For our ECS service definition we have this snippet at the
`load_balancer`.
The `target_group_arn` is being pupulated by an external service that
returns the arn based on a simple string from our microservices list.
If the arn changed, this would not cause a recreation of the service and
leaving a dangling pointer to an arn that does not exist anymore.
```
load_balancer {
target_group_arn = "${lookup(var.target_group_mapping, element(values(var.microservices), count.index))}"
container_name = "${element(values(var.microservices), count.index)}"
container_port = "${var.container_port}"
}
```
The fix is adding another field to the set that's creating the ELB/ALB
definition. From looking into the git history seems this code was
created prior to ALB thus not having this field available at the time.
Service is being recreated as expected, no other services are affected
(expected behavior)
Fixes: #13216
Prior to Terraform 0.9.2, we always set placement_strategies to
lowercase. Therefore, people using it in Terraform 0.9.2 are getting
continual diffs:
```
-/+ aws_ecs_service.mongo
cluster: "arn:aws:ecs:us-west-2:187416307283:cluster/terraformecstest1" => "arn:aws:ecs:us-west-2:187416307283:cluster/terraformecstest1"
deployment_maximum_percent: "200" => "200"
deployment_minimum_healthy_percent: "100" => "100"
desired_count: "1" => "1"
name: "mongodb" => "mongodb"
placement_strategy.#: "1" => "1"
placement_strategy.1676812570.field: "instanceid" => "" (forces new resource)
placement_strategy.1676812570.type: "spread" => "" (forces new resource)
placement_strategy.3946258308.field: "" => "instanceId" (forces new resource)
placement_strategy.3946258308.type: "" => "spread" (forces new resource)
task_definition: "arn:aws:ecs:us-west-2:187416307283:task-definition/mongodb:1991" => "arn:aws:ecs:us-west-2:187416307283:task-definition/mongodb:1991"
Plan: 1 to add, 0 to change, 1 to destroy.
```
This adds a DiffSuppression func to make sure this doesn't trigger a
ForceNew resource:
```
% terraform plan ✹ ✭
[WARN] /Users/stacko/Code/go/bin/terraform-provider-aws overrides an internal plugin for aws-provider.
If you did not expect to see this message you will need to remove the old plugin.
See https://www.terraform.io/docs/internals/internal-plugins.html
Refreshing Terraform state in-memory prior to plan...
The refreshed state will be used to calculate this plan, but will not be
persisted to local or remote state storage.
aws_ecs_cluster.default: Refreshing state... (ID: arn:aws:e...ecstest1)
aws_ecs_task_definition.mongo: Refreshing state... (ID: mongodb)
aws_ecs_service.mongo: Refreshing state... (ID: arn:aws:e.../mongodb)
No changes. Infrastructure is up-to-date.
This means that Terraform did not detect any differences between your
configuration and real physical resources that exist. As a result, Terraform
doesn't need to do anything.
```
```
```
AWS API requires ECS placement strategies "field" attribute to be
"memory" or "cpu" (lowercase) when type=bin, but these read back as
"MEMORY" and "CPU" (uppercase) respectively.
PR #11565 (which fixed separately reported #11644) deals with this by
always lowering the case of the resource received from the API, but this
breaks for other "field" values (e.g. "instanceId" -> "instanceid").
This PR only lowers the case of the returned resource when field
"MEMORY" or "CPU". Haven't checked if any other fields need this
treatment.
The API asks you to send lower case values, but returns uppercase ones.
Here we lowercase the returned API values.
There is no migration here because the field in question is nested in a
set, so the hash will change regardless. Anyone using this feature now
has it broken anyway.
Allows users from govcloud and other regions (aws-cn) to now use the following resources correctly:
```
- data "aws_billing_service_account"
- data "aws_elb_service_account"
- resource "aws_cloudfront_origin_access_identity"
- resource "aws_ecs_service"
- resource "aws_iam_saml_provider"
- resource "aws_lambda_permission"
- resource "aws_sns_topic_policy"
```
Updates ECS task_definition documentation, and schema validation functions to match the AWS API documentation.
Updates ECS service documentation, and schema validation functions match the AWS API documentation.
Change the `RetryFunc` from a plain `error` return type to a
specialized `RetryError` which must decide whether it is
retryable or not.
Add `RetryableError` / `NonRetryableError` factory functions that
callers are meant to use to build up these errors.
This makes it eminently clear whether or not a given error is
retryable from inside the client code.
Goal here is to _not_ change any behavior, simply reflect the
existing behavior with the new, clearer, API.
All of these RetryErrors were meant to fail right away, but instead
caused retry looping because the typecheck in the implementation of
`resource.Retry()` only catches the value type, and not the pointer
type.
Refs #5537
Radek Simko