Terraform EMR模块未声明core_instance_group却在计划中出现致集群重复重建

问题现象 #

• 使用terraform-aws-modules/emr/aws 1.0.0版本部署EMR集群，已配置master_instance_fleet和core_instance_fleet
• 执行terraform plan时，计划输出中出现未声明的core_instance_group和master_instance_group，标记为待创建（+）
• Terraform状态里这两个字段值为null，导致哪怕集群无任何修改，每次执行plan都会提示需要重建集群

原因分析 #

这个版本的EMR模块在处理Instance Fleet配置时，没有自动禁用Instance Group相关参数。Terraform会将未设置的Instance Group参数视为null，但AWS EMR的API逻辑是：集群采用Fleet模式时，不应存在Instance Group资源。这种状态不一致会让Terraform误判为需要修改集群，进而触发重建操作。

解决方法 #

在模块配置中显式将master_instance_group和core_instance_group设为null，强制Terraform忽略这两个资源块。修改后的完整配置如下：

module "emr_trino_cluster" {
    source                  = "terraform-aws-modules/emr/aws"
    version                 = "1.0.0"
    name                    = "${local.resources_prefix}-trino"
    log_uri                 = "s3n://${s3_log}/logs/elasticmapreduce/"
    release_label           = var.trino_emr_version
    visible_to_all_users    = true
    applications            = ["Trino", "Hive", "Hue"]
    create_service_iam_role = false
    service_iam_role_arn    = aws_iam_role.emr_default_role.arn
    scale_down_behavior     = "TERMINATE_AT_TASK_COMPLETION"
    ebs_root_volume_size    = 20
    step_concurrency_level  = 1
    termination_protection  = false
    vpc_id                  = var.vpc_id
    
    master_instance_fleet = {
        name                      = "cfnMaster"
        target_on_demand_capacity = 1

        instance_type_configs = [
            {
                instance_type     = "r6g.xlarge"
                weighted_capacity = 1
                ebs_config = [{
                    iops                 = 0
                    size                 = 64
                    type                 = "gp3"
                    volumes_per_instance = 1
                }]
            }
        ]
    }

    core_instance_fleet = {
        name                      = "coreNinja"
        target_on_demand_capacity = 1
        instance_type_configs = [
            {
                bid_price_as_percentage_of_on_demand_price = 60
                ebs_config = [{
                    iops                 = 0
                    size                 = 64
                    type                 = "gp3"
                    volumes_per_instance = 1
                }]
                instance_type     = "r6g.xlarge"
                weighted_capacity = 1
            }
        ]
    }

    # 新增这两行，显式禁用Instance Group
    master_instance_group = null
    core_instance_group = null

    create_iam_instance_profile = false
    ec2_attributes = {
        instance_profile                  = aws_iam_instance_profile.ec2_attributes_profile.name
        key_name                          = aws_key_pair.ec2_key_name.key_name
        subnet_id                         = var.private_subnet_ids[0]
        additional_master_security_groups = aws_security_group.trino_additional_master_security_groups.id
        additional_slave_security_groups  = aws_security_group.trino_slave_security_groups.id
    }

    tags = local.tags
}

验证操作 #

1. 重新执行terraform plan，确认core_instance_group和master_instance_group不再出现在变更列表中
2. 执行terraform apply更新状态，后续再执行terraform plan将不会再出现重建集群的提示

额外说明 #

• 该模块的2.0.0及以上版本已修复此问题，模块会自动根据是否配置Fleet来处理Instance Group参数，若能升级模块版本，也可直接通过升级解决问题
• 若暂时无法升级，显式设置null是最直接有效的解决方案

内容的提问来源于stack exchange，提问作者Marksman