SaaS Starter Kit

Copyright 2019, Geeks Accelerator twins@geeksaccelerator.com

Sponsored by Copper Valley Telecom

The SaaS Starter Kit is a set of libraries for building scalable software-as-a-service (SaaS) applications that helps preventing both misuse and fraud. The goal of this project is to provide a proven starting point for new projects that reduces the repetitive tasks in getting a new project launched to production that can easily be scaled and ready to onboard enterprise clients. It uses minimal dependencies, implements idiomatic code and follows Golang best practices. Collectively, the toolkit lays out everything logically to minimize guess work and enable engineers to quickly load a mental model for the project. This inturn will make current developers happy and expedite on-boarding of new engineers.

This project should not be considered a web framework. It is a starter toolkit that provides a set of working examples to handle some of the common challenges for developing SaaS using Golang. Coding is a discovery process and with that, it leaves you in control of your project’s architecture and development.

SaaS product offerings generally provide a web-based service using a subscription model. They typically provide at least two main components: a REST API and a web application.

To see screen captures of the web app and auto-generated API documentation, check out this Google Slides deck: https://docs.google.com/presentation/d/1WGYqMZ-YUOaNxlZBfU4srpN8i86MU0ppWWSBb3pkejM/edit#slide=id.p

You are welcome to add comments to the Google Slides.

Motivation

When getting started building SaaS, we believe that is important for both the frontend web experience and the backend business logic (business value) be developed in the same codebase - using the same language for the frontend and backend development in the same single repository. We believe this for two main reasons:

1. Lower barrier for and accelerate onboarding of new engineers developing the SaaS by making it easy for them to load a complete mental model of the codebase.
2. Minimize potential bottlenecks and eliminate complexities of coordinating development across repositories, with potentially different teams responsible for the different repositories.

Once the SaaS product has gained market traction and the core set of functionality has been identified to achieve product-market fit, the functionality could be re-written with a language that would improve user experience or further increase efficiency. Two good examples of this would be:

1. Developing an iPhone or Android app. The front end web application provided by this project is responsive to support mobile devices. However, there may be a point that developing native would provide an enhanced experience.
2. The backend business logic has a set of methods that handle small data transformations on a massive scale. If the code for this is relatively small and can easily be rewritten, it might make sense to rewrite this directly in C or Rust. This is a very rare case as GoLang is already a preformat language.

There are five areas of expertise that an engineer or engineering team must do for a project to grow and scale. Based on our experience, a few core decisions were made for each of these areas that help you focus initially on building the business logic.

1. Micro level - The semantics that cover how data is defined, the relationships and how the data is being captured. This project tries to minimize the connection between packages on the same horizontally later. Data models should not be part of feature functionality. Hopefully these micro level decisions help prevent cases where 30K lines of code rely on a single data model which makes simple one line changes potentially high risk.
2. Macro level - The architecture and its design provides basic project structure and the foundation for development. This project provides a good set of examples that demonstrate where different types of code can reside.
3. Business logic - The code for the business logic facilitates value generating activities for the business. This project provides an example Golang package that helps illustrate the implementation of business logic and how it can be delivered to clients.
4. Deployment and Operations - Get the code to production! This sometimes can be a challenging task as it requires a knowledge of a completely different expertise - DevOps. This project provides a complete continuous build pipeline that will push the code to production with minimal effort using serverless deployments to AWS Fargate with GitLab CI/CD.
5. Observability - Ensure the code is running as expected in a remote environment. This project implements Datadog to facilitate exposing metrics, logs and request tracing to obverse and validate your services are stable and responsive for your clients (hopefully paying clients).

Description

The example project is a complete starter kit for building SasS with GoLang. It provides two example services:

• Web App - Responsive web application to provide service to clients. Includes user signup and user authentication for direct client interaction via their web browsers.
• Web API - REST API with JWT authentication that renders results as JSON. This allows clients and other third-party companies to develop deep integrations with the project.

The example project also provides these tools:

• Schema - Creating, initializing tables of Postgres database and handles schema migration.
• Dev Ops - Deploying project to AWS with GitLab CI/CD.

