Skip to content

AWS-Setup.md

Latest commit

 

History

History
129 lines (104 loc) · 4.75 KB

AWS-Setup.md

File metadata and controls

129 lines (104 loc) · 4.75 KB
Raw

AWS Configuration Guide (for RedLine13 Distributed Load)

RedLine13 launches load-generator EC2 instances in your own AWS account using IAM credentials you provide. This guide covers the minimum, least- privilege setup. You stated AWS is not configured yet — do these steps when you are ready to run distributed (cloud) tests. Local execution needs none of this.

1. AWS Account Setup

  1. Create/identify the AWS account that will host load generators (ideally a dedicated non-prod account so load-test EC2 cost and blast radius are isolated).
  2. Pick your primary region (e.g. us-east-1) and the additional regions you intend to load from (see multi-region table in RedLine13-Guide.md).
  3. Confirm EC2 vCPU service quotas in each region are high enough for your largest test. 50k VUs ≈ 25 × c5.2xlarge = 200 vCPUs — request a quota increase in advance (On-Demand Standard instances quota), as new accounts default low.

2. IAM Permissions (least privilege)

Create a dedicated IAM user redline13-loadgen with programmatic access (access key + secret). Attach a policy granting only what RedLine13 needs to launch/terminate generators:

{
  "Version": "2012-10-17",
  "Statement": [
    {
      "Sid": "RedLine13LoadGenerators",
      "Effect": "Allow",
      "Action": [
        "ec2:RunInstances",
        "ec2:TerminateInstances",
        "ec2:StartInstances",
        "ec2:StopInstances",
        "ec2:DescribeInstances",
        "ec2:DescribeInstanceStatus",
        "ec2:DescribeImages",
        "ec2:DescribeKeyPairs",
        "ec2:CreateKeyPair",
        "ec2:DescribeSecurityGroups",
        "ec2:CreateSecurityGroup",
        "ec2:AuthorizeSecurityGroupEgress",
        "ec2:AuthorizeSecurityGroupIngress",
        "ec2:DescribeVpcs",
        "ec2:DescribeSubnets",
        "ec2:DescribeRegions",
        "ec2:CreateTags"
      ],
      "Resource": "*"
    }
  ]
}

Tighten further in production by constraining Resource/conditions to specific regions, instance types, and a tag (e.g. Project=Perf-Lab). Rotate the access key regularly and store it only in the RedLine13 account settings.

Add the access key + secret under RedLine13 → Account → AWS Settings.

3. EC2 Requirements

  • AMI: RedLine13 supplies its own agent AMI per region — you don't build one.
  • Instance types: see scaling table (t3.medium → c5.2xlarge). Compute- optimised c5/c6i give the best threads-per-dollar for CPU-bound load gen.
  • Key pair: RedLine13 can create one; keep it if you want SSH access to a generator for debugging.
  • EBS: default volume is sufficient; load generators are stateless.
  • Lifecycle: generators are ephemeral — launched at test start, terminated at test end. Verify termination after each run to avoid cost leaks.
  • Spot (optional): for large, non-time-critical runs, Spot instances cut cost substantially; tolerate occasional generator loss.

4. Security Groups

Two directions matter:

A. Generator → SUT (egress / SUT ingress). The SUT must accept inbound traffic from the generators' public IPs on the app ports (443/80).

  • For public endpoints: usually already open.
  • For private/internal the target application: whitelist the generator Elastic IPs, or place generators in a VPC/subnet peered to the SUT, or use a VPN. Coordinate with the network team before the test window.

B. Controller → Generator (management). RedLine13 needs to reach each generator. Its launch flow creates/uses a security group allowing the agent's management port and SSH (22) from RedLine13. Allow this in the IAM policy (above) so the SG can be created automatically.

[RedLine13 Controller] --(mgmt/SSH 22)--> [Generator SG] --(443)--> [SUT SG]

Restrict generator inbound SSH to RedLine13/your office CIDR, not 0.0.0.0/0.

5. Cost Considerations

You pay AWS directly for generator EC2 time (RedLine13 charges its own subscription separately). Rough On-Demand estimates (us-east-1, Linux):

Instance ~vCPU ~Approx $/hr Use
t3.medium 2 ~$0.04 small Load/Smoke
c5.large 2 ~$0.085 ~1k VUs
c5.xlarge 4 ~$0.17 ~5–10k VUs
c5.2xlarge 8 ~$0.34 50k VUs tier

Example: 50k-VU run = 25 × c5.2xlarge × 1 hr ≈ ~$8.50 of EC2 for the hour (prices vary by region/time; verify current pricing). Cost controls:

  • Terminate generators promptly; confirm zero running instances post-run.
  • Tag everything Project=Perf-Lab and set an AWS Budget + alarm.
  • Prefer fewer, larger generators over many tiny ones for big runs.
  • Use Spot for exploratory large runs.
  • Keep soak tests on the smallest instance count that sustains the target rate (long duration multiplies cost).