Table of Contents generated with DocToc

• rmq-eddy-current
  • Install and setup
    • Example
  • Scenario
  • When to use a rmq-eddy-current
  • When to not use and eddy

rmq-eddy-current

A small TS node package for offsetting work to RMQ with the intention of consuming the work by the emitter (or 1 of the emitters in the cluster)

Install and setup

Ensure you install the peer dependencies then follow the example below.

Example

If you have many connections you might want to connect them all in a single static class for ease of access throughout your app. You may also want to inject a function from elsewhere as the consumeCb function. As the emit and parse is from the same service, you now have a typed payload to play with.

// import it
import { Eddy } from 'rmq-eddy-current/build/Eddy';

// set the settings for the connection, verbose here for demo
const baseConfig =  {
  protocol: 'amqp',
  hostname: 'mrrabbit.domain.com',
  port: 5672,
  username: 'guest',
  password: 'guest',
  verboseLogging: false,
  queue: `q.somequeue`,
  dleQueue: 'q.dle_queue',
  dleExchange: 'myapp.dleExchange',
  exchange: 'myapp.serviceNameExchange',
  exchangeType: 'direct'
}

// declare the interface for the data shape to be sent via the q
export interface IsomeQueue{
  id: string,
  campaignId: string
}

// initialise it
export const someQueue = new Eddy<IsomeQueue>({
    ...baseConfig,
    consumeCb: async (obj: IsomeQueue): Promise<ConsumerResponse> => {
      // Example thing to be called
      await BuiltwithRepository.patchRecordAsUsed(obj.id, obj.campaignId)
      
      // and anything else...
      
      // then return
      return { processed: true, requeue: false }
});

// Connect the eddy to your rmq 
await someQueue.connect();

// Publish to it
someQueue.publish({
  id: '321354654lop',
  campaignId: '00thgdfhsfgh'
})

Scenario

1. You have *n services in 1 cluster, ie they are clones of each other
2. A job comes in to process 20k records
3. 1 of the services in your cluster throws each record to the Q, 1 by 1, in seconds
4. All services in your cluster consume the q (as they are all connected to it)

When to use a rmq-eddy-current

Typically, you would employ dedicated workers for this kind of work. However, there are complexities with workers, firstly they must be co-ordinated and maintained.

When you have highly infrequent workloads (typically backoffice kind of stuff) 10k records to parse here and there, it is far more cost-effective to let the service which enqueues the work to also process the work. Assuming performance of that service is not mission critical. Think of it like an eddy current in a stream, the main body of water for the service is flowing but little eddy's quickly spin up do their work then spin down.

When to not use and eddy

When doing the work would impact the performance of the app to an audience which should not be affected... for example, your customers. Or when the work load is fair consistent, like every 2nd day you need to chunk through a lot of data.

This was not designed for sending data between 2 different services.