Building with AWS OpenSearch Serverless

This repo contains a CDK application that deploys a stack with an AWS OpenSearch Serverless application that is fed by a DynamoDB product database. There is a API Gateway with Lambda fronting OpenSearch for allowing people to do queries against the data.

The premise of this application was to explore the new (in preview at this time) OpenSearch Serverless application and deploy it in a realistic way to see how it works and evaluate the product.

System Overview

Deployment

If you have CDK installed globally then you can call cdk direct, if not you can use npx cdk - also see the Getting started guide.

1. Run npm install in the application directory.
2. Deploy with cdk deploy --all
3. Destroy with cdk destroy --all

Beware, OpenSearch Serverless is not cheap so don't run it for very long!

After deployment then you need to create an index. An Index cannot be created in the collection through CDK/CloudFormation - and rather than building something using a Custom Resource I added the index manually for this Spike:

• Once created then browse to the OpenSearch Service Console and go down to the Serverless menu.
• Select dashboard and launch the OpenSearch Dashboards URL
• Login with a valid access key
• In OpenSearch then browse to Management and Dev Tools
• On the left then enter PUT product-index and click run
• On the left menu, under Management choose Stack Management
• Click Index Patterns
• Click Create index pattern
• Type product-index in Index pattern name
• Click Next step
• Click Create index pattern

You can search in the Discover section when products are indexed:

Add Products

Once the index exists then you can add products through DynamoDB:

• Browse to DynamoDB
• Choose Explore items
• Select the table with the prefix AwsServerlessSearchStack-ProductCatalogueConstructProducts
• Click Create item
• Enter a unique productId
• Click Add new attribute and add a string called title and give it a value
• Repeat for description and price
• Click Create item

At this point the stream should be processed by lambda and the item indexed in search within a second. You can also update rows and delete rows for that to be reflected in the index. Obviously you can import to DynamoDB, use APIs, etc as well.

Try the API

In the frontend folder is a basic HTML page that you can run locally to test your API - you need to open the index.html file and update the API Gateway URL from your deployment.