dynamodb-as-eventstore

Using DynamoDB as EventStore with SQS and Lambda.

Article at https://learnings.aleixmorgadas.dev/p/dynamodb-as-eventstore

Setup local environment

I assume you have:

• Terraform installed
• AWS Account
• Python Environment

Install Python Dependencies

ℹ️ I used Linux/Debian for this project, you might need to adapt the commands to your operating system.

python3 -m venv venv
source venv/bin/activate
pip install -r requirements.txt

You might need python3-venv as OS dependency installed.

Deployment

⚠️ The provided solution will incur in your billing as some services are not part of the AWS Free Tier

Before Deployment, you should check:

• infrastructure/main.tf. The region
• infrastructure/variables.tf. Set the read and write capacity of DynamoDB that you need. Default read-capacity 1, write-capacity 1. Also, prevent destroy for the event-store is set to false as default.

At infrastructure folder, execute ./run.sh.

Destroy Infrastructure ⛑️

At infrastructure folder, execute ./destroy.sh.

Solution Design. Trade-offs

• Queue isn't a FIFO (First In-First Out).

It has consequences in the order the events are stored. Even though they contain a timestamp, they might get stored in a different order they have been produced.

In case of low traffic it might not be a real issue. But in case the application starts emitting events intensively that require them to be ordered, you might need to switch from a normal Queue to a FIFO Queue or Kinesis.

• Hash Key and Sort Key

Hash Key = ID Sort Key = Timestamp

The idea is to have the Events Ordered by Timestamp.

• Items don't expire

I haven't configured the Time to Live (TTL) in DynamoDB. The table will grow in infinitum.w

• Only one Index
  You might want to add more indexes to query the events as you like.

  • Info in Terraform to add them in DynamoDB table configuration.

  • Info in AWS About Global Secondary Indexes and why you should define them to perform queries instead of scans.

Event Base

All events received by the lambda must follow the next schema:

Field	Type	Required	Description
Timestamp	Date	true	Time when event was created
EventType	String	true	Event identifier. Format EVENT_EXAMPLE
ID	String	true	Id of the entity
EmitterId	String	true	Id of the user or system that caused the event
EmitterType	String	true	SYSTEM or USER is what we use. It's a free text
Data	Object	false	fields of the event

In case of not following the schema, the event will be rejected.

Example:

{
	"Timestamp"	: 1596719980,
	"EventType"	: "EVENT_TEST",
	"ID"		: "53877b2e-e496-11ea-b678-9b42a44d5e59",
	"EmitterId"	: "6b04fd08-e496-11ea-8176-9f85e075569c",
	"EmitterType": "USER",
	"Data"		: {
		"foo": "bar"
	}
}

Technical Debt

• AWS Permissions

The different AWS Roles have too much permissions in their policies. It will be convenient that you restricted those to what's really required.

• AWS CloudWatch

It will be convenient to setup a CloudWatch Dashboard with the main metrics to check regularly to know how the system is behaving.