scripts/gateway/alb_gateway_update_cluster.py

import asyncio
import sys
import time
import requests
import alb_utils
import logging
import os
import docker
import gateway_config

from concurrent.futures import ThreadPoolExecutor

health_instances = []
ess_instances = []
remove_container_instances = []


def server_health_check(ecs_client, instance_id):
    """
    服务健康检查
    :param ecs_client: 客户端连接
    :param instance_id: instanceId
    :return:
    """
    global health_instances
    ip_address = alb_utils.get_ip_address(ecs_client=ecs_client, instance_id=instance_id)
    while True:
        health_check_url = f"http://{ip_address}:9000/healthcheck"
        try:
            http_code = requests.get(health_check_url).status_code
        except:
            logging.info(f"images is downloading ip：{ip_address}")
            http_code = 0

        if http_code == 200:
            health_instances.append((instance_id, ip_address))
            logging.info(f"health check success, instance: {instance_id}/{ip_address}")
            break
        else:
            time.sleep(10)


async def ess_instance(ecs_client, alb_client, ess_count, max_workers, version, port):
    """
    扩容机器并运行新服务
    :param ecs_client: 购买机器客户端连接
    :param alb_client: 修改负载均衡权限
    :param ess_count: 扩容数量
    :param max_workers: 线程数
    :param version: 版本标记
    :param port: 后端服务器使用的端口
    :return:
    """
    # 1. 购买机器并启动
    ess_instance_ids = alb_utils.create_multiple_instances(
        amount=ess_count,
        ecs_client=ecs_client,
        **gateway_config.instance_config_j,
    )
    time.sleep(60)

    # 2. 发送启动脚本到机器上
    alb_utils.send_file_to_ecs(ecs_client=ecs_client, instance_id_list=ess_instance_ids, **gateway_config.start_sh)
    logging.info(f"send start shell file finished, count: {len(ess_instance_ids)} instances: {ess_instance_ids}")
    # 3. 启动服务
    server_start_sh = os.path.join(gateway_config.start_sh['target_dir'], gateway_config.start_sh['name'])
    server_start_commend = f"sh {server_start_sh} {version}"
    alb_utils.run_command(ecs_client=ecs_client, instance_ids=ess_instance_ids, command=server_start_commend)
    logging.info(f"start server finish, count: {len(ess_instance_ids)} instances: {ess_instance_ids}")
    # 4. 异步探活
    global health_instances
    health_instances = []
    loop = asyncio.get_running_loop()
    executor = ThreadPoolExecutor(max_workers=max_workers)
    tasks = [
        loop.run_in_executor(executor, server_health_check, *args) for args in
        [(ecs_client, instance_id) for instance_id in ess_instance_ids]
    ]
    await asyncio.wait(tasks)
    logging.info(f"health instances count: {len(health_instances)}, instances: {health_instances}")
    # 5. 挂载流量
    if len(health_instances) == len(ess_instance_ids):
        # 所有机器探活成功
        time.sleep(20)
        health_instance_ids = [instance_id for instance_id, _ in health_instances]
        # for server_group_id in gateway_config.server_group_id_list:
            # for instance_id in health_instance_ids:
        alb_utils.add_servers_to_server_group(alb_client, gateway_config.server_group_id_list, health_instance_ids, weight=0, port=port)
        logging.info(f"Successfully added count: {len(health_instance_ids)} health_instance_ids {health_instance_ids} to server groups {gateway_config.server_group_id_list}.")

        time.sleep(20)
        logging.info(f"start update weight count: {len(health_instance_ids)} instances: {health_instance_ids} server groups: {gateway_config.server_group_id_list}.")
        add_weight_list = [(10, 5), (20, 5), (40, 5), (60, 5), (80, 5), (100, 5)]
        # add_weight_list = [(10, 10), (20, 10), (40, 10), (60, 10), (80, 10), (100, 10)]
        alb_utils.update_server_group_servers_attribute(alb_client,
                                                    server_group_id_list=gateway_config.server_group_id_list,
                                                    instance_id_list=health_instance_ids,
                                                    weight_list=add_weight_list,
                                                    port=port)
        global ess_instances
        ess_instances.extend(health_instance_ids)
        logging.info(f"ess count: {ess_count}, "
                     f"create count: {len(ess_instance_ids)}, "
                     f"finished count: {len(health_instance_ids)}")
    else:
        logging.info(f"ess count: {ess_count}, "
                     f"create count: {len(ess_instance_ids)}, "
                     f"health count: {len(health_instances)}")
        sys.exit()


def remove_container_image(ecs_client, instance_id, container_name_list):
    """
    移除旧容器并删除旧镜像
    :param ecs_client: 客户端连接
    :param instance_id: instanceId type-string
    :param container_name: 容器名称 type-string
    :return:
    """
    ip_address = alb_utils.get_ip_address(ecs_client=ecs_client, instance_id=instance_id)
    logging.info(f"服务器信息：{instance_id}/{ip_address}")
    client = docker.DockerClient(base_url=f'tcp://{ip_address}:2375', timeout=60)
    # 移除旧的容器
    container_remove_retry = 3
    i = 0
    while True:
        if i >= container_remove_retry:
            logging.error(f"容器不存在或者无法删除当前容器, instance = {instance_id}/{ip_address}")
            sys.exit()
        try:
            flag = False
            for container_name in container_name_list:
                try:
                    container_id = client.containers.get(container_name)
                    container_id.remove(force=True)
                    flag = True
                    break
                except:
                    continue
            if flag:
                break
        except Exception as e:
            i += 1

