@castore/message-queue-adapter-sqs-s3
v2.3.1
Published
DRY Castore MessageQueue definition using AWS SQS and S3
Downloads
316
Readme
SQS + S3 Message Bus Adapter
DRY Castore MessageQueue
definition using AWS SQS and AWS S3.
This adapter works like the SQS Message Queue Adapter (it actually uses it under the hood), excepts that entry sizes are checked before publishing messages to EventBridge. If they are over the 256KB limit, they are written on a s3 bucket instead, and a message is sent containing a pre-signed URL, as recommended by AWS.
Do not forget to set a lifecycle configuration on your s3 bucket to delete the written objects after the presigned URL has expired to avoid high s3 bills! 🤑
📥 Installation
# npm
npm install @castore/message-queue-adapter-sqs-s3
# yarn
yarn add @castore/message-queue-adapter-sqs-s3
This package has @castore/core
, @aws-sdk/client-sqs
(above v3), @aws-sdk/client-s3
(above v3) and @aws-sdk/s3-request-presigner
(above v3) as peer dependencies, so you will have to install them as well:
# npm
npm install @castore/core @aws-sdk/client-sqs @aws-sdk/client-s3 @aws-sdk/s3-request-presigner
# yarn
yarn add @castore/core @aws-sdk/client-sqs @aws-sdk/client-s3 @aws-sdk/s3-request-presigner
👩💻 Usage
import { SQSClient } from '@aws-sdk/client-sqs';
import { S3Client } from '@aws-sdk/client-s3';
import { SQSS3MessageBusAdapter } from '@castore/message-queue-adapter-sqs-s3';
const sqsClient = new SQSClient({});
const s3Client = new S3Client({});
const messageQueueAdapter = new SQSS3MessageQueueAdapter({
queueUrl: 'https://sqs.us-east-1.amazonaws.com/111122223333/my-super-queue',
sqsClient,
s3BucketName: 'my-bucket-name',
s3Client,
// 👇 Optional s3 prefix for temporary data
s3Prefix: 'temporary-storage/',
// 👇 Optional s3 presignature expiration in seconds (defaults to 900)
s3PreSignatureExpirationInSec: 3600
});
// 👇 Alternatively, provide a getter
const messageQueueAdapter = new SQSS3MessageQueueAdapter({
queueUrl: () => process.env.MY_QUEUE_URL,
s3BucketName: () => process.env.MY_BUCKET_NAME
...
});
const appMessageQueue = new NotificationMessageQueue({
...
messageQueueAdapter
})
This will directly plug your MessageQueue to SQS and S3 🙌
🤔 How it works
You can read the SQS Message Queue Adapter documentation for regular cases.
When an entry is oversized, its Detail
is saved as a JSON object in the provided s3 bucket. It's key is a concatenation of the constructor s3Prefix
option, the eventStoreId
and aggregateId
of the event and the current timestamp:
const key = 'temporary-storage/POKEMONS/pikachu1/2020-01-01T00:00:00.000Z';
If the event is a notification or state-carrying event, the version
is also added to the mix:
// 👇 Date is suffixed by the version
const key = 'temporary-storage/POKEMONS/pikachu1/2020-01-01T00:00:00.000Z#3';
On the listeners side, you can use the SQSS3MessageBusMessage
TS type to type your argument, and the parseMessage
util to fetch the message if it has been uploaded to S3 (it passes it through otherwise):
On the worker side, you can use the SQSS3MessageQueueMessage
TS type to type your argument, and the parseBody
util to fetch the message if it has been uploaded to S3 (it passes it through otherwise):
import {
SQSS3MessageQueueMessage,
parseBody,
} from '@castore/message-queue-adapter-sqs-s3';
const appMessagesWorker = async ({ Records }: SQSS3MessageQueueMessage) => {
for (const { body } of Records) {
// 🙌 Correctly typed!
const recordBody = await parseBody<typeof appMessageQueue>(body);
}
};
Note that parseBody
uses fetch
under the hood, so you will have to provide it if your version of node doesn't:
import fetch from 'node-fetch';
import {
SQSS3MessageQueueMessage,
parseBody,
} from '@castore/message-queue-adapter-sqs-s3';
const appMessagesWorker = async ({ Records }: SQSS3MessageQueueMessage) => {
for (const { body } of Records) {
// 🙌 Correctly typed!
const recordBody = await parseBody<typeof appMessageQueue>(body, { fetch });
}
};
🔑 IAM
The publishMessage
method requires the sqs:SendMessage
IAM permission on the provided SQS queue, as well as the s3:putObject
and s3:getObject
IAM permissions on the provided s3 bucket at the desired keys (e.g. my-bucket-name/temporary-storage/*
).
The parseBody
util doesn't require any permission as the messageURL is pre-signed.