It's sad that terraform is not backward compatible.
data "aws_security_group" "security_groupdev" {
filter {
name = "group-name"
values = ["SecurityGroupdev"]
}
}
resource "aws_instance" "ec2_instance" {
count = "${var.ec2_instance_count}"
...
}
resource "aws_network_interface_sg_attachment" "sg_attachment" {
security_group_id = "${data.aws_security_group.security_groupdev.id}"
network_interface_id = "${aws_instance.ec2_instance.primary_network_interface_id}"
}
but after upgrading it to Terraform 0.12 I have started facing issues and I am not able to get the syntax for TF0.12.
Error: Missing resource instance key
on ..\resources\ec2_instance\main.tf line 101, in resource "aws_network_interface_sg_attachment" "sg_attachment":
101: network_interface_id = "${aws_instance.ec2_instance.primary_network_interface_id}"
Because aws_instance.ec2_instance has "count" set, its attributes must be
accessed on specific instances.
I tried "${aws_instance.ec2_instance[count.index].primary_network_interface_id}" but no luck.
Related
Most of our ECS Services are not in the "new" format that allow tags to be set.
We recently added default tags to our aws provider e.g.:
provider "aws" {
region = local.workspace["aws_region"]
profile = local.workspace["aws_profile"]
default_tags {
tags = {
Environment = local.workspace["releaseStage"]
Owner = "terraform"
Project = "infrastructure"
Account = local.workspace["releaseStage"]
}
}
}
However, if we run terraform apply, it barks at ecs service resources as they don't support tagging:
Error: error updating ECS Service (arn:aws:ecs:us-east-1:xxxx:service/myservice) tags: error tagging resource (arn:aws:ecs:us-east-1:xxxx:service/myservice): InvalidParameterException: Long arn format must be used for tagging operations
If I override the tags in the resource e.g.:
resource "aws_ecs_service" "myservice" {
name = "myservice"
...
tags = {
Environment = ""
Owner = ""
Project = ""
Account = ""
}
}
It works, but I never get a clean terraform plan as it always needs to evaluate the merged tags.
Is there a way to exclude tagging of default_tags with certain resources?
You should be able to use lifecycle block with ignore_changes to tags which will exclude any external changes. I'm not sure whether this would work. But you can try.
I have defined an aws_eks_cluster and aws_eks_node_group as follows:
resource "aws_eks_cluster" "example" {
count = var.create_eks_cluster ? 1 : 0
name = local.cluster_name
role_arn = aws_iam_role.example[count.index].arn
vpc_config {
subnet_ids = [
aws_subnet.main2.id,
aws_subnet.main3.id
]
security_group_ids = [
module.network.security_group_allow_all_from_client_ip,
module.network.security_group_main_id
]
endpoint_private_access = true
endpoint_public_access = false
}
# Ensure that IAM Role permissions are created before and deleted after EKS Cluster handling.
# Otherwise, EKS will not be able to properly delete EKS managed EC2 infrastructure such as Security Groups.
depends_on = [
aws_iam_role_policy_attachment.example-AmazonEKSClusterPolicy,
aws_iam_role_policy_attachment.example-AmazonEKSVPCResourceController,
]
}
resource "aws_eks_node_group" "example" {
count = var.create_eks_cluster ? 1 : 0
cluster_name = aws_eks_cluster.example[count.index].name
node_group_name = random_uuid.deployment_uuid.result
node_role_arn = aws_iam_role.eks-node-group-example[count.index].arn
subnet_ids = [
aws_subnet.main2.id,
aws_subnet.main3.id
]
scaling_config {
desired_size = 1
max_size = 5
min_size = 1
}
# Ensure that IAM Role permissions are created before and deleted after EKS Node Group handling.
# Otherwise, EKS will not be able to properly delete EC2 Instances and Elastic Network Interfaces.
depends_on = [
aws_iam_role_policy_attachment.example-AmazonEKSWorkerNodePolicy,
aws_iam_role_policy_attachment.example-AmazonEKS_CNI_Policy,
aws_iam_role_policy_attachment.example-AmazonEC2ContainerRegistryReadOnly,
]
}
How can I retrieve the KubeConfig?
I have seen that the kubeconfig is available as an output on the eks module.
Do I need to replace aws_eks_cluster and aws_eks_node_group with the eks module?
The EKS module composes a kubeconfig based on a template.
You can include that template alongside your terraform code.
You will need to provide default values for all the variables in the templatefile function call and reference your own EKS resource name. It's fine to drop all the coalescelist functions too.
e.g.:
locals {
kubeconfig = templatefile("templates/kubeconfig.tpl", {
kubeconfig_name = local.kubeconfig_name
endpoint = aws_eks_cluster.example.endpoint
cluster_auth_base64 = aws_eks_cluster.example.certificate_authority[0].data
aws_authenticator_command = "aws-iam-authenticator"
aws_authenticator_command_args = ["token", "-i", aws_eks_cluster.example.name]
aws_authenticator_additional_args = []
aws_authenticator_env_variables = {}
})
}
output "kubeconfig" { value = local.kubeconfig }
I am new to Terraform and this is my first script trying it out
provider "aws" {
profile = "default"
region = "us-east-1"
}
resource "aws_instance" "example" {
ami = "ami-2757f631"
instance_type = "t2.micro"
}
I have the script above stored on my windows desktop in C:\TerraformScripts\First.tf
Now when I run it the first time, the script executes and creates a new instance for me. I wanted to run it the second time with just changing the name from example to example2. I assumed it would create a new instance with the same configuration since I changed the name against the resource setting. But it instead destroyed the instance I had created on the first run and then recreated it again. Why is this happening without my specifying destroy?