    # 删除旧镜像
    images_remove_retry = 3
    j = 0
    while True:
        if j >= images_remove_retry:
            logging.error(f"镜像不存在，无法获取到镜像ID, instance = {instance_id}/{ip_address}")
            sys.exit()
        try:
            images = client.images.list()
            for image in images:
                client.images.remove(force=True, image=image.tags[0])
                time.sleep(2)
            global remove_container_instances
            remove_container_instances.append(instance_id)
            break
        except Exception as e:
            i += 1


async def update_instance(ecs_client, alb_client, instance_ids, max_workers, version, port):
    """
    线上机器更新
    :param ecs_client:
    :param alb_client: alb客户端连接
    :param instance_ids: instanceId type-list
    :param max_workers:
    :param version: 版本标记
    :param port: 后端服务器使用的端口
    :return:
    """
    sub_index = len(instance_ids)//2
    instance_ids_group = [instance_ids[:sub_index], instance_ids[sub_index:]]
    update_finished_count = 0
    for instance_id_list in instance_ids_group:
        logging.info(f"update instances: {instance_id_list}")
        # 1. 摘流量
        alb_utils.update_server_group_servers_attribute(alb_client=alb_client,
                                                    server_group_id_list=gateway_config.server_group_id_list,
                                                    instance_id_list=instance_id_list,
                                                    weight_list=[(0, 20)],
                                                    port=port)
        logging.info(f"set weight with 0 finished, count: {len(instance_id_list)} instances: {instance_id_list}")
        # 2. 异步移除旧容器并删除旧镜像
        global remove_container_instances
        remove_container_instances = []
        # container_name = 'longvideoapi'
        container_name_list = ['piaoquan-gateway', 'gateway']
        loop = asyncio.get_running_loop()
        executor = ThreadPoolExecutor(max_workers=max_workers)
        tasks = [
            loop.run_in_executor(executor, remove_container_image, *args) for args in
            [(ecs_client, instance_id, container_name_list) for instance_id in instance_id_list]
        ]
        await asyncio.wait(tasks)
        logging.info(f"remove container & images finished, instances: {remove_container_instances},"
                     f" count: {len(remove_container_instances)}")
        if len(remove_container_instances) < len(instance_id_list):
            logging.error(f"remove container image failed| "
                          f"request count: {len(instance_id_list)}, removed count: {len(remove_container_instances)}")
            sys.exit()
        # 3. 发送启动脚本到机器上
        alb_utils.send_file_to_ecs(ecs_client=ecs_client, instance_id_list=instance_id_list, **gateway_config.start_sh)
        logging.info(f"send start shell file finished, instances: {instance_id_list}, count: {len(instance_id_list)}")
        # 4. 启动服务
        server_start_sh = os.path.join(gateway_config.start_sh['target_dir'], gateway_config.start_sh['name'])
        server_start_commend = f"sh {server_start_sh} {version}"
        alb_utils.run_command(ecs_client=ecs_client, instance_ids=instance_id_list, command=server_start_commend)
        # 5. 异步探活
        global health_instances
        health_instances = []
        loop = asyncio.get_running_loop()
        executor = ThreadPoolExecutor(max_workers=max_workers)
        tasks = [
            loop.run_in_executor(executor, server_health_check, *args) for args in
            [(ecs_client, instance_id) for instance_id in instance_id_list]
        ]
        await asyncio.wait(tasks)
        logging.info(f"health instances: {health_instances}, count: {len(health_instances)}")
        # 6. 挂载流量
        if len(health_instances) == len(instance_id_list):
            # 所有机器探活成功
            time.sleep(10)
            health_instance_ids = [instance_id for instance_id, _ in health_instances]

            add_weight_list = [(10, 5), (20, 5), (40, 5), (60, 5), (80, 5), (100, 5)]
            # add_weight_list = [(10, 10), (20, 10), (40, 10), (60, 10), (80, 10), (100, 10)]
            alb_utils.update_server_group_servers_attribute(alb_client,
                                                        server_group_id_list=gateway_config.server_group_id_list,
                                                        instance_id_list=health_instance_ids,
                                                        weight_list=add_weight_list,
                                                        port=port)
            logging.info(f"finished instances: {health_instances}, count: {len(health_instances)}")
            update_finished_count += len(health_instances)
            logging.info(f"update finished: {update_finished_count}/{len(instance_ids)}")
        else:
            logging.info(f"health instances: {health_instances}, count: {len(health_instances)}")
            sys.exit()


