Skip to content

Sample serverless application that reads events from Amazon Kinesis Data Streams and batches records to Amazon Timestream via Apache Flink app, visualizing results via Grafana.

License

Notifications You must be signed in to change notification settings

aws-samples/amazon-kinesis-timestream-grafana

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

22 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Amazon Timestream with Amazon Kinesis and Grafana Demo

Sample application that reads events from Amazon Kinesis Data Streams and batches records to Amazon Timestream, visualizing results via Grafana.

Overall architecture

The overall serverless architecture will work on events streamed to Amazon Kinesis Data Streams, pulled by an Apache Flink analytics application hosted on Amazon Kinesis Data Analytics to be batch-inserted in Amazon Timestream database. Finally, data will be visualized directly from the Timestream database by a Grafana dashboard using the Timestream datasource plugin.

Data sent from IoT device, to Amazon Kinesis Data Stream to Amazon Kinesis Data Analytics for Apache Flink to Amazon Timestream then visualized through Grafana

The sample setup will assume that events are being streamed via Amazon Kinesis service. However, this is not a precondition as any other streaming service like Kafka provisioned on EC2 or using Amazon Managed Streaming for Apache Kafka(Amazon MSK) can be used in a similar setup. To simulate devices streaming data to the Kinesis data stream, an AWS Lambda function produces events and pushes them to Kinesis data stream, the events streamed will contain information similar to the below sample.

DeviceID Timestamp temperature humidity voltage watt
b974f43a-2f04-11eb-adc1-0242ac120002 2020-09-13 11:49:42.5352820 15.5 70.3 39.7 301.44

In Timestream data will be modeled as follows

DeviceID (Dimension) measure_value::double measure_name measure_type time
b974f43a-2f04-11eb-adc1-0242ac120002 15.5 temperature DOUBLE 2020-09-13 11:49:42.5352820
b974f43a-2f04-11eb-adc1-0242ac120002 70.3 humidity DOUBLE 2020-09-13 11:49:42.5352820
b974f43a-2f04-11eb-adc1-0242ac120002 39.7 voltage DOUBLE 2020-09-13 11:49:42.5352820
b974f43a-2f04-11eb-adc1-0242ac120002 301.4 watt DOUBLE 2020-09-13 11:49:42.5352820

The Device ID is mapped as a Dimension and the Property fields measured mapped as measure_name , finally the value of the measure is mapped to the measure_value .datatype is set as double in this case. Check the best practices on mapping and data modeling for a better insight on mapping your data.

Getting started

1. Building and packaging Amazon Kinesis Data Analytics for Apache Flink application

This project provides 2 sample applications built with different toolsets. You can use either one of those as the application to be deployed.

  1. To build an application using Java and Apache Maven, refer to instructions here
  2. To build an application using Kotlin and Gradle, refer to instructions here

2. Deploy infrastructure

Infrastructure deployment will automatically use packaged application jar and upload it to an Amazon S3 bucket. The infrastructure utilizes multiple stacks built using an AWS CDK project with Python3 language. For more information on working with the CDK and Python, check the following guide. To deploy all stacks use the --all option when invoking cdk deploy

  1. Navigate to cdk folder
  2. Follow instructions here to create virtual environment and build stacks
  3. Make sure CDK environment is bootstrapped in the account and region you're deploying stacks to, as the stacks utilize assets to deploy the Kinesis Data Analytics Flink application
    $ cdk bootstrap
  4. Deploy infrastructure using packaged applications
    1. To deploy infrastructure and use Java application as basis for Kinesis analytics application, you can directly deploy the CDK stacks
      $ cdk deploy --all
    2. To deploy infrastructure and use Kotlin application as basis for Kinesis analytics application, you can customize the stacks using context variables
      $ cdk deploy --context kda_path=../analytics-kotlin/build/libs/analytics-timestream-kotlin-sample-all.jar --all

Check deployed resource and kick-off pipeline

Once CDK stacks are deployed successfully you can check created AWS resources. You can directly run script ./setup.sh or follow below instructions.

  1. Amazon Kinesis Data Stream

    Deployed through stack amazon-kinesis-stream-stack and is ready to receive events from sample producer.

