hassy

//
// Custom Artillery functions to print the latency distribution for a single HTTP request
//
// Example usage: (in your scenario)
// - get:
//     url: "/some/endpoint"
//     beforeRequest: ["recordStartTime"]
//     afterResponse: ["logLatencyDelta"]
//     name: "Optional name for the endpoint for the report"

hassy

hassy

Tracing your own app

Set up lltrace-aws-sdk

Wait for enough activity to be traced

(a quick load-test is good if you're impatient)

Run lltrace

hassy

What can be traced

• S3 putObject and uploadFile
• Invoking other lambdas
• Publishing to SNS topics
• Custom actions (e.g. write to Neo4J)

Sampling

By default only 10% of traced actions actually get logged.

hassy

How it works

SNS and S3 notifications are easy

Lambda activity - not quite so

Solution: monkey-patch aws-sdk 🙈

hassy

What we wanted

To recap:

• 20 lambdas
• 10 SNS topics (more than one subscriber/publisher for some)
• S3 buckets (triggering lambdas)

hassy

AWS API Gateway

GraphQL

Lambda

hassy

config:
  target: "http://localhost:3003"
  phases:
    - duration: 1
      arrivalRate: 1
scenarios:
  - flow:
      - get:
          url: "/"
      - get:

hassy

- post:
    url: "/some/url"
    formData:
      firstName: "Jerry"
      lastName: "Winston"
    headers:
      authorization: "Bearer {{access_token}}"

hassy

//
// Lambda's timeout needs to be >5 seconds, 10 should do
//

var startedAt = new Date();

var interval = setInterval(function () {
    console.log(startedAt, new Date());
}, 1000);