Merge pull request #15 from druids/feat/custom-autoscaling-support

SQS autoscaling support

Author: radimsuckr

autoscaling.tf

@@ -5,7 +5,6 @@ resource "aws_appautoscaling_target" "ecs_target" {
   resource_id        = "service/${var.cluster_name}/${var.name}"
   scalable_dimension = "ecs:service:DesiredCount"
   service_namespace  = "ecs"
-  depends_on         = [aws_ecs_service.application]
 
   lifecycle {
     replace_triggered_by = [
@@ -14,17 +13,18 @@ resource "aws_appautoscaling_target" "ecs_target" {
       aws_ecs_service.application.id
     ]
   }
+
+  depends_on = [aws_ecs_service.application]
 }
 
-// metric used for auto scale
 resource "aws_cloudwatch_metric_alarm" "service_cpu_high" {
   count               = var.scheduling_strategy == "REPLICA" ? 1 : 0
   alarm_name          = "${local.name_underscore}_cpu_utilization_high"
   comparison_operator = "GreaterThanOrEqualToThreshold"
-  evaluation_periods  = "1"
+  evaluation_periods  = 1
   metric_name         = "CPUUtilization"
   namespace           = "AWS/ECS"
-  period              = "60"
+  period              = 60
   statistic           = "Average"
   threshold           = var.scale_up
 
@@ -34,17 +34,16 @@ resource "aws_cloudwatch_metric_alarm" "service_cpu_high" {
   }
 
   alarm_actions = [aws_appautoscaling_policy.up[0].arn]
-  depends_on    = [aws_appautoscaling_policy.up[0]]
 }
 
 resource "aws_cloudwatch_metric_alarm" "service_cpu_low" {
   count               = var.scheduling_strategy == "REPLICA" ? 1 : 0
   alarm_name          = "${local.name_underscore}_cpu_utilization_low"
   comparison_operator = "LessThanThreshold"
-  evaluation_periods  = "1"
+  evaluation_periods  = 1
   metric_name         = "CPUUtilization"
   namespace           = "AWS/ECS"
-  period              = "60"
+  period              = 60
   statistic           = "Average"
   threshold           = var.scale_down
 
@@ -54,15 +53,54 @@ resource "aws_cloudwatch_metric_alarm" "service_cpu_low" {
   }
 
   alarm_actions = [aws_appautoscaling_policy.down[0].arn]
-  depends_on    = [aws_appautoscaling_policy.down[0]]
+}
+
+resource "aws_cloudwatch_metric_alarm" "sqs_up" {
+  count = var.scheduling_strategy == "REPLICA" && var.sqs_scaling_enabled ? 1 : 0
+
+  alarm_name          = "${local.name}-up"
+  comparison_operator = "GreaterThanOrEqualToThreshold"
+  evaluation_periods  = var.sqs_scaling_evaluation_periods
+  metric_name         = var.sqs_scaling_metric_name
+  namespace           = "AWS/SQS"
+  period              = var.sqs_scaling_period
+  statistic           = "Maximum"
+  threshold           = var.sqs_scaling_threshold
+
+  dimensions = {
+    QueueName = var.sqs_scaling_queue_name
+  }
+
+  alarm_actions     = [aws_appautoscaling_policy.sqs_up[0].arn]
+  alarm_description = "Monitors ApproximateAgeOfOldestMessage in SQS queue"
+}
+
+resource "aws_cloudwatch_metric_alarm" "sqs_down" {
+  count = var.scheduling_strategy == "REPLICA" && var.sqs_scaling_enabled ? 1 : 0
+
+  alarm_name          = "${local.name}-down"
+  comparison_operator = "LessThanThreshold"
+  evaluation_periods  = var.sqs_scaling_evaluation_periods
+  metric_name         = var.sqs_scaling_metric_name
+  namespace           = "AWS/SQS"
+  period              = var.sqs_scaling_period
+  statistic           = "Maximum"
+  threshold           = var.sqs_scaling_threshold
+
+  dimensions = {
+    QueueName = var.sqs_scaling_queue_name
+  }
+
+  alarm_actions     = [aws_appautoscaling_policy.sqs_down[0].arn]
+  alarm_description = "Monitors ApproximateAgeOfOldestMessage in SQS queue"
 }
 
