nats-load-traffic

Load test NATS with configurable traffic patterns.

Description

This application enables various load profiles to be constructed and applied to a NATS server or cluster. Setup multiple controlled and concurrent submission loads to test NATS performance or downstream consumer/workers performance.

Additionally, this app collects metrics about submissions and enables them to be forwarded to a prometheus remote write server for further analysis.

Getting Started

Examples

Please see /examples for a demo

Running

The app can be run from a pre-built docker container

version: "3.7"
services:
  nats-load-traffic:
    container_name: nats-load-traffic
    image: evanofslack/nats-load-traffic:latest
    restart: unless-stopped
    volumes:
      - ./config:/config
    command: -c /config/config.yaml -m /config/msg.json

Alternatively, build the executable from source

git clone https://github.com/evanofslack/nats-load-traffic
cd nats-load-traffic/cmd/nats-load-traffic
go build ./...

Parameters

Runtime parameters can be configured through flags:

usage: nats-load-traffic [-c config path] [-p payload path]
  -c string
        path to config file (default "config.yaml")
  -p string
        path to message payload file (default "msg.json")

Configuration

Configuration file is a yaml formatted file. It provides the endpoints for the nats server and the remote write prometheus server (if enabled).

The app can take any number of profiles. It will run concurrent load for each provided profile.

remote_write:
  enabled: true
  url: http://localhost:9090/api/v1/write
  interval: 5s

nats:
  url: nats://localhost:4222

profiles:
  - name: job-1     # human readable profile name
    subject: test.1 # nats subject to submit to
    load: constant  # type of load (constant, periodic, random, rare)
    rate: 600       # max submit rate (msg/sec)
    duration: 60m   # how long to run the load

  - name: job-2
    subject: test.2
    load: random
    rate: 600
    duration: 60m

Metrics

If remote write is enabled, all publish metrics will be sent to the prometheus remote write target. There is one metric with a couple of labels that is recorded: nats_load_traffic_submissions_total

Here is an example plot of the metric (which also exhibits the four different load profiles): rate(nats_load_traffic_submissions_total{result="success"}[30s])