When you shape a map with a for expression in Terraform, it sorts and lays out the keys in ascending order.
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 > {for k, v in {"a":3, "c":2, "d":1}: v => k} { "1" = "d" "2" = "c" "3" = "a" } > values({for k, v in {"a":3, "c":2, "d":1}: v => k}) [ "d", "c", "a", ]
Using this, let’s retrieve the Subnet IDs obtained from a data resource as a list ordered by AZ (a → c → d).
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 data "aws_subnets" "public" { filter { name = "vpc-id" values = ["xxxxx"] } filter { name = "tag:Name" values = ["public-*"] } } data "aws_subnet" "public" { for_each = toset(data.aws_subnets.public.ids) id = each.value } values({ for id, s in data.aws_subnet.public : s.availability_zone => id }) # [ # "subnet-xxx", # ap-northeast-1a # "subnet-yyy", # ap-northeast-1c # "subnet-zzz", # ap-northeast-1d # ]
This is handy when, for example, you launch 10 EC2 instances and want to distribute the subnet AZs in order as a, c, d.
1 2 3 4 5 6 7 8 9 locals { az_count = length(data.aws_availability_zones.available.names) } resource "aws_instance" "app" { count = 10 # NOTE: By using a map type, we take advantage of the keys being ordered in ascending order to arrange the AZs as a → c → d subnet_id = values({ for id, s in data.aws_subnet.public : s.availability_zone => id })[count.index % local.az_count]
That’s all. I hope this is helpful.
