Skip to content


Switch branches/tags

Name already in use

A tag already exists with the provided branch name. Many Git commands accept both tag and branch names, so creating this branch may cause unexpected behavior. Are you sure you want to create this branch?


Failed to load latest commit information.
Latest commit message
Commit time


semantic-release npm version GitHub
"Buy Me A Coffee"

Serverless-postgres is a wrapper for node-pg Node.js module. It is heavily inspired by Jeremy Daly's serverless-mysql package.

Why I need this module?

In a serverless application a function can scale almost "infinitely" by creating separate container instances for each concurrent user. Each container can correspond to a database connection which, for performance purposes, is left opened for further re-utilization. If we have a sudden spike of concurrent traffic, the available connections can be quickly maxed out by other competing functions. If we reach the max connections limit, Postgres will automatically reject any frontend trying to connect to its backend. This can cause heavy disruption in your application.

What does it do?

Serverless-postgres adds a connection management component specifically for FaaS based applications. By calling the method .clean() at the end of your functions, the module will constantly monitor the status of all the processes running in the PostgreSQL backend and then, based on the configuration provided, will garbage collect the "zombie" connections. If the client fails to connect with "sorry, too many clients already" error, the module will retry using trusted backoff algorithms.

NOTE: This module should work with any PostgreSQL server. It has been tested with AWS's RDS Postgres, Aurora Postgres, and Aurora Serverless.

Feel free to request additional features and contribute =)


  • Default connections filtering (>= v2): this feature leverage postgres application_name to differentiate clients created by this library and others, this will avoid terminating connections belonging to long-running process, batch jobs, ect... By default, we set the same application_name parameter for all the serverless clients, if you wish you can change it by just specifying it in the client config:
    const client = new ServerlessClient({
      application_name: 'my_client',


npm i serverless-postgres


Declare the ServerlessClient outside the lambda handler

const ServerlessClient = require('serverless-postgres')

const client = new ServerlessClient({
  user: process.env.DB_USER,
  host: process.env.DB_HOST,
  database: process.env.DB_NAME,
  password: process.env.DB_PASSWORD,
  port: process.env.DB_PORT,
  debug: true,
  delayMs: 3000,

const handler = async (event, context) => {
  await client.connect();
  const result = await client.query(`SELECT 1+1 AS result`);
  await client.clean();
  return {
    body: JSON.stringify({ message: result.rows[0] }),
    statusCode: 200

You can set the configuration dynamically if your secret is stored in a vault

const ServerlessClient = require('serverless-postgres')

const client = new ServerlessClient({
  host: process.env.DB_HOST,
  database: process.env.DB_NAME,
  port: process.env.DB_PORT,

const handler = async (event, context) => {
  const { user, password } = await getCredentials('my-secret')
    user, password
  await client.connect();
  // of the code

Configuration Options

Property Type Description Default
config Object A node-pg configuration object as defined here {}
maxConnsFreqMs Integer The number of milliseconds to cache lookups of max_connections. 60000
manualMaxConnections Boolean if this parameters is set to true it will query to get the maxConnections values, to maximize performance you should set the maxConnections yourself false
maxConnections Integer Max connections of your PostgreSQL. I highly suggest to set this yourself 100
strategy String Name of your chosen strategy for cleaning up "zombie" connections, allowed values minimum_idle_time or ranked minimum_idle_time
minConnectionIdleTimeSec Integer The minimum number of seconds that a connection must be idle before the module will recycle it. 0.5
maxIdleConnectionsToKill Integer or null The amount of max connection that will get killed. Default is ALL null
connUtilization Number The percentage of total connections to use when connecting to your PostgreSQL server. A value of 0.80 would use 80% of your total available connections. 0.8
debug Boolean Enable/disable debugging logs. false
capMs Integer Maximum number of milliseconds between connection retries. 1000
baseMs Integer Number of milliseconds added to random backoff values. 2
delayMs Integer Additional delay to add to the exponential backoff. 1000
maxRetries Integer Maximum number of times to retry a connection before throwing an error. 3
processCountCacheEnabled Boolean Enable caching for get process count. False
processCountFreqMs Integer The number of milliseconds to cache lookups of process count. 6000
allowCredentialsDiffing Boolean If you are using dynamic credentials, such as IAM, you can set this parameter to true and the client will be refreshed false
library Function Custom postgres library require('pg')


  • Serverless-postgres depends on pg package and usually you do not need to install it on your own. As some users have observed, if you have installed it on your own, and it is a different version, this package might misbehave.