Apologies, if I may have missed out something in the documentation, but I couldn't see it when I looked.
Thanks.
Terraform is a declarative language, which means that the script you write is telling terraform the state you want to get to (then terraform works out how to get there). It's effectively like saying "I want you to make sure I have an aws_instance", rather than "I want you to create an aws_instance".
If I'm understanding correctly, you are probably aiming to do this:
provider "aws" {
profile = "default"
region = "us-east-1"
}
resource "aws_instance" "example" {
ami = "ami-2757f631"
instance_type = "t2.micro"
}
resource "aws_instance" "example2" {
ami = "ami-2757f631"
instance_type = "t2.micro"
}
If you run terraform apply now, you will have two EC2 instances regardless of how many were created by the script previously. That's because under the hood, terraform is tracking the resources it's created previously for that script in a state file, comparing them to the current script, then working out what actions to take to make them line up.
Alternatively, you could use the count parameter to get multiple copies of the same resource:
provider "aws" {
profile = "default"
region = "us-east-1"
}
resource "aws_instance" "example" {
count = 2
ami = "ami-2757f631"
instance_type = "t2.micro"
}
I'm trying to put the RPC Permissions, along with the users and their password on an external database. I've followed the documentation for Corda v. 3.3 (https://docs.corda.net/clientrpc.html#rpc-security-management).
It says that I need to create a "security" field for the node in question and fill out all the necessary information. I've done it, but as soon as I try to deploy the Node, it gives me this error:
"Could not set unknown property 'security' for object of type net.corda.plugins.Node."
The node's information looks like this in the build.gradle document:
node {
name "O=myOrganisation,L=Lisbon,C=PT"
p2pPort 10024
rpcSettings {
address("localhost:10025")
adminAddress("localhost:10026")
}
security = {
authService = {
dataSource = {
type = "DB"
passwordEncryption = "SHIRO_1_CRYPT"
connection = {
jdbcUrl = "localhost:3306"
username = "*******"
password = "*******"
driverClassName = "com.mysql.jdbc.Driver"
}
}
}
}
cordapps = [
"$project.group:cordapp:$project.version"
]
}
You are confusing two syntaxes:
The syntax for configuring a node block inside a Cordform task such as deployNodes
The syntax for configuring a node directly via node.conf
The security settings are for inside node.conf. You have to create the node first, then modify the node's node.conf with these settings once it has been created.
Corda 4 will introduce an extraConfig option for use inside Cordfrom node blocks, as described here.
I have a set of terraform codes in a directory called myproject:
\myproject\ec2.tf
\myproject\provider.tf
\myproject\s3.tf
....
the provider.tf shows:
provider "aws" {
region = "us-west-1"
profile = "default"
}
so, if I terraform apply in myproject folder, a set of aws resources are launched in us-west-1 under my account.
Now I want to introduce a AWS Glue resource, which is only available in a different region us-west-2. then how do I layout glue.tf file?
Currently I store it in a sub-directory under myproject and run terraform apply in that sub-directory i.e.
\myproject\glue\glue.tf
\myproject\glue\another_provider.tf
another_provider.tf is:
provider "aws" {
region = "us-west-2"
profile = "default"
}
Is it the only way to store a file launching resources in different regions? any better way?
If there is no better way, then I need to have another backend file in glue sub-folder as well, besides, some common variables in myproject directory cannot be shared.
--------- update:
I followed the link posted by Phuong Nguyen,
provider "aws" {
region = "us-west-1"
profile = "default"
}
provider "aws" {
alias = "oregon"
region = "us-west-2"
profile = "default"
}
resource "aws_glue_connection" "example" {
provider = "aws.oregon"
....
}
But I saw:
Error: aws_glue_connection.example: Provider doesn't support resource: aws_glue_connection
you can use provider alias to define multiple providers, .e.g.
# this is default provider
provider "aws" {
region = "us-west-1"
profile = "default"
}
# additional provider
provider "aws" {
alias = "west-2"
region = "us-west-2"
profile = "default"
}
and then in your glue.tf, you can refer to alias provider as:
resource "aws_glue_job" "example" {
provider = "aws.west-2"
# ...
}
More details at Multiple Provider Instances section: https://www.terraform.io/docs/configuration/providers.html
Read my comment ...
Which basically means that you should keep out aws profiles and regions and what not from your terraform code as much as possible and use them as configuration as follows:
terraform {
required_version = "1.0.1"
required_providers {
aws = {
version = ">= 3.56.0"
source = "hashicorp/aws"
}
}
backend "s3" {}
}
provider "aws" {
region = var.region
profile = var.profile
}
Than use tfvars configuration files:
cat cnf/env/spe/prd/tf/03-static-website.backend-config.tfvars
profile = "prd-spe-rcr-web"
region = "eu-north-1"
bucket = "prd-bucket-spe"
foobar = "baz"
which you will apply during the terraform plan and apply calls as follows:
terraform -chdir=$tf_code_path plan -var-file=<<line-one-^^^>>.tfvars
terraform -chdir=$tf_code_path plan -var-file=<<like-the-one-^^^>>.tfvars -auto-approve
As a rule of thumb you SHOULD separate your code and configuration always, the more mixed they are the deeper you will get into troubles ... this applies to ANY programming language / project etc. Now some wise heads will argue that terraform code is in itself configuration , but no it is not. The terraform code in your application is the declarative source code, which is used to provision your binary infrastructure used by the application source code etc. in your application ...