Creating multiple AWS environment sizes for different use cases

Project Details and Overview

A B2B application implemented as a single and multi-tenant environment on AWS with custom builds for mobile apps for both iOS and Android devices that communicate with the back-end services via API. Each application is custom-built for a specific customer and the AWS environment is created in an automated way with specific parameters. The application ecosystem is written in Go and consists of six different micro-services each playing a different role. All the micro-services communicate via gRPC and connect to a PostgreSQL database (except for the front-end application which is written in AngularJS).

Challenges

The client wanted to accelerate the production of their next-gen platforms, migrate their existing microservices from bare-metal servers to cloud-native servers, increase automation in infrastructure management, and optimize the distribution of different workloads.

The request was to build scalable, cost-efficient and a highly accessible platform for two use cases:
a) single-tenant environments
b) multi-tenant environment

Both environments should be created in a short period and to accommodate the expected workload.

For this purpose, three different sizes of environments should be available:
1) demo environment with the latest builds
2) medium environment where only core services should be highly available
3) high traffic environment where all the services are load-balanced and highly available

This would allow the client to choose a specific environment and its size in order to have everything up and running in a short period of time and minimize the AWS cost. CI/CD pipeline was required in order to allow for zero-downtime deployments. The challenge was to make sure that gRPC would work in an Auto-Scaling environment and that the different microservices can communicate with each other via gRPC.

Our Solution

We used Ansible for configuration management in order to provision the servers for each role and then used Packer to bake the initial Ansible plays into the AWS AMI images to create images that are ready to be used in Auto Scaling Groups.

Using Terraform we took care of provisioning and updating the AWS cloud infrastructure resources. We then combined all the elements into Jenkins where the client’s requirements were fine-grained to specific environments and infrastructure sizes. From here the client can run new deployments with zero downtime by connecting to AWS CodeDeploy.

An example of an architecture for the high-availability solution: