基于SQS队列深度的Auto Scaling Group扩容异常排查求助

我用Terraform实现以下流程：创建S3存储桶及事件通知，配置S3触发的Lambda函数将消息发送到SQS队列；同时创建基于SQS队列深度（公式：消息数/实例数）的Auto Scaling Group（ASG），最大实例数设为5。我已经创建了对应的自动扩容策略，知道这类目标追踪策略无需告警即可自行触发，但执行terraform apply成功后，向SQS发送消息时ASG并没有新增实例。请问我哪里出错或者遗漏了什么？

初始代码 #

resource "aws_sqs_queue" "file_checker_sqs_queue" {
  name                      = var.sqs_queue_name
  max_message_size          = 256000
  message_retention_seconds = 86400
  visibility_timeout_seconds = 30
  receive_wait_time_seconds = 10
}

data "aws_availability_zones" "available" {
  state = "available"
  # Only retrieve the main AZs, no local zones.
  filter {
    name = "opt-in-status"
    values = ["opt-in-not-required"]
  }
}

data "aws_ami" "audio_ami" {
  most_recent = true

  filter {
    name   = "image-id"
    values = ["ami-067d1e60475437ey2"]
  }
}

resource "aws_launch_configuration" "audio_asg_launch_config" {
  name_prefix   = var.launch_config_name_prefix
  image_id      = data.aws_ami.audio_ami.id
  instance_type = "t2.micro"

  lifecycle {
    create_before_destroy = true
  }
}

resource "aws_autoscaling_group" "audio_auto_scaling_group" {
  availability_zones        = data.aws_availability_zones.available.names
  name                      = var.asg_name
  max_size                  = 5
  min_size                  = 0
  health_check_grace_period = 300
  health_check_type         = "EC2"
  force_delete              = true
  launch_configuration      = aws_launch_configuration.audio_asg_launch_config.name
}

resource "aws_autoscaling_policy" "asg_audio_scale_up_policy" {
  autoscaling_group_name = var.asg_name
  name                   = var.asg_policy_name
  policy_type            = "TargetTrackingScaling"
  depends_on             = [ aws_autoscaling_group.audio_auto_scaling_group ]
    
  target_tracking_configuration {
    target_value = 1
    customized_metric_specification {
      metrics {
        label = "Get the queue size (the number of messages waiting to be processed)"
        id    = "m1"
        metric_stat {
          metric {
            namespace   = "AWS/SQS"
            metric_name = "ApproximateNumberOfMessagesVisible"
            dimensions {
              name  = "QueueName"
              value = var.sqs_queue_name
            }
          }
          stat = "Sum"
        }
        return_data = false
      }
      metrics {
        label = "Get the group size (the number of InService instances)"
        id    = "m2"
        metric_stat {
          metric {
            namespace   = "AWS/AutoScaling"
            metric_name = "GroupInServiceInstances"
            dimensions {
              name  = "AutoScalingGroupName"
              value = var.asg_name
            }
          }
          stat = "Average"
        }
        return_data = false
      }
      metrics {
        label       = "Calculate the backlog per instance"
        id          = "e1"
        expression  = "IF(m2 == 0, m1, m1 / m2)"
        return_data = true
      }
    }
  }
}

更新后代码 #

resource "aws_launch_template" "audio_asg_launch_template" {
  name_prefix   = var.launch_config_name_prefix
  image_id      = data.aws_ami.audio_ami.id
  instance_type = "t2.micro"
}
resource "aws_autoscaling_group" "audio_auto_scaling_group" {
  availability_zones        = data.aws_availability_zones.available.names
  name                      = var.asg_name
  max_size                  = var.asg_max_size
  min_size                  = var.asg_min_size
  desired_capacity = 0
  health_check_grace_period = var.asg_health_check_grace_period
  health_check_type         = "EC2"
  force_delete              = true
  launch_template {
    id      = aws_launch_template.audio_asg_launch_template.id
    version = "$Latest"
  }
}
resource "aws_autoscaling_policy" "asg_audio_scale_up_policy" {
  autoscaling_group_name = aws_autoscaling_group.audio_auto_scaling_group.name
  name                   = var.asg_policy_name
  policy_type            = "TargetTrackingScaling"
  
  target_tracking_configuration {
    target_value = 1
    customized_metric_specification {
      metrics {
        label = "Get the queue size (the number of messages waiting to be processed)"
        id    = "m1"
        metric_stat {
          metric {
            namespace   = "AWS/SQS"
            metric_name = "ApproximateNumberOfMessagesVisible"
            dimensions {
              name  = "QueueName"
              value = aws_sqs_queue.file_checker_sqs_queue.name
            }
          }
          stat = "Sum"
        }
        return_data = false
      }
      metrics {
        label = "Get the group size (the number of InService instances)"
        id    = "m2"
        metric_stat {
          metric {
            namespace   = "AWS/AutoScaling"
            metric_name = "GroupInServiceInstances"
            dimensions {
              name  = "AutoScalingGroupName"
              value = aws_autoscaling_group.audio_auto_scaling_group.name
            }
          }
          stat = "Average"
        }
        return_data = false
      }
      metrics {
        label       = "Calculate the backlog per instance"
        id          = "e1"
        expression  = "IF(m2 == 0, m1, m1 / m2)"
        return_data = true
      }
    }
  }
}

可能的问题及排查方向 #

• 指标计算逻辑缺陷：当前表达式IF(m2 == 0, m1, m1 / m2)在实例数为0时，计算值等于队列消息数。而你的target_value=1，如果队列只有1条消息，计算值刚好等于目标值，不会触发扩容。建议调整表达式为IF(m2 == 0, 100, m1/m2)，这样只要队列有消息，计算值就会远大于目标值，触发ASG扩容到至少1个实例。
• CloudWatch指标延迟：CloudWatch指标默认的统计周期是5分钟，目标追踪策略需要等待指标聚合完成后才会执行扩缩容判断。发送消息后需等待至少5分钟再观察ASG状态。
• ASG权限缺失：检查ASG关联的IAM角色是否具备读取CloudWatch指标、执行EC2实例创建的权限。虽然Terraform默认会创建基础权限，但如果有自定义IAM策略，可能遗漏了cloudwatch:GetMetricStatistics、ec2:RunInstances等关键权限。
• 启动模板配置不全：你的启动模板未配置安全组、密钥对等必要参数，可能导致实例无法成功启动。即使ASG触发扩容，实例启动失败也不会显示在ASG的实例列表中。需补充安全组（允许必要的网络访问）和密钥对配置。
• 策略关联状态异常：登录AWS控制台，进入ASG详情页查看自动扩缩容策略是否已成功关联，以及策略是否有报错信息（比如指标无法读取）。
• 目标值设置不合理：target_value=1意味着每个实例只处理1条消息，这个阈值过于严格。可以根据实际处理能力调整为更高的值（比如5或10），避免频繁扩缩容，同时确保触发逻辑更符合预期。

内容的提问来源于stack exchange，提问作者dnd1993