It contains the following features:

• Minimal web application using standard html/template package.
• Auto-documented REST API.
• Middleware integration.
• Database support using Postgres.
• Cache and key value store using Redis
• CRUD based pattern.
• Role-based access control (RBAC).
• Account signup and user management.
• Distributed logging and tracing.
• Integration with Datadog for enterprise-level observability.
• Testing patterns.
• Build, deploy and run application using Docker, Docker Compose, and Makefiles.
• Vendoring dependencies with Modules, requires Go 1.12 or higher.
• Continuous deployment pipeline.
• Serverless deployments with AWS ECS Fargate.
• CLI with boilerplate templates to reduce repetitive copy/pasting.
• Integration with GitLab for enterprise-level CI/CD.

Accordingly, the project architecture is illustrated with the following diagram.

Example project

With SaaS, a client subscribes to an online service you provide them. The example project provides functionality for clients to subscribe and then once subscribed they can interact with your software service.

The initial contributors to this project are building this saas-starter-kit based on their years of experience building enterprise B2B SaaS. Particularly, this saas-starter-kit is based on their most recent experience building the B2B SaaS for standard operating procedure software (written entirely in Golang). Please refer to the Keeni.Space website, its SOP software pricing and its signup process. The SaaS web app is then available at app.keeni.space. They plan on leveraging this experience and build it into a simplified set example services for both a web API and a web app for SaaS businesses.

For this example, projects will be the single business logic package that will be exposed to users for management based on their role. Additional business logic packages can be added to support your project. It’s important at the beginning to minimize the connection between business logic packages on the same horizontal level.

This project provides the following functionality to users:

New clients can sign up which creates an account and a user with role of admin.

• Users with the role of admin can manage users for their account.
• Authenticated users can manage their projects based on RBAC.

The project implements RBAC with two basic roles for users: admin and user.

• The role of admin provides the ability to perform all CRUD actions on projects and users.
• The role of user limits users to only view projects and users.

Of course, this example implementation of RBAC can be modified and enhanced to meet your requirements.

The project groups code in three distinct directories:

• Cmd - all application stuff (routes and http transport)
• Internal - all business logic (compiler protections)
• Platform - all foundation stuff (kit)

All business logic should be contained as a package inside the internal directory. This enables both the web app and web API to use the same API (Golang packages) with the only main difference between them is their response, HTML or JSON.

The web API and web app have similar functionality. While the web app is meant for humans to experience and requires a friendly UI, the web API is meant for customers or third-party partners of your SaaS to programmatically integrate. To help show the similarities and differences between the pages in the web app and similar endpoints in the web API, we have created this diagram below. Since it is very detailed, you can click on the image to see the larger version.

Local Installation

Docker is required to run this project on your local machine. This project uses multiple third-party services that will be hosted locally via Docker.

• Postgres - Transactional database to handle persistence of all data.
• Redis - Key / value storage for sessions and other data. Used only as ephemeral storage.
• Datadog - Provides metrics, logging, and tracing.

An AWS account is required for deployment for the following AWS dependencies:

• Secret Manager - Provides store for private key used for JWT.
• S3 - Host static files on S3 with additional CDN support with CloudFront.
• ECS Fargate - Serverless deployments of application.
• RDS - Cloud hosted version of Postgres.
• Route 53 - Management of DNS entries.

Getting the project

Clone the repo into your desired location. This project uses Go modules and does not need to be in your GOPATH. You will need to be using Go >= 1.11.

You should now be able to clone the project.

$ git clone git@gitlab.com:geeks-accelerator/oss/saas-starter-kit.git
$ cd saas-starter-kit/

If you have Go Modules enabled, you should be able compile the project locally. If you have Go Modules disabled, see the next section.

Go Modules

This project is using Go Module support for vendoring dependencies.

We are using the tidy command to maintain the dependencies and make sure the project can create reproducible builds.

$ GO111MODULE=on go mod tidy

It is recommended to use at least Go 1.12 and enable go modules.

$ echo "export  GO111MODULE=on" >> ~/.bash_profile

Installing Docker

Docker is a critical component and required to run this project.

https://docs.docker.com/install/

Running The Project

There is a docker-compose file that knows how to build and run all the services. Each service has its own a dockerfile.

Before using docker-compose, you need to copy sample.env_docker_compose to .env_docker_compose that docker will use. When you run docker-compose up it will run all the services including the main.go file for each Go service. The following services will run:

• web-api
• web-app
• postgres
• mysql

Running the project

Use the docker-compose.yaml to run all of the services, including the 3rd party services. The first time to run this command, Docker will download the required images for the 3rd party services.

