net-ilb

Internal Load Balancer Module

This module allows managing a GCE Internal Load Balancer and integrates the forwarding rule, regional backend, and optional health check resources. It's designed to be a simple match for the compute-vm module, which can be used to manage instance templates and instance groups.

Issues

TODO(ludoo): check if this is still the case after splitting out MIG from compute-vm

There are some corner cases (eg when switching the instance template from internal service account to an externally managed one) where Terraform raises a cycle error on apply. In these situations, run successive applies targeting resources used in the template first then the template itself, and the cycle should be fixed.

One other issue is a Provider produced inconsistent final plan error which is sometimes raised when switching template version. This seems to be related to this open provider issue, but it's relatively harmless since the resource is updated, and subsequent applies raise no errors.

Examples

Externally managed instances

This examples shows how to create an ILB by combining externally managed instances (in a custom module or even outside of the current root module) in an unmanaged group. When using internally managed groups, remember to run terraform apply each time group instances change.

module "ilb" {
  source        = "./fabric/modules/net-ilb"
  project_id    = var.project_id
  region        = "europe-west1"
  name          = "ilb-test"
  service_label = "ilb-test"
  vpc_config = {
    network    = var.vpc.self_link
    subnetwork = var.subnet.self_link
  }
  group_configs = {
    my-group = {
      zone = "europe-west1-b"
      instances = [
        "instance-1-self-link",
        "instance-2-self-link"
      ]
    }
  }
  backends = [{
    group = module.ilb.groups.my-group.self_link
  }]
  health_check_config = {
    http = {
      port = 80
    }
  }
}
# tftest modules=1 resources=4

End to end example

This example spins up a simple HTTP server and combines four modules:

• nginx from the cloud-config-container collection, to manage instance configuration
• compute-vm to manage the instance template and unmanaged instance group
• this module to create an Internal Load Balancer in front of the managed instance group

Note that the example uses the GCE default service account. You might want to create an ad-hoc service account by combining the iam-service-account module, or by having the GCE VM module create one for you. In both cases, remember to set at least logging write permissions for the service account, or the container on the instances won't be able to start.

module "cos-nginx" {
  source = "./fabric/modules/cloud-config-container/nginx"
}

module "instance-group" {
  source     = "./fabric/modules/compute-vm"
  for_each   = toset(["b", "c"])
  project_id = var.project_id
  zone       = "europe-west1-${each.key}"
  name       = "ilb-test-${each.key}"
  network_interfaces = [{
    network    = var.vpc.self_link
    subnetwork = var.subnet.self_link
    nat        = false
    addresses  = null
  }]
  boot_disk = {
    image = "projects/cos-cloud/global/images/family/cos-stable"
    type  = "pd-ssd"
    size  = 10
  }
  tags = ["http-server", "ssh"]
  metadata = {
    user-data = module.cos-nginx.cloud_config
  }
  group = { named_ports = {} }
}

module "ilb" {
  source        = "./fabric/modules/net-ilb"
  project_id    = var.project_id
  region        = "europe-west1"
  name          = "ilb-test"
  service_label = "ilb-test"
  vpc_config = {
    network    = var.vpc.self_link
    subnetwork = var.subnet.self_link
  }
  ports = [80]
  backends = [
    for z, mod in module.instance-group : {
      group          = mod.group.self_link
      balancing_mode = "UTILIZATION"
    }
  ]
  health_check_config = {
    http = {
      port = 80
    }
  }
}
# tftest modules=3 resources=7

Variables

name	description	type	required	default
name	Name used for all resources.	string	✓
project_id	Project id where resources will be created.	string	✓
region	GCP region.	string	✓
vpc_config	VPC-level configuration.	object({…})	✓
address	Optional IP address used for the forwarding rule.	string		null
backend_service_config	Backend service level configuration.	object({…})		{}
backends	Load balancer backends, balancing mode is one of 'CONNECTION' or 'UTILIZATION'.	list(object({…}))		[]
description	Optional description used for resources.	string		"Terraform managed."
global_access	Global access, defaults to false if not set.	bool		null
group_configs	Optional unmanaged groups to create. Can be referenced in backends via outputs.	map(object({…}))		{}
health_check	Name of existing health check to use, disables auto-created health check.	string		null
health_check_config	Optional auto-created health check configuration, use the output self-link to set it in the auto healing policy. Refer to examples for usage.	object({…})		{…}
labels	Labels set on resources.	map(string)		{}
ports	Comma-separated ports, leave null to use all ports.	list(string)		null
protocol	IP protocol used, defaults to TCP.	string		"TCP"
service_label	Optional prefix of the fully qualified forwarding rule name.	string		null

Outputs

name	description	sensitive
backend_service	Backend resource.
backend_service_id	Backend id.
backend_service_self_link	Backend self link.
forwarding_rule	Forwarding rule resource.
forwarding_rule_address	Forwarding rule address.
forwarding_rule_id	Forwarding rule id.
forwarding_rule_self_link	Forwarding rule self link.
group_self_links	Optional unmanaged instance group self links.
groups	Optional unmanaged instance group resources.
health_check	Auto-created health-check resource.
health_check_self_id	Auto-created health-check self id.
health_check_self_link	Auto-created health-check self link.