The application reads data from a DynamoDB database. It's the only app that we own that does so, and it's the only one that seems to experience a memory leak. So, we decided to leave this bit as the issue might be in New Relic's instrumentation for the AWS SDK.
We include a terraform script to create the database in deployments/infrastructure. Adjust the region and table name as desired, and simply run terraform init and terraform apply.
We also include a JS script to populate the database with some sample data. Go into deployments/db-population and run:
npm init
node ./populate.jsYou might need to set the right AWS credentials to conduct the previous steps.
We ran the application on a kubernetes cluster. We have not included any configuration in this repo, as we believe the exact configuration you would need to use will most likely differ substantially. In any case, here are some details that might be useful:
- We use an ALB to expose the application
- Resource limits:
- Memory: 500Mi
- Resource requests:
- cpu: 300m
- memory: 50Mi
The application needs the following environment variables to be specified:
- NRML_TABLENAME: DynamoDB table name
- NRML_AWS_REGION: AWS region where the table is
- NRML_NEWRELIC_LICENSE: New Relic license key
See this README file for details on how to generate traffic for the application.
You can enable pprof by setting the environment variable NRML_ENABLE_PPROF to true. This uses gin-contrib/pprof. See more details here.