Recently I was creating sets of AWS instances. To configure themselves
they needed access to some script functions and a place to share
configuration information between them. They would operate as a cluster
but there was no way for terraform to tell each machine the ip addresses
of the other members. However terraform could inform the last-allocated
instance the ip addresses of all the other members so I used an s3
bucket to share around that configuration information as well as some
other things.
However just creating a single s3 bucket didn’t work across regions.
While the namespace for AWS buckets appears to be global, access appears
to be limited per-region. Therefore I created one bucket per region
like so.
First I set the AWS provider and made a number of aliases for the
different regions. Note that I use AWS profiles to access config
information This saves storing credentials in configs and makes it easier
and more clear to access multiple AWS accounts.
dist.tf
:
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
| variable "profile" {}
provider "aws" {
region = "eu-west-1"
profile = "${var.profile}"
}
provider "aws" {
alias = "eu-central-1"
region = "eu-central-1"
profile = "${var.profile}"
}
provider "aws" {
alias = "eu-west-2"
region = "eu-west-2"
profile = "${var.profile}"
}
provider "aws" {
alias = "eu-west-3"
region = "eu-west-3"
profile = "${var.profile}"
}
module "eu-west-1-s3-dist" {
source = "./modules/s3-dist"
region = "eu-west-1"
}
module "eu-central-1-s3-dist" {
source = "./modules/s3-dist"
region = "eu-central-1"
providers = { aws = "aws.eu-central-1" }
}
module "eu-west-2-s3-dist" {
source = "./modules/s3-dist"
region = "eu-west-2"
providers = { aws = "aws.eu-west-2" }
}
module "eu-west-3-s3-dist" {
source = "./modules/s3-dist"
region = "eu-west-3"
providers = { aws = "aws.eu-west-3" }
}
|
The last four stanzas source this directory which contains two files.
The first is the terraform config, modules/s3-dist/dist.tf
:
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
| variable "region" {}
resource "aws_s3_bucket" "dist" {
bucket = "dist-${var.region}"
acl = "private"
region = "${var.region}"
lifecycle {
prevent_destroy = true
}
tags {
Name = "dist"
Role = "infrastructure"
}
}
resource "aws_s3_bucket_object" "script" {
bucket = "dist-${var.region}"
key = "script"
source = "${path.module}/script"
# filemd5() is in Terraform >=0.11.12
etag = "${filemd5("${path.module}/script")}"
}
|
The second is the script, modules/s3-dist/script
:
1
2
| #!/usr/local/bin/bash
echo "Lots of shell functions."
|
The instances are then told the region specific path to get the script
(this gets past user-data length restrictions for one thing) and they
also use this to synchronise certain information about their cluster.