$ cp sample.env_docker_compose .env_docker_compose
$ docker-compose up

Default configuration is set which should be valid for most systems.

Use the docker-compose.yaml file to configure the services differently using environment variables when necessary.

How we run the project

We like to run the project where the services run in the background of our CLI. This can be done by using the -d with the docker-compose up --build command:

$ docker-compose up --build -d

Then when we want to see the logs, we can use the docker-compose logs command:

$ docker-compose logs

Or we can tail the logs using this command:

$ docker-compose logs -f

Stopping the project

You can hit ctrl-C in the terminal window that ran docker-compose up.

Once that shutdown sequence is complete, it is important to run the docker-compose down command.

$ <ctrl>C
$ docker-compose down

Running docker-compose down will properly stop and terminate the Docker Compose session.

Note: None of the containers are setup by default with volumes and all data will be lost with docker-compose down. This is specifically important to remember regarding the postgres container. If you would like data to be persisted across builds locally, update docker-compose.yaml to define a volume.

Re-starting a specific Go service for development

When writing code in an iterative fashion, it is nice to have your change automatically rebuilt. This project uses github.com/gravityblast/fresh to recompile your services that will include most changes.

Fresh is a command line tool that builds and (re)starts your web application everytime you save a Go or template file. 

The (Fresh configuration file](https://gitlab.com/geeks-accelerator/oss/saas-starter-kit/blob/master/fresh-auto-reload.conf) is located in the project root. By default the following folders are watched by Fresh:

• handlers
• static
• templates

Any changes to internal/* or additional project dependencies added to go.mod will require the service to be rebuilt.

docker-compose up  --build -d web-app

Forking your own copy

1. Checkout the project

2. Update references.

flist=`grep -r "geeks-accelerator/oss/saas-starter-kit" * | awk -F ':' '{print $1}' | sort | uniq`
for f in $flist; do echo $f; sed -i "" -e "s#geeks-accelerator/oss/saas-starter-kit#geeks-accelerator/oss/aurora-cam#g" $f; done


flist=`grep -r "saas-starter-kit" * | awk -F ':' '{print $1}' | sort | uniq`
for f in $flist; do echo $f; sed -i "" -e "s#saas-starter-kit#aurora-cam#g" $f; done


flist=`grep -r "example-project" * | awk -F ':' '{print $1}' | sort | uniq`
for f in $flist; do echo $f; sed -i "" -e "s#example-project#aurora-cam#g" $f; done

3. Create a new AWS Policy with the following details:

Name:   SaasStarterKitDevServices 
Description: Defines access for saas-starter-kit services. 
Policy Document: {
                     "Version": "2012-10-17",
                     "Statement": [
                         {
                             "Sid": "DefaultServiceAccess",
                             "Effect": "Allow",
                             "Action": [
                                 "s3:HeadBucket",
                                 "s3:ListObjects",
                                 "s3:PutObject",
                                 "s3:PutObjectAcl",
                                 "cloudfront:ListDistributions",
                                 "ec2:DescribeNetworkInterfaces",
                                 "ec2:DeleteNetworkInterface",
                                 "ecs:ListTasks",
                                 "ecs:DescribeServices",
                                 "ecs:DescribeTasks",
                                 "ec2:DescribeNetworkInterfaces",
                                 "route53:ListHostedZones",
                                 "route53:ListResourceRecordSets",
                                 "route53:ChangeResourceRecordSets",
                                 "ecs:UpdateService",
                                 "ses:SendEmail",
                                 "ses:ListIdentities",
                                 "ses:GetAccountSendingEnabled",
                                 "secretsmanager:ListSecretVersionIds",
                                 "secretsmanager:GetSecretValue",
                                 "secretsmanager:CreateSecret",
                                 "secretsmanager:UpdateSecret",
                                 "secretsmanager:RestoreSecret",
                                 "secretsmanager:DeleteSecret"
                             ],
                             "Resource": "*"
                         },
                         {
                             "Sid": "ServiceInvokeLambda",
                             "Effect": "Allow",
                             "Action": [
                                 "iam:GetRole",
                                 "lambda:InvokeFunction",
                                 "lambda:ListVersionsByFunction",
                                 "lambda:GetFunction",
                                 "lambda:InvokeAsync",
                                 "lambda:GetFunctionConfiguration",
                                 "iam:PassRole",
                                 "lambda:GetAlias",
                                 "lambda:GetPolicy"
                             ],
                             "Resource": [
                                 "arn:aws:iam:::role/*",
                                 "arn:aws:lambda:::function:*"
                             ]
                         },
                         {
                             "Sid": "datadoglambda",
                             "Effect": "Allow",
                             "Action": [
                                 "cloudwatch:Get*",
                                 "cloudwatch:List*",
                                 "ec2:Describe*",
                                 "support:*",
                                 "tag:GetResources",
                                 "tag:GetTagKeys",
                                 "tag:GetTagValues"
                             ],
                             "Resource": "*"
                         }
                     ]
                 }