def remove_instances(ecs_client, alb_client, instance_ids, port):
    """
    停止并释放机器
    :param ecs_client:
    :param alb_client:
    :param instance_ids: instanceId type-list
    :param port: 后端服务器使用的端口
    :return: None
    """
    # 1. 摘流量
    weight_list = [(0, 20)]  # 设置权重为0，等待20秒
    try:
        alb_utils.update_server_group_servers_attribute(alb_client=alb_client,
                                                    server_group_id_list=gateway_config.server_group_id_list,
                                                    instance_id_list=instance_ids,
                                                    weight_list=weight_list,
                                                    port=port)
    except Exception as e:
        logging.error(f"Failed to set instance weight: {e}")
        sys.exit()

    time.sleep(10)
    # 2.移除alb
    # for server_group_id in gateway_config.server_group_id_list:
    try:
        alb_utils.remove_servers_from_server_group(alb_client=alb_client, server_group_id_list=gateway_config.server_group_id_list,
                                               instance_ids=instance_ids, port=port)
        logging.info(f"Successfully removed instances: {instance_ids} from server group {gateway_config.server_group_id_list}.")
    except Exception as e:
        logging.error(f"Failed to remove instances: {instance_ids} from server group {gateway_config.server_group_id_list}: {e}")

    logging.info(f"Remove from ALB finished, instances: {instance_ids}")

    # 3. 停止机器
    alb_utils.stop_instances(ecs_client=ecs_client, instance_ids=instance_ids)
    # stop_response = alb_utils.stop_instances(ecs_client=ecs_client, instance_ids=instance_ids)
    # if stop_response.get('Code') is None:
    #     logging.info(f"Instances stop finished, instances: {instance_ids}")
    # else:
    #     logging.error(f"Failed to stop instances: {stop_response}")
    #     sys.exit()

    # 4. 判断机器运行状态是否为Stopped
    # while True:
    stopped_instances = alb_utils.get_instances_status(ecs_client=ecs_client, instance_ids=instance_ids)
        # response = alb_utils.get_instances_status(ecs_client=ecs_client, instance_ids=instance_ids)
        # if response.get('Code') is None:
        #     instances_list = response.get('InstanceStatuses').get('InstanceStatus')
        #     stopped_instances = [instance.get('InstanceId') for instance in instances_list if
        #                          instance.get('Status') == 'Stopped']
        #     if len(stopped_instances) == len(instance_ids):
        #         logging.info(f"Instances stopped status set success, instances: {stopped_instances}")
        #         break
        #     else:
        #         logging.info(f"Stopped instances count = {len(stopped_instances)}, instances: {stopped_instances}")
        #         time.sleep(5)
        # else:
        #     logging.error(response)
        #     sys.exit()
    # 5. 释放机器
    alb_utils.release_instances(ecs_client=ecs_client, instance_ids=stopped_instances)
    # release_response = alb_utils.release_instances(ecs_client=ecs_client, instance_ids=stopped_instances)
    # if release_response.get('Code') is None:
    #     logging.info(f"Release instances finished, instances: {stopped_instances}")
    # else:
    #     logging.error(f"Release instances fail!!!")
    #     sys.exit()


def main():
    try:
        version = sys.argv[1]
        alb_client = alb_utils.connect_alb_client(
            access_key_id=gateway_config.alb_client_params['access_key_id'],
            access_key_secret=gateway_config.alb_client_params['access_key_secret'],
            endpoint=gateway_config.alb_client_params['endpoint']
        )
        ecs_client = alb_utils.connect_client(access_key_id=gateway_config.ecs_client_params['access_key_id'],
                                          access_key_secret=gateway_config.ecs_client_params['access_key_secret'],
                                          region_id=gateway_config.ecs_client_params['region_id'])

        # 1. 获取alb下所有机器
        online_instance_ids = alb_utils.list_server_group_servers(alb_client=alb_client, server_group_id=gateway_config.server_group_id_list[0])
        online_instance_count = len(online_instance_ids)
        logging.info(f"online instance count: {online_instance_count} instance_ids: {online_instance_ids}")

        # 2. 扩容机器并启动新服务 扩容数量：线上机器数量//2
        logging.info(f"ess instances start ...")
        ess_instance_count = online_instance_count // 2
        logging.info(f"ess instance count: {ess_instance_count}")
        asyncio.run(ess_instance(ecs_client=ecs_client, alb_client=alb_client,
                                 ess_count=ess_instance_count, max_workers=2, version=version, port=gateway_config.port))
        logging.info(f"ess instances end!")

        # 3. 原有机器进行更新
        logging.info(f"update online instances start ...")
        asyncio.run(update_instance(ecs_client=ecs_client, alb_client=alb_client,
                                    instance_ids=online_instance_ids, max_workers=8, version=version, port=gateway_config.port))
        logging.info(f"update online instances end!")

        # 4. 停止并释放扩容机器
        logging.info(f"stop & release instances start ...")
        remove_instances(ecs_client=ecs_client, alb_client=alb_client, instance_ids=ess_instances, port=gateway_config.port)
        logging.info(f"stop & release instances end!")
    except Exception as e:
        logging.error(e)
        sys.exit()


if __name__ == '__main__':
    main()