Knex Dynamic Connection

This module is meant to patch knex and add support for defining dynamic connection configuration.

Table of contents

• Why you need it?
• Is it reliable?
• How does it actually work?
• Will this work in the future?

Why you need it?

Knex.js doesn't have inbuilt support for read/write connections. One can create two seperate instances of knex for read and write, but that still doesn't completely solves the problem. For example:

const Knex = require('knex')

const writeConfig = {
  client: 'pg',
  connection: {
  }
}
const writer = Knex(writeConfig)

const readConfig = {
  client: 'pg',
  connection: {
  }
}
const reader = Knex(readConfig)

Now, if you want to use multiple servers for read operations, you cannot do that, since knex.js allows only one connection server and will pool connections within that server.

Following is not possible

const readConfig = {
  client: 'pg',
  connection: [
    {
      host: ''
    },
    {
      host: ''
    }
  ]
}

With the help of this module, you can make knex create a connection using the dynamic config for every query.

npm i knex-dynamic-connection

const Knex = require('knex')
const { patchKnex } = require('knex-dynamic-connection')

const readConfig = {
  client: 'pg',
  connection: {},
  replicas: [
    {
      host: '',
    },
    {
      host: '',
    }
  ],
}

const knex = Knex(readConfig)
let roundRobinCounter = 0

patchKnex(knex, (originalConfig) => {
  const node = roundRobinCounter++ % originalConfig.replicas.length
  return originalConfig.replicas[node]
})

The patchKnex method overwrites the acquireRawConnection on all the dialects and make them fetch the config from your callback vs reading it from a static source.

Is it reliable?

Yes!

1. I have copied the code of acquireRawConnection from the knex codebase and have just made one line of change to read the config from a different source.
2. I have written tests for select, insert, transactions and schema methods.
3. The code is tested against mssql, mysql, mysql2 and pg.
4. The connection is still managed inside the pool, so don't worry about any extra maintaince overhead.

How does it actually work?

Knex.js makes use of tarn.js for managing pool resources and everytime pool needs a connection, knexjs calls acquireRawConnection on the dialect in use.

The dialect creates a new connection to the database server, but uses static configuration. I have just added a patch, which will rely on your closure to return the config vs using the same static config all the time.

Will this work in the future?

I am using Knex.js to write the ORM of https://adonisjs.com/ and will keep a close eye on the releases of Knex to make sure that I incorporate any changes made of the underlying code and keep this module upto date. If knex.js team plans to re-write the entire codebase (which is less likely to happen), then I will pitch this change to be a first class citizen.

Can I help?

Yes, there are currently no tests for oracle. It will be great, if you can help set it up using docker