Auroraのクラスタのスナップショットを既存のAurora環境に復元する

restore_snapshot.rb

require 'logger'
require 'aws-sdk'

LOGGER = Logger.new(STDOUT)

def current_timestamp
  Time.new.strftime("%Y-%m-%d-%H-%M-%S")
end

# 最大一時間待つ
def wait_for(wait_time: 60, retry_max: 60)
  i = 0
  loop {
    result = yield
    break result if result
    i += 1
    break false if i >= retry_max
    sleep wait_time
  }
end

def get_instance_class(rds, identifier)
  rds.describe_db_instances(db_instance_identifier: identifier).db_instances.first.db_instance_class
end

def main
  src_db_cluster_identifier = "ams-api-alpha"
  dest_db_cluster_identifier = "ams-api-stg"

  rds = Aws::RDS::Client.new(region: 'ap-northeast-1')

  LOGGER.info("find latest snapshot")
  #対象クラスタの最新のスナップショット取得
  src_snapshot = rds.describe_db_cluster_snapshots(db_cluster_identifier: src_db_cluster_identifier).db_cluster_snapshots.first

  if src_snapshot
    # 置き換え対象のクラスタ取得
    dest_cluster = rds.describe_db_clusters(db_cluster_identifier: dest_db_cluster_identifier).db_clusters.first
    vpc_security_groups = dest_cluster.vpc_security_groups
    engine = dest_cluster.engine
    instance_classes = {}

    # 置き換え対象のクラスタ内のインスタンスをすべて削除
    dest_cluster.db_cluster_members.each do |member|
      instance_classes[member.db_instance_identifier] = get_instance_class(rds, member.db_instance_identifier)
      LOGGER.info("delete dest instances: #{member.db_instance_identifier}")
      rds.delete_db_instance(
        db_instance_identifier: member.db_instance_identifier
      )
    end

    result = wait_for {
      LOGGER.info("check instances ...")
      cluster = rds.describe_db_clusters(db_cluster_identifier: dest_db_cluster_identifier).db_clusters.first
      count = cluster.db_cluster_members.count
      count == 0
    }
    # かなり時間かかるので待ち足りない可能性もある
    raise "インスタンスの削除に失敗しました" unless result
    LOGGER.info("wait delete instances: end")

    # 置き換え対象のクラスタを削除
    LOGGER.info("delete dest cluster")
    rds.delete_db_cluster(
      db_cluster_identifier: dest_db_cluster_identifier,
      skip_final_snapshot: true
    )

    # 削除されるのを待つ
    LOGGER.info("delete cluster: start")
    result = wait_for {
      begin
        LOGGER.info("check delete cluster ...")
        rds.describe_db_clusters(db_cluster_identifier: dest_db_cluster_identifier)
        false
      rescue Aws::RDS::Errors::DBClusterNotFoundFault
        # 見つからない場合例外になるのでそれで削除確認
        true
      end
    }
    raise "削除に失敗しました" unless result
    LOGGER.info("delete cluster: end")

    # スナップショットから復元
    rds.restore_db_cluster_from_snapshot(
      db_cluster_identifier: dest_db_cluster_identifier,
      engine: engine,
      snapshot_identifier: src_snapshot.db_cluster_snapshot_identifier,
      vpc_security_group_ids: vpc_security_groups.map(&:vpc_security_group_id),
      db_subnet_group_name: dest_cluster.db_subnet_group
    )

    # 復元が完了するまで待つ
    LOGGER.info("restore cluster: start")
    result = wait_for {
      begin
        LOGGER.info("check restore cluster ...")
        cluster = rds.describe_db_clusters(db_cluster_identifier: dest_db_cluster_identifier)
        cluster.db_clusters.first.status == "available"
      rescue Aws::RDS::Errors::DBClusterNotFoundFault
        false
      end
    }
    LOGGER.info("restore cluster: end")

    # クラスタパラメータグループ変更
    rds.modify_db_cluster(
      db_cluster_identifier: dest_db_cluster_identifier,
      db_cluster_parameter_group_name:  dest_cluster.db_cluster_parameter_group
    )

    # 復元後インスタンス起動
    LOGGER.info("start cluster members(instances): start")
    dest_cluster.db_cluster_members.each do |member|
      rds.create_db_instance(
        db_instance_identifier: member.db_instance_identifier,
        db_instance_class: instance_classes[member.db_instance_identifier],
        engine: dest_cluster.engine,
        db_cluster_identifier: dest_cluster.db_cluster_identifier
      )
    end
    result = wait_for {
      cluster = rds.describe_db_clusters(db_cluster_identifier: dest_db_cluster_identifier).db_clusters.first
      count = cluster.db_cluster_members.count
      puts count
      count == dest_cluster.db_cluster_members.size
    }
    LOGGER.info("check cluster instances in aws console")
    LOGGER.info("restore done!")
  end
end

main