GCP locations configuration
Private Locations on your GCP account.
Instance specifications
We recommend that you use for your own load generator instances with at least 4 cores.
As Gatling load tests tend to be CPU bound, we recommend using instances of the “Compute Optimized” family.
As a result, we recommend using c3-highcpu-4
instances or larger.
You might want to tune the Xmx
JVM options to half of the physical memory.
See jvm-options
configuration below.
If you don’t, the JVM will use a max heap size of 1/4th of the physical memory.
Permissions
GCP private locations require the control plane to have GCP access rights configured in order to instantiate virtual machines.
GCP private locations require the control plane to have credentials configured in order to instantiate virtual machines and associated resources.
Access rights can be set through a service account associated with your control plane.
Check GCP and Gatling documentation pages for more details:
System requirements
GCP private locations rely on some dependencies.
So when using a custom image, make sure following are available:
- cloud-init integration.
- jq a lightweight and flexible command-line JSON processor.
- curl a command line tool and library for transferring data with URLs
- Java runtime environment: OpenJDK 64bits LTS versions: 11, 17 or 21 (see Gatling prerequisites)
Control plane configuration file
control-plane {
# Control plane token
token = "cpt_example_c7oze5djp3u14a5xqjanh..."
# Control plane token with an environment variable
token = ${?CONTROL_PLANE_TOKEN}
# Control plane description (optional)
description = "my control plane description"
# Locations configurations
locations = [
{
# Private location ID, must be prefixed by prl_, only consist of numbers 0-9,
# lowercase letters a-z, and underscores, with a max length of 30 characters
id = "prl_private_location_example"
# Private location description (optional)
description = "Private Location on GCP"
# Private location type
type = "gcp"
# GCP location name, as listed by GCP CLI:
# gcloud compute zones list
zone = "europe-west3-a"
# Instance template (alternative to machine)
# instance-template = "example-template"
# Machine configuration (alternative to instance template)
machine {
# Virtual machine type, as listed by GCP CLI:
# gcloud compute machine-types list --filter="zone:( europe-west3-a )"
type = "c3-highcpu-4"
# Configure load generators instances as preemptible or not. (optional, default: false)
# preemptible = true
# Engine (optional, default classic)
engine = "classic" # Possible values: classic or javascript
# Certified image configuration
image {
type = "certified"
java = "latest" # See engine section
}
# Custom image configuration (alternative to certified image)
# Base your location on a custom image from its name
# image {
# type = "custom"
# project = "my-project"
# id = "image-name"
# }
# Custom image configuration (alternative to certified image)
# Base your location on a custom image from its image family
# image {
# type = "custom"
# project = "my-project"
# family = "image-family"
# }
# Storage configuration
disk {
# Disk size in Gb (mininum 20Gb)
sizeGb = 20
}
# Network interface (optional)
network-interface {
# Network name on your project (optional)
# Not needed if subnetwork is configured
# network = "gatling-network"
# Subnetwork name on your project (optional)
# subnetwork = "gatling-subnetwork-europe-west3"
# Associate external IP to instance (optional, default to true)
# See Cloud NAT when set to false
# with-external-ip = true
}
}
# GCP project id as returned by GCP CLI:
# gcloud projects list
project = "my-project-id"
# Java configuration (the following configuration properties are optional)
# System properties (optional)
system-properties {
"java.net.preferIPv6Addresses" = "true"
}
# Overwrite JAVA_HOME definition (optional)
# java-home = "/usr/lib/jvm/zulu"
# JVM Options (optional)
# Default ones, that can be overridden with precedence:
# [
# "-XX:MaxInlineLevel=20",
# "-XX:MaxTrivialSize=12",
# "-XX:+IgnoreUnrecognizedVMOptions",
# "--add-opens=java.base/java.nio=ALL-UNNAMED",
# "--add-opens=java.base/jdk.internal.misc=ALL-UNNAMED"
# ]
# Based on your instance configuration, you may want to update Xmx and Xms values.
# jvm-options = ["-Xmx4G", "-Xms512M"]
}
]
}
Engine
The engine specified for a location determines the compatible package formats (JavaScript or JVM) for Gatling packages.
Each engine (classic
or javascript
) supports specific Java versions, where latest
is defaulted.
The table below outlines the supported Java versions for certified Gatling images:
Engine | Supported Java Versions |
---|---|
classic | 21 or latest |
javascript | latest |
javascript
engine, only the latest Java version is supported, which corresponds to the GraalVM version used to run Gatling with JavaScript.Internet access for your Load Generators instances
Cloud NAT (Network Address Translation) lets certain resources in Google Cloud create outbound connections to the internet or to other Virtual Private Cloud (VPC) networks. Cloud NAT supports address translation for established inbound response packets only. It does not allow unsolicited inbound connections.
More info on Cloud NAT overview.
Instances with no external IP
An instance without an external IP cannot access to internet without a Cloud NAT configured on the network, for the region. Load generators must have access to some outbound domains in order to run (see Private locations introduction).
In the GCP management console, open the Cloud NAT (or search for “Cloud NAT” in the search bar).
Set the Gateway:
- name
- region base on your location zone
- create a router
Set static IPs
Cloud NAT gateway can be configured with static IP addresses. See limits of concurrent connections with Cloud NAT on Cloud NAT port reservation and make sure to provide enough static IP addresses based on the load you need to generate.
Configure instances using Terraform
Gatling provides Terraform modules to set up GCP infrastructure for Private Locations. One module specifies the load generator location(s), and the second module deploys the control plane. To use the Terraform module, visit our dedicated GitHub repository