When creating the autoscaling policy and cloudwatch alarm resource in Terraform v13, they are created fine. However, when load testing the endpoint, they successfully scale up instances, but fail to scale down when CPU utilization hits the necessary % for a period of time. The error looks like this:

"historySummary": "Failed to execute AutoScaling action: No step adjustment found for metric value [5.99763732496649, 2.7634547331059975] and breach delta -4.00236267503351"

Listed below are the terraform resources:

Autoscaling Policy -

resource "aws_appautoscaling_policy" "frontend_down" {
  name               = "${var.name}_frontend_scale_down"
  service_namespace  = "ecs"
  resource_id        = "service/${aws_ecs_cluster.main.name}/${aws_ecs_service.frontend.name}"
  scalable_dimension = "ecs:service:DesiredCount"

  step_scaling_policy_configuration {
    adjustment_type         = "ChangeInCapacity"
    cooldown                = 30
    metric_aggregation_type = "Maximum"

    step_adjustment {
      metric_interval_lower_bound = 0
      scaling_adjustment          = -1
    }
  }

  depends_on = [aws_appautoscaling_target.frontend_target]
}

Cloudwatch Alarm -

resource "aws_cloudwatch_metric_alarm" "frontend_service_cpu_low" {
  alarm_name          = "${var.name}_cpu_utilization_low_fe"
  parison_operator = "LessThanOrEqualToThreshold"
  evaluation_periods  = "2"
  metric_name         = "CPUUtilization"
  namespace           = "AWS/ECS"
  period              = "60"
  statistic           = "Average"
  threshold           = "10"

  dimensions = {
    ClusterName = var.ecs_cluster_name
    ServiceName = var.ecs_service_name_frontend
  }

  alarm_actions = [var.autoscaling_down_arn_frontend]

  tags = {
    Name        = "${var.name}-autoscaling"
    BillingCode = var.billing_code_tag
    Environment = var.environment_tag
  }
}

When creating the autoscaling policy and cloudwatch alarm resource in Terraform v13, they are created fine. However, when load testing the endpoint, they successfully scale up instances, but fail to scale down when CPU utilization hits the necessary % for a period of time. The error looks like this:

"historySummary": "Failed to execute AutoScaling action: No step adjustment found for metric value [5.99763732496649, 2.7634547331059975] and breach delta -4.00236267503351"

Listed below are the terraform resources:

Autoscaling Policy -

resource "aws_appautoscaling_policy" "frontend_down" {
  name               = "${var.name}_frontend_scale_down"
  service_namespace  = "ecs"
  resource_id        = "service/${aws_ecs_cluster.main.name}/${aws_ecs_service.frontend.name}"
  scalable_dimension = "ecs:service:DesiredCount"

  step_scaling_policy_configuration {
    adjustment_type         = "ChangeInCapacity"
    cooldown                = 30
    metric_aggregation_type = "Maximum"

    step_adjustment {
      metric_interval_lower_bound = 0
      scaling_adjustment          = -1
    }
  }

  depends_on = [aws_appautoscaling_target.frontend_target]
}

Cloudwatch Alarm -

resource "aws_cloudwatch_metric_alarm" "frontend_service_cpu_low" {
  alarm_name          = "${var.name}_cpu_utilization_low_fe"
  parison_operator = "LessThanOrEqualToThreshold"
  evaluation_periods  = "2"
  metric_name         = "CPUUtilization"
  namespace           = "AWS/ECS"
  period              = "60"
  statistic           = "Average"
  threshold           = "10"

  dimensions = {
    ClusterName = var.ecs_cluster_name
    ServiceName = var.ecs_service_name_frontend
  }

  alarm_actions = [var.autoscaling_down_arn_frontend]

  tags = {
    Name        = "${var.name}-autoscaling"
    BillingCode = var.billing_code_tag
    Environment = var.environment_tag
  }
}

• javascript
• amazon-web-services
• terraform
• amazon-cloudwatch
• autoscaling

Share Improve this question Follow asked Apr 8, 2021 at 7:31 narliechollernarliecholler 47988 silver badges2222 bronze badges 1

• figured out the reason, it was because on scaling down policy I was using 'metric_interval_lower_bound' instead of 'metric_interval_upper_bound'. When scaling down it provides a negative delta when paring to the alarm threshold and cloudwatch metric and thus 0 bees the upper bound. When scaling up, you use lower bound as it provides a positive delta. – narliecholler Commented Apr 8, 2021 at 13:50

Add a ment  | 

2 Answers 2

Sorted by: Reset to default 8

figured out the reason, it was because on scaling down policy I was using 'metric_interval_lower_bound' instead of 'metric_interval_upper_bound'. When scaling down it provides a negative delta when paring to the alarm threshold and cloudwatch metric and thus 0 bees the upper bound. When scaling up, you use lower bound as it provides a positive delta

@narliecholler As you figured out, you need to check the values set for the auto scaling.

For instance:

prod-cpu-ScaleUp: CPUUtilization >= 50
Take the action: Add 2 tasks when 65 <= CPUUtilization

In above case, you will receive similar below error message because the scaling doesn't know how to handle the CPUUtilization actions between 50 and 65.

"No step adjustment found for metric value [52.01] and breach delta 2.01"