Create a new user with programmatic access and directly attach it the policy SaasStarterKitDevServices

4. Create a new docker-compose config file

 cp sample.env_docker_compose .env_docker_compose 

5. Update .env_docker_compose with the Access key ID and Secret access key

6. Update .gitlab-ci.yml with relevant details.

Optional. Set AWS and Datadog Configs

By default the project will compile and run without AWS configs or other third-party dependencies.

As you start utilizing AWS services in this project and/or ready for deployment, you will need to start specifying AWS configs in a docker-compose file. You can also set credentials for other dependencies in the new docker-compose file too.

The sample docker-compose file is not loaded since it is named sample, which allows the project to run without these configs.

To set AWS configs and credentials for other third-party dependencies, you need to create a copy of the sample environment docker-compose file without "sample" prepending the file name.

Navigate to the root of the project. Copy sample.env_docker_compose to .env_docker_compose.

$ cd $GOPATH/src/geeks-accelerator/oss/saas-starter-kit
$ cp sample.env_docker_compose .env_docker_compose

The example the docker-compose file specifies these environmental variables. The $ means that the variable is commented out.

$ AWS_ACCESS_KEY_ID=
$ AWS_SECRET_ACCESS_KEY=
AWS_REGION=us-east-1
$ AWS_USE_ROLE=false
$ DD_API_KEY=

In your new copy of the example docker-compose file ".env_docker_compose", set the AWS configs by updating the following environment variables: AWS_ACCESS_KEY_ID, AWS_SECRET_ACCESS_KEY, and AWS_REGION. Remember to remove the $ before the variable name.

As noted in the Local Installation section, the project is integrated with Datadog for observability. You can specify the API key for your Datadog account by setting the environment variable: DD_API_KEY.

Web API

cmd/web-api

REST API is available to clients for supporting deeper integrations. This API is also a foundation for third-party integrations. The API implements JWT authentication that renders results as JSON to clients.

Once the web-api service is running it will be available on port 3001. http://127.0.0.1:3001/

This web-api service is not directly used by the web-app service to prevent locking the functionally required for internally development of the web-app service to the same functionality exposed to clients via this web-api service. This separate web-api service can be exposed to clients and be maintained in a more rigid/structured process to manage client expectations.

The web-app will have its own internal API, similar to this external web-api service, but not exposed for third-party integrations. It is believed that in the beginning, having to define an additional API for internal purposes is worth for the additional effort as the internal API can handle more flexible updates.

For more details on this service, read web-api readme

API Documentation

Documentation for this API service is automatically generated using swag. Once this web-api service is running, it can be accessed at /docs

http://127.0.0.1:3001/docs/

Web App

cmd/web-app

Responsive web application that renders HTML using the html/template package from the standard library to enable direct interaction with clients and their users. It allows clients to sign up new accounts and provides user authentication with HTTP sessions. The web app relies on the Golang business logic packages developed to provide an API for internal requests.

Once the web-app service is running it will be available on port 3000. http://127.0.0.1:3000/

While the web-api service is rocking, this web-app service is still in development. Only the signup functionality works in order for a user to create the initial user with role of admin and a corresponding account for their organization. If you would like to help, please email twins@geeksinthewoods.com.

For more details on this service, read web-app readme

Schema

cmd/schema

Schema is a minimalistic database migration helper that can manually be invoked via CLI. It provides schema versioning and migration rollback.

The schema for the entire project is defined globally and is located inside internal: internal/schema

Keeping a global schema helps ensure business logic can be decoupled across multiple packages. It is a firm belief that data models should not be part of feature functionality. Globally defined structs are dangerous as they create large code dependencies. Structs for the same database table can be defined by package to help mitigate large code dependencies.

