Skip to content

Latest commit

 

History

History
92 lines (57 loc) · 3.3 KB

README.md

File metadata and controls

92 lines (57 loc) · 3.3 KB
Raw

Serializable Isolation Level explained

Cover

A small project showing serializable transactions usage and capability, made as part of the article It's not always about locking data. The goal of this project is to have a self-contained test showing a possible use-case of SERIALIZABLE Isolation Level.

📜 Overview

This small app emulates a bank account withdrawal process.

API Endpoint: Withdrawals

This endpoint allows users to initiate a withdrawal request. If you pass serializable parameter as false (or don't pass it at all), the withdrawal request will be handled using READ COMMITED isolation level, which will fail validating the daily withdrawal limit of $5000 in case of two or more concurrent requests for same user. Otherwise, if you pass serializable as true, the request will be handle using SERIALIZABLE isolation level, protecting the database from an inconsistent state.

Be aware that this endpoint has a sleep of 5 seconds before committing changes on database. This is needed to reproduce the scenario without having a high load environment. You can check it at WithdrawalService.

HTTP Method

POST

Endpoint URL

/withdrawals

Query parameters

serializable={false|true}

Request Headers

  • Content-Type: application/json

Request Body

The request body should be a JSON object with the following fields:

{
  "amount": 1000,
  "user": 1
}

💻 Prerequisites

Before you begin, make sure you have met the following requirements:

  • NodeJS v18 or above
  • Docker v27 or above

☕ Preparing

Clone this repo and issue a npm install command at the root directory to install its dependencies.

🚀 Running

To run this project, you have two options: either start the application and send concurrent requests manually (in that case you also need a PostgreSQL instance running locally), or run the integration tests, which include this scenario and starts a PostgreSQL container automatically.

Running manually

  • Create a .env file and add your PostgreSQL URI there, such as DATABASE_URL=postgres://myuser:mypass@localhost:5432/mydatabase.
  • Issue a npx sequelize-cli db:migrate to run all database migrations (create tables, etc.)
  • Now you're good to go, just issue a npm start to start the app
  • Then you can send requests, e.g.:
curl --request POST \
  --url 'http://localhost:3000/withdrawals?serializable=false' \
  --header 'Content-Type: application/json' \
  --data '{
	"amount": 1000,
	"user": 1
}'

Remember to run at least 2 requests simultaneously for same user and to play with serializable query parameter.

Running integration tests

  • Issue a npm test and inspect test results

We're covering these scenarios:

  1. Two concurrent requests for same user to be handled by READ COMMITED isolation level. This test is intentionally failing due to inability of applying the limit.
  2. Two concurrent requests for same user to be handled by SERIALIZABLE isolation level. This test is running fine.

📝 License

This project is licensed under the GNU General Public License v3.0. See the LICENSE file for more details.