AWS for DevOps — DynamoDB for DevOps

9.3 — DynamoDB for DevOps

Amazon DynamoDB is a fully managed NoSQL database with single-digit millisecond performance at any scale. For DevOps, it’s the go-to database for serverless backends, Terraform state locking, and event-driven pipelines.

Key Concepts

Concept	Description	Analogy
Table	Collection of items	A spreadsheet
Item	A single record (row)	A row in the spreadsheet
Attribute	A field on an item	A column value
Partition Key (PK)	Primary identifier — must be unique	Employee ID
Sort Key (SK)	Optional — enables range queries	Date, version number

Creating a Table

$ aws dynamodb create-table \
    --table-name Deployments \
    --attribute-definitions \
        AttributeName=ServiceName,AttributeType=S \
        AttributeName=DeploymentId,AttributeType=S \
    --key-schema \
        AttributeName=ServiceName,KeyType=HASH \
        AttributeName=DeploymentId,KeyType=RANGE \
    --billing-mode PAY_PER_REQUEST \
    --tags Key=Team,Value=DevOps
{
    "TableDescription": {
        "TableName": "Deployments",
        "TableStatus": "CREATING",
        "BillingModeSummary": { "BillingMode": "PAY_PER_REQUEST" }
    }
}

On-Demand vs Provisioned

Mode	Pricing	Scaling	Best For
On-Demand	Per request ($1.25/million writes)	Instant, automatic	Unpredictable traffic, new workloads
Provisioned	Per RCU/WCU per hour	Manual or auto-scaling	Predictable traffic, cost optimisation

CRUD Operations

# Put an item
$ aws dynamodb put-item \
    --table-name Deployments \
    --item '{
        "ServiceName": {"S": "api-service"},
        "DeploymentId": {"S": "deploy-2026-001"},
        "Status": {"S": "SUCCESS"},
        "Version": {"S": "v2.1.0"},
        "Timestamp": {"N": "1711800000"}
    }'

# Get an item
$ aws dynamodb get-item \
    --table-name Deployments \
    --key '{"ServiceName": {"S": "api-service"}, "DeploymentId": {"S": "deploy-2026-001"}}' \
    --query 'Item.{Service:ServiceName.S,Status:Status.S,Version:Version.S}'
{
    "Service": "api-service",
    "Status": "SUCCESS",
    "Version": "v2.1.0"
}

# Query all deployments for a service
$ aws dynamodb query \
    --table-name Deployments \
    --key-condition-expression "ServiceName = :svc" \
    --expression-attribute-values '{":svc": {"S": "api-service"}}' \
    --scan-index-forward false \
    --limit 5

DevOps Use Cases

Use Case	Table Design
Terraform state locking	PK=LockID — prevents concurrent applies
Deployment tracking	PK=ServiceName, SK=DeploymentId
Feature flags	PK=FlagName, attributes=enabled, percentage
Configuration store	PK=Environment, SK=ConfigKey

TTL (Time to Live)

Automatically delete old items:

$ aws dynamodb update-time-to-live \
    --table-name Deployments \
    --time-to-live-specification Enabled=true,AttributeName=ExpiresAt

Set ExpiresAt to a Unix timestamp — DynamoDB deletes the item after that time (within 48 hours).

Common Mistakes

Mistake	Why It Hurts	Fix
Using Scan instead of Query	Reads entire table — slow and expensive	Design keys for Query access patterns
Provisioned mode for new workloads	Under/over-provisioned, throttling or waste	Start with On-Demand, switch later
No TTL on temporary data	Table grows forever	Enable TTL for logs, sessions, locks
Hot partition (one PK gets all traffic)	Throttling on that partition	Distribute writes across partition keys

Quick Reference

Task	Command
Create table	aws dynamodb create-table --table-name NAME --billing-mode PAY_PER_REQUEST ...
Put item	aws dynamodb put-item --table-name NAME --item '{...}'
Get item	aws dynamodb get-item --table-name NAME --key '{...}'
Query	aws dynamodb query --table-name NAME --key-condition-expression "PK = :val"
Delete table	aws dynamodb delete-table --table-name NAME

---

Next → SAM — Serverless Application Model