How do I correctly configure my notebook instance to use AWS IoT Analytics?

Follow these steps to make sure the IAM role you are using to create the notebook instance has the required permissions:

1. Go to the SageMaker console and create a notebook instance.

2. Fill in the details and choose create a new role. Make a note of the role ARN.

3. Create the notebook instance. This also creates a role that SageMaker can use.

4. Go to the IAM console and modify the newly created SageMaker role. When you open that role, it should have a managed policy.

5. Click add inline policy, choose IoTAnalytics as the service, and under read permission, select GetDatasetContent.

6. Review the policy, add a policy name, and then create it. The newly created role now has policy permission to read a dataset from AWS IoT Analytics.

7. Go to AWS IoT Analytics console and create notebooks in the notebook instance.

8. Wait for the notebook instance to be in the "In Service" state.

9. Choose create notebooks, and select the notebook instance you created. This creates a Jupyter notebook with the selected template that can access your datasets.

Why can't I create notebooks in an instance?

• Make sure you create a notebook instance with the correct IAM policy. (Follow the steps in the previous question.)

• Make sure the notebook instance is in the "In Service" state. When you create an instance, it starts in a "Pending" state. It usually takes about five minutes for it to go into the "In Service" state. If the notebook instance goes into the "Failed" state after about five minutes, check the permissions again.

Why aren't I seeing my datasets in Amazon QuickSight?

Amazon QuickSight might need permission to read your AWS IoT Analytics dataset content. To give permission, follow these steps.

1. Choose your account name in the upper‐right corner of Amazon QuickSight and choose Manage QuickSight.

2. In the left navigation pane, choose Security & permissions. Under QuickSight access to AWS services, verify that access is granted to AWS IoT Analytics.

   1. If AWS IoT Analytics doesn't have access, choose Add or remove.

   2. Choose the box next to AWS IoT Analytics and then select Update. This gives Amazon QuickSight permission to read your dataset content.

3. Try again to visualize your data.

Make sure that you choose the same AWS Region for both AWS IoT Analytics and Amazon QuickSight. Otherwise, you might have issues accessing the AWS resources. For the list of supported Regions, see AWS IoT Analytics endpoints and quotas and Amazon QuickSight endpoints and quotas in the Amazon Web Services General Reference.

Why am I not seeing the containerize button on my existing Jupyter Notebook?

• This is caused by a missing AWS IoT Analytics Containerization Plugin. If you created you SageMaker notebook instance before August 23, 2018, you need to manually install the plugin by following the instructions in Containerizing a notebook.

• If you don't see the containerize button after creating the SageMaker notebook instance from the AWS IoT Analytics console or manually installing it, contact AWS IoT Analytics technical support.

Why is my containerization plugin installation failing?

• Usually, the plugin installation fails because of missing permissions in the SageMaker notebook instance. For the required permissions for the notebook instance, see Permissions and add the required permissions to the notebook instance role. If the problem persists, create a new notebook instance from the AWS IoT Analytics console.

• You can safely ignore the following message in the log if it appears during installation of the plugin: "To initialize this extension in the browser every time the notebook (or other app) loads."

Why is my containerization plugin throwing an error?

• Containerization can fail and generate errors for multiple reasons. Make sure that you're using the correct kernel before containerizing your notebook. Containerized kernels begin with the "Containerized" prefix.

• Since the plugin creates and saves a docker image in an ECR repository, make sure that your notebook instance role has sufficient permissions to read, list and create ECR repositories. For the required permissions for the notebook instance, see Permissions and add the required permissions to the notebook instance role.

• Also make sure that the name of the repository complies with ECR requirements. ECR repository names must start with a letter and can contain only lower-case letters, numbers, hyphens, underscores, and forward slashes.

• If the containerization process fails with the error:"This instance has insufficient free space to run containerization" try using a larger instance to resolve the issue.

• If you see connection errors or an image creation error, please retry. If the problem persists, restart the instance and install the latest plugin version.

Why don't I see my variables during the containerization?

• The AWS IoT Analytics containerization plugin automatically recognizes all variables in your notebook after it runs the notebook with the "Containerized" kernel. Use one of the containerized kernels to run the notebook, and then perform containerization.

What variables can I add to my container as an input?

• You can add any variable whose value you want to modify during the runtime as an input to your container. This enables your to run the same container with different parameters that need to be supplied at the time of dataset creation. The AWS IoT Analytics containerization Jupyter plugin simplifies this process by automatically recognizing the variables in the notebook and making them available as part of the containerization process.

How do I set my container output as an input for subsequent analysis?

• A specific S3 location where the executed artifacts can be stored is created for each run of your container dataset. To access this output location, create a variable with type outputFileUriValue in your container dataset. The value of this variable should be an S3 path that is used for storing your additional output files. To access these saved artifacts in subsequent runs, you can use the getDatasetContent API and pick the appropriate output file required for the subsequent run.

Why is my container dataset failing?

• Make sure that you're passing the correct executionRole to the container dataset. The trust policy of the executionRole must include both iotanalytics.amazonaws.com and sagemaker.amazonaws.com.

• If you see AlgorithmError as the reason for the failure, try to debug your container code manually. This happens if there is a bug in the container code or the execution role doesn't have permission to execute the container. If you containerized by using the AWS IoT Analytics Jupyter plugin, create a new SageMaker notebook instance with the same role as the executionRole of the containerDataset and try running the notebook manually. If the container was created outside of the Jupyter plugin, try manually running the code and limiting the permission to the executionRole.