Huge News!Announcing our $40M Series B led by Abstract Ventures.Learn More
Socket
Sign inDemoInstall
Socket

kinesis-client-library

Package Overview
Dependencies
Maintainers
1
Versions
17
Alerts
File Explorer

Advanced tools

Socket logo

Install Socket

Detect and block malicious and high-risk dependencies

Install

kinesis-client-library

Process Kinesis streams and automatically scale up or down as shards split or merge.

  • 0.3.6
  • latest
  • Source
  • npm
  • Socket score

Version published
Weekly downloads
2
decreased by-50%
Maintainers
1
Weekly downloads
 
Created
Source

Node Kinesis Client Library

Build Status

Based on the AWS Kinesis Client Library for Java, reimplemented in Node.js.

  • Build Kinesis consumers easily
  • Automatically scale up or down as streams split or merge
  • Allow distributed processing without any extra code

Install with npm install kinesis-client-library --save.

I am using this in production, but it's very new so there's a good chance it will (1) have some bugs and (2) go through a breaking change at some point. Feedback welcome, use with care.

Why?

I had a stack in Node.js. Working with and deploying a Java app just to consume a stream was too inconvenient, so I made this.

Terminology

  • Consumer: A single process that is responsible for consuming a single Kinesis shard. Consumers are created and managed by clusters; they should not be created directly directly.
  • Cluster: A process that manages consumers. A single cluster can have many consumers (or 1, or 0) that all run on the same machine. Clusters should be created by the CLI.
  • Network: All of the consumers processing a single Kinesis stream for a single application. Networks can have many clusters that are run on one or many machines.

Networks are meant to be distributed (though they don't have to be) and will automatically rebalance shards across clusters. Each cluster will try to process an equal number of shards, so it's a good idea to make sure each cluster has equivalent resources (e.g. clusters should have similar memory and CPU available). Networks will automatically pick up new shards as splits or merges happen.

Multiple networks can independently process the same stream as long as the stream has enough capacity. In this case, you should be sure to give each set of processors a unique table flag.

Usage

Before you can do anything with this library, you need to create a Kinesis stream. Everything from this point forward assumes you have an existing stream.

CLI

Clusters are designed to be launched from the command line. The library exposes a launch-kinesis-cluster executable.

$ launch-kinesis-cluster --help
Usage:

--help  (Display this message)

Required flags:
--consumer [Path to consumer file]
--table [DynamoDB table name]
--stream [Kinesis stream name]

Optional flags:
--start-at [Starting iterator type] ("trim_horizon" or "latest", defaults to "trim_horizon")
--capacity.[read|write] [Throughput] (DynamoDB throughput for *new* tables, defaults to 10 for each)
--aws.[option] [Option value]  (e.g. --aws.region us-west-2)
--http [port]  (Start HTTP server, port defaults to $PORT)
--log-level [level] (Logging verbosity, uses Bunyan log levels)
--dynamo-endpoint (Use a cusotm endpoint for the DynamoDB service)
--local-dynamo (Whether or not to use a local implementation of DynamoDB, defaults to false)
--local-dynamo-directory (Directory to store local DB, defaults to temp directory)
--kinesis-endpoint (Use a custom endpoint for the Kinesis service)
--local-kinesis (Use a local implementation of Kinesis, defaults to false)
--local-kinesis-port (Port to access local Kinesis on, defaults to 4567)
--local-kinesis-no-start (Assume a local Kinesis server is already running, defaults to false)
--num-records (Maximum number of records to get in each Kinesis query, defaults to the Kinesis maximum of 10000)

Notes:

  • table is used as a DynamoDB table name that is unique to the network. If the table doesn't exist yet, it is created.
  • stream must match a Kinesis stream that already exists.
  • The (optional) HTTP server is primarily meant to be used as a health check.
  • AWS options are based to the AWS.Service constructor, so any other credential strategies (e.g. environment variables, EC2 IAM roles) will be used automatically.

Consumer

Consumers are implemented in JavaScript by calling AbstractConsumer.extend() with an object that implements some/all of these methods:

  • processRecords: Accepts an array of Record objects (Record.Data will be a Buffer object) and a callback. Pass true as the second argument to the callback when you want to save a checkpoint — i.e. when you have processed a chunk of data.
  • processResponse: Accepts a full response from Kinesis' GetRecords API and a callback. Pass true as the second argument to the callback when you want to save a checkpoint — i.e. when you have processed a chunk of data.
  • initialize: Called when a consumer is spawned, before any records are processed. Accepts a callback that must be called to start record processing.
  • shutdown: Called when a consumer is about to exit. Accepts a callback that must be called to complete shutdown; if the callback is not called without 30 seconds the process exits anyway.

A consumer MUST implement either processRecords or processResponse.

Example consumer implementation

This consumer uploads records to S3 in 50 megabyte batches.

// AWS config skipped for brevity
var s3 = new require('aws-sdk').S3()
var kcl = require('kinesis-client-library')

var newlineBuffer = new Buffer('\n')

kcl.AbstractConsumer.extend({
  // create places to hold some data about the consumer
  initialize: function (done) {
    this.cachedRecords = []
    this.cachedRecordsSize = 0
    // This MUST be called or processing will never start
    // That is really really really bad
    done()
  },

  processRecords: function (records, done) {
    // Put each record into our list of cached records (separated by newlines) and update the size
    records.forEach(function (record) {
      this.cachedRecords.push(record.Data)
      this.cachedRecords.push(newlineBuffer)
      this.cachedRecordsSize += (record.Data.length + newlineBuffer.length)
    }.bind(this))

    // not very good for performance
    var shouldCheckpoint = this.cachedRecordsSize > 50000000

    // Get more records, but not save a checkpoint
    if (! shouldCheckpoint) return done()

    // Upload the records to S3
    s3.putObject({
      Bucket: 'my-bucket-name',
      Key: 'path/to/records/' + Date.now(),
      Body: Buffer.concat(this.cachedRecords)
    }, function (err)  {
      if (err) return done(err)

      this.cachedRecords = []
      this.cachedRecordsSize = 0

      // Pass `true` to checkpoint the latest record we've received
      done(null, true)
    }.bind(this))
  }
})
Example consumer launch

Assume the code above is in lib/consumer.js and you've got a package with kinesis-client-library declared as a dependency.

For convenience, we'll assume your AWS credentials are defined as environment variables. Using the command line, you could launch a cluster like this.

launch-kinesis-cluster \
  --consumer lib/consumer.js \
  --table MyKinesisConsumerApp \
  --stream MyKinesisStreamName \
  --aws.region us-east-1

Keywords

FAQs

Package last updated on 12 Feb 2017

Did you know?

Socket

Socket for GitHub automatically highlights issues in each pull request and monitors the health of all your open source dependencies. Discover the contents of your packages and block harmful activity before you install or update your dependencies.

Install

Related posts

SocketSocket SOC 2 Logo

Product

  • Package Alerts
  • Integrations
  • Docs
  • Pricing
  • FAQ
  • Roadmap
  • Changelog

Packages

npm

Stay in touch

Get open source security insights delivered straight into your inbox.


  • Terms
  • Privacy
  • Security

Made with ⚡️ by Socket Inc