 resource "aws_appautoscaling_policy" "up" {
   count              = var.scheduling_strategy == "REPLICA" ? 1 : 0
   name               = "${local.name_underscore}_scale_up"
-  service_namespace  = "ecs"
-  resource_id        = "service/${var.cluster_name}/${var.name}"
-  scalable_dimension = "ecs:service:DesiredCount"
+  resource_id        = aws_appautoscaling_target.ecs_target[0].resource_id
+  scalable_dimension = aws_appautoscaling_target.ecs_target[0].scalable_dimension
+  service_namespace  = aws_appautoscaling_target.ecs_target[0].service_namespace
 
   step_scaling_policy_configuration {
     adjustment_type         = "ChangeInCapacity"
@@ -74,16 +112,14 @@ resource "aws_appautoscaling_policy" "up" {
       scaling_adjustment          = 1
     }
   }
-
-  depends_on = [aws_appautoscaling_target.ecs_target]
 }
 
 resource "aws_appautoscaling_policy" "down" {
   count              = var.scheduling_strategy == "REPLICA" ? 1 : 0
   name               = "${local.name_underscore}_scale_down"
-  service_namespace  = "ecs"
-  resource_id        = "service/${var.cluster_name}/${var.name}"
-  scalable_dimension = "ecs:service:DesiredCount"
+  resource_id        = aws_appautoscaling_target.ecs_target[0].resource_id
+  scalable_dimension = aws_appautoscaling_target.ecs_target[0].scalable_dimension
+  service_namespace  = aws_appautoscaling_target.ecs_target[0].service_namespace
 
   step_scaling_policy_configuration {
     adjustment_type         = "ChangeInCapacity"
@@ -95,6 +131,44 @@ resource "aws_appautoscaling_policy" "down" {
       scaling_adjustment          = -1
     }
   }
+}
 
-  depends_on = [aws_appautoscaling_target.ecs_target]
+resource "aws_appautoscaling_policy" "sqs_up" {
+  count = var.scheduling_strategy == "REPLICA" && var.sqs_scaling_enabled ? 1 : 0
+
+  name               = "${local.name_underscore}_sqs_up"
+  resource_id        = aws_appautoscaling_target.ecs_target[0].resource_id
+  scalable_dimension = aws_appautoscaling_target.ecs_target[0].scalable_dimension
+  service_namespace  = aws_appautoscaling_target.ecs_target[0].service_namespace
+
+  step_scaling_policy_configuration {
+    adjustment_type         = "ChangeInCapacity"
+    cooldown                = var.cooldown
+    metric_aggregation_type = "Maximum"
+
+    step_adjustment {
+      metric_interval_lower_bound = 0
+      scaling_adjustment          = 1
+    }
+  }
+}
+
+resource "aws_appautoscaling_policy" "sqs_down" {
+  count = var.scheduling_strategy == "REPLICA" && var.sqs_scaling_enabled ? 1 : 0
+
+  name               = "${local.name_underscore}_sqs_down"
+  resource_id        = aws_appautoscaling_target.ecs_target[0].resource_id
+  scalable_dimension = aws_appautoscaling_target.ecs_target[0].scalable_dimension
+  service_namespace  = aws_appautoscaling_target.ecs_target[0].service_namespace
+
+  step_scaling_policy_configuration {
+    adjustment_type         = "ChangeInCapacity"
+    cooldown                = var.cooldown
+    metric_aggregation_type = "Maximum"
+
+    step_adjustment {
+      metric_interval_lower_bound = 0
+      scaling_adjustment          = -1
+    }
+  }
 }

container.tf

@@ -121,4 +121,7 @@ module "container_definition_nginx" { // Nginx task
       awslogs-stream-prefix = "ecs"
     }
   }
+
+  environment = []
+  secrets     = []
 }

service.tf

@@ -37,7 +37,6 @@ resource "aws_ecs_service" "application" {
     }
   }
 
-
   lifecycle {
     ignore_changes = [desired_count]
   }

variables.tf

@@ -267,3 +267,33 @@ variable "image_scanning" {
   default = false
   type    = bool
 }
+
+variable "sqs_scaling_enabled" {
+  default = false
+  type    = bool
+}
+
+variable "sqs_scaling_evaluation_periods" {
+  default = 60
+  type    = number
+}
+
+variable "sqs_scaling_metric_name" {
+  default = "ApproximateAgeOfOldestMessage"
+  type    = string
+}
+
+variable "sqs_scaling_period" {
+  default = 120
+  type    = number
+}
+
+variable "sqs_scaling_queue_name" {
+  default = null
+  type    = string
+}
+
+variable "sqs_scaling_threshold" {
+  default = 60
+  type    = number
+}