    Amazon EventBridge 1 min rate event triggers AWS Lambda to send JSON data to Amazon Kinesis Data Stream

    Producer resources are deployed through stack sample-kinesis-stream-producer. You can check the Lambda function monitoring and logs to make sure it's being regularly called and sending events to the stream.

    For more information on how the producer works check documentation.

  2. Amazon Kinesis Data Analytics for Apache Flink Application

    Deployed through stack amazon-kinesis-analytics. Although the application is created through the stack it would still not be running.

    To run the application and kick-off the pipeline, simply pickup the application name from the stack output KdaApplicationName.

    Follow instructions under Run the Application section or run the following command

    $ aws kinesisanalyticsv2 start-application --application-name amazon-kinesis-analytics \
      --run-configuration '{ "ApplicationRestoreConfiguration": { "ApplicationRestoreType": "SKIP_RESTORE_FROM_SNAPSHOT" } }'
  3. Amazon S3 bucket to store the JAR package for the application and application role

    Deployed through stack flink-source-bucket

  4. Amazon Timestream Database

    Deployed through stack amazon-timestream

  5. Grafana dashboard deployment

    Deployed through stack grafana. To check the created Grafana check the output MyFargateServiceServiceURL....

    1. Grafana deployed using Amazon ECS on AWS Fargate for compute
    2. Amazon Elastic File System (Amazon EFS) for storage
    3. AWS SecretsManager Secret storing Grafana admin user password. Check the stack output GrafanaAdminSecret for AWS Secrets Manager Secret's id storing the Grafana admin user's password.

3. Visualizing Amazon Timestream values through Grafana

To help you get started with data visualization, we have created a sample dashboard in Grafana that visualizes data sent to Timestream from the sample producer. If you invoked the setup.sh script, it would have automatically performed these steps.

You can also check the following video tutorial or complete guide for more information.

  1. Install Grafana Datasource plugin

    The Grafana Amazon Timestream datasource plugin is automatically installed through the infrastructure stack.

  2. Create Grafana Datasource

    To create a Timeseries data source, go to Datasources, click on Add Datasource, search for Timestream, and select the Timestream datasource. You can also use programmatic means described here.

    To use the programmatic means to create a data source:

    1. Create an API token to be used. Make sure to replace <stack grafana.GrafanaAdminSecret.secret_value> with the sceret value, escaping any characters in the password like $ or " with a preceding \. And replace <stack grafana.MyFargateServiceServiceURL...> with the value from stack output.
    $ grafana_token=$(\
     curl -X POST -u "admin:<stack grafana.GrafanaAdminSecret.secret_value>" \
     -H "Content-Type: application/json" \
     -d '{"name":"apikeycurl", "role": "Admin"}' \
     <stack grafana.MyFargateServiceServiceURL...>/api/auth/keys \
     | jq -r .key)
    1. Create Amazon Timestream datasource
    $ curl -X POST --insecure \
     -H "Content-Type: application/json" -H "Authorization: Bearer ${grafana_token}" \
     -d @./cdk/stacks/grafana/datasource.json \
     <stack grafana.MyFargateServiceServiceURL...>/api/datasources
  3. Create Grafana Dashboard

    Grafana dashboards can be exported and imported as JSON. You can find a dashboard sample json in here. The defined dashboard provides sample variables and visualizations by querying directly from Timestream database. The dashboard assumes the Timestream datasource is your default datasource.

    To import sample dashboard you can follow instructions here or use programmatic means described here.

    To use the programmatic means to create a dashboard:

    $ curl -X POST --insecure \
     -H "Content-Type: application/json" -H "Authorization: Bearer ${grafana_token}" \
     -d @./cdk/stacks/grafana/dashboard.json \
     <stack grafana.MyFargateServiceServiceURL...>/api/dashboards/db
    1. To check created dashboard navigate to
    2. Login using admin credentials.
    3. Check created dashboard under "Dashboards" -> "Manage" Grafana dashboards screenshot

Clean up

To delete all created stack resources you can use

  $ cdk destroy --all

Security

See CONTRIBUTING for more information.

License

This library is licensed under the MIT-0 License. See the LICENSE file.

About

Sample serverless application that reads events from Amazon Kinesis Data Streams and batches records to Amazon Timestream via Apache Flink app, visualizing results via Grafana.

Topics

Resources

License

Code of conduct

Security policy

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published