Splat doesn't work properly

[hashibot]

This issue was originally opened by @slikk66 as hashicorp/terraform#13397. It was migrated here as part of the provider split. The original body of the issue is below.

---

Hi there,

Thank you for opening an issue. Please note that we try to keep the Terraform issue tracker reserved for bug reports and feature requests. For general usage questions, please see: https://www.terraform.io/community.html.

Terraform Version

terraform -v

Terraform v0.9.2

Affected Resource(s)

Please list the resources as a list, for example:

• aws_elasticache_cluster (at least)

If this issue appears to affect multiple resources, it may be an issue with Terraform's core, so please mention this. (think it does)

Expected Behavior

create output of multiple cache nodes using splat:

output "memcached_ha_endpoints" {
  value = ["${aws_elasticache_cluster.memcached_ha.cache_nodes.*.address}"]
}

Try to read in these addresses for DNS:

resource "aws_route53_record" "memcached_ha" {

  count   = "${var.cache_engine == "memcached_ha" ? "${var.cache_ha_node_count}" : 0}"

  zone_id = "${aws_route53_zone.dns_internal_zone.zone_id}"
  name    = "${format("memcached-%03d.%s", count.index+1, "${var.internal_dns_tld}")}"
  type    = "CNAME"
  ttl     = "300"
  records = ["${var.memcached_ha_endpoints[count.index]}"]
}

Actual Behavior

Error running plan: 1 error(s) occurred:

* module.dns.module.dns_internal.aws_route53_record.memcached_ha: 1 error(s) occurred:

* module.dns.module.dns_internal.aws_route53_record.memcached_ha[1]: index 1 out of range for list var.memcached_ha_endpoints (max 1) in:

${var.memcached_ha_endpoints[count.index]}

However, if you define the outputs using hardcoded indexes, it works:

output "memcached_ha_endpoints" {
  value = [
    "${aws_elasticache_cluster.memcached_ha.cache_nodes.0.address}",
    "${aws_elasticache_cluster.memcached_ha.cache_nodes.1.address}"
  ]
}

[imauriciovargas]

Using version 0.10.0

variable "num_nodes" {
  defaul  = 2
}

resource "aws_elasticache_cluster" "memcached" {
  cluster_id = "foo"
  engine = "memcached"
  num_cache_nodes = "${var.num_nodes}"
  ...
}

resource "aws_cloudwatch_metric_alarm" "CPU" {
  count                     = "${var.num_nodes}"
  depends_on                = ["aws_elasticache_cluster.memcached"]
  ...
  dimensions                {
      CachedClusterId = "${aws_elasticache_cluster.memcached.id}",
      CachedNodeId    = "${element(aws_elasticache_cluster.memcached.cache_nodes.*.id, count.index)}"
  }
}

Error applying plan:

2 error(s) occurred:

* module.memcached.aws_cloudwatch_metric_alarm.CPU[1]: Resource 'aws_elasticache_cluster.memcached' does not have attribute 'cache_nodes.*.id' for variable 'aws_elasticache_cluster.memcached.cache_nodes.
*.id'
* module.memcached.aws_cloudwatch_metric_alarm.CPU[0]: Resource 'aws_elasticache_cluster.memcached' does not have attribute 'cache_nodes.*.id' for variable 'aws_elasticache_cluster.memcached.cache_nodes.
*.id'

Checking the aws_elasticache_cluyster output for cache_nodes:

output "memcached_nodes" {
    value = "${aws_elasticache_cluster.memcached.cache_nodes.*.id}"
}

Output should return a list of node IDs, however it returns a string with the number for nodes in cluster:

"outputs": {
                "memcached_nodes": {
                    "sensitive": false,
                    "type": "string",
                    "value": "2"
                }
            },

If remove the splat *;

output "memcached_nodes" {
    value = "${aws_elasticache_cluster.memcached.cache_nodes}"
}

returns a list with all the node info as it should.

"outputs": {
                "memcached_nodes": {
                    "sensitive": false,
                    "type": "list",
                    "value": [
                        {
                            "address": "foo.ncubqb.0001.use1.cache.amazonaws.com",
                            "availability_zone": "us-east-1b",
                            "id": "0001",
                            "port": "11211"
                        },
                        {
                            "address": "foo.ncubqb.0002.use1.cache.amazonaws.com",
                            "availability_zone": "us-east-1c",
                            "id": "0002",
                            "port": "11211"
                        }
                    ]
                }
            },

It seems like splat * just returns count. Also tried with other attributes and get the same issue.

The workaround is to manually create each resource and reference the nodes individually (in this case CloudWatch alarm).

[bflad]

Hi folks 👋 This issue is resolved in Terraform 0.12, which fully supports indexed splat (*) notation with list attribute values.

Given this configuration:

terraform {
  required_providers {
    aws = "2.20.0"
  }
  required_version = "0.12.5"
}

provider "aws" {
  region = "us-east-2"
}

resource "aws_elasticache_cluster" "test" {
  cluster_id      = "bflad-testing"
  engine          = "memcached"
  node_type       = "cache.t2.micro"
  num_cache_nodes = 3
}

output "test1" {
  value = aws_elasticache_cluster.test.cache_nodes[*].address
}

output "test2" {
  value = element(aws_elasticache_cluster.test.cache_nodes[*].address, 1)
}

output "test3" {
  value = join(",", formatlist("%s:%s", aws_elasticache_cluster.test.cache_nodes[*].address, aws_elasticache_cluster.test.cache_nodes[*].port))
}

We can get the various expected outputs that were problematic in Terraform 0.11 and earlier:

Outputs:

test1 = [
  "bflad-testing.asxu6y.0001.use2.cache.amazonaws.com",
  "bflad-testing.asxu6y.0002.use2.cache.amazonaws.com",
  "bflad-testing.asxu6y.0003.use2.cache.amazonaws.com",
]
test2 = bflad-testing.asxu6y.0002.use2.cache.amazonaws.com
test3 = bflad-testing.asxu6y.0001.use2.cache.amazonaws.com:11211,bflad-testing.asxu6y.0002.use2.cache.amazonaws.com:11211,bflad-testing.asxu6y.0003.use2.cache.amazonaws.com:11211

Enjoy! 🚀

[ghost]

I'm going to lock this issue because it has been closed for 30 days ⏳. This helps our maintainers find and focus on the active issues.

If you feel this issue should be reopened, we encourage creating a new issue linking back to this one for added context. Thanks!