The example schema package provides two separate methods for handling schema migration:

• Migrations - List of direct SQL statements for each migration with defined version ID. A database table is created to persist executed migrations. Upon run of each schema migration run, the migraction logic checks the migration database table to check if it’s already been executed. Thus, schema migrations are only ever executed once. Migrations are defined as a function to enable complex migrations so results from query manipulated before being piped to the next query.
• Init Schema - If you have a lot of migrations, it can be a pain to run all them. For example, when you are deploying a new instance of the app into a clean database. To prevent this, use the initSchema function that will run as-if no migration was run before (in a new clean database).

Another bonus with the globally defined schema is that it enables the testing package to spin up database containers on-demand and automatically include all the migrations. This allows the testing package to programmatically execute schema migrations before running any unit tests.

Accessing Postgres

To login to the local Postgres container, use the following command:

docker exec -it saas-starter-kit_postgres_1 /bin/bash
bash-5.0# psql -U postgres shared

The example project currently only includes a few tables. As more functionality is built into both the web-app and web-api services, the number of tables will expand. You can use the show tables command to list them.

shared=# \dt
             List of relations
 Schema |      Name      | Type  |  Owner   
--------+----------------+-------+----------
 public | accounts       | table | postgres
 public | migrations     | table | postgres
 public | projects       | table | postgres
 public | users          | table | postgres
 public | users_accounts | table | postgres
(5 rows)

Deployment

This project includes a complete build pipeline that relies on AWS and GitLab. The presentation "SaaS Starter Kit - Setup GitLab CI / CD" has been made available on Google Docs that provides a step by step guide to setting up a build pipeline using your own AWS and GitLab accounts.

The .gitlab-ci.yaml file includes the following build stages:

stages:
  - build:dev     # Build containers with configs targeting dev env.
  - migrate:dev   # Run database migration against the dev database.
  - deploy:dev    # Deploy the containers built for dev env to AWS ECS. 
  - build:stage   # Build containers with configs targeting stage env.
  - migrate:stage # Run database migration against the stage database.
  - deploy:stage  # Deploy the containers built for stage env to AWS ECS. 
  - build:prod    # Build containers with configs targeting prod env.
  - migrate:prod  # Run database migration against the prod database.
  - deploy:prod   # Deploy the containers built for prod env to AWS ECS. 

Currently .gitlab-ci.yaml only defines jobs for the first three stages. The remaining stages can be chained together so each job is dependant on the previous or run jobs for each target environment independently.

A build tool called devops has been included apart of this project. Devops handles creating AWS resources and deploying your services with minimal additional configuration. You can customizing any of the configuration in the code. While AWS is already a core part of the saas-starter-kit, keeping the deployment in GoLang limits the scope of additional technologies required to get your project successfully up and running. If you understand Golang, then you will be a master at devops with this tool.

Refer to the README for setup details.

Development Notes

Country / Region / Postal Code Support

This project uses geonames.org to populate database tables for countries, postal codes and timezones that help facilitate standardizing user input. To keep the schema script quick for dev, the postal codes for only country code US are loaded. This can be changed as needed in geonames.go.

Datadog

Datadog has a custom init script to support setting multiple expvar urls for monitoring. The docker-compose file then can set a single env variable.

DD_EXPVAR=service_name=web-app env=dev url=http://web-app:4000/debug/vars|service_name=web-api env=dev url=http://web-api:4001/debug/vars

Postgres and future MySQL support

Postgres is only supported based on its dependency of sqlxmigrate. MySQL should be easy to add to sqlxmigrate after determining a better method for abstracting the create table and other SQL statements from the main testing logic.

SQLx bindvars

When making new packages that use sqlx, bind vars for mysql are ? where as postgres is $1.

To database agnostic, sqlx supports using ? for all queries and exposes the method Rebind to remap the placeholders to the correct database.

sqlQueryStr = db.Rebind(sqlQueryStr)

For additional details refer to bindvars

What's Next

We are in the process of writing more documentation about this code. We welcome you to make enhancements to this documentation or just send us your feedback and suggestions ; )

Join us on Gopher Slack

If you are having problems installing, troubles getting the project running or would like to contribute, join the channel #saas-starter-kit on Gopher Slack