server-deploy/rov_server_grayscale_update_with_num.py

import asyncio
import sys
import time
import requests
import utils
import logging
import os
import docker
import rov_server_config

from concurrent.futures import ThreadPoolExecutor

logging.basicConfig(level=logging.INFO,
                    format='%(asctime)s %(filename)s[line:%(lineno)d] %(levelname)s %(message)s',
                    datefmt='%a, %d %b %Y %H:%M:%S')

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


def server_health_check(client, instance_id):
    """
    服务健康检查
    :param client: 客户端连接
    :param instance_id: instanceId
    :return:
    """
    global health_instances
    ip_address = utils.get_ip_address(client=client, instance_id=instance_id)
    while True:
        health_check_url = f"http://{ip_address}:5001/healthcheck"
        try:
            http_code = requests.get(health_check_url).status_code
        except:
            logging.info("images is downloading")
            http_code = 0

        if http_code == 200:
            health_instances.append((instance_id, ip_address))
            break
        else:
            time.sleep(10)


async def ess_instance(create_client, slb_client, ess_count, max_workers, version):
    """
    扩容机器并运行新服务
    :param create_client: 购买机器客户端连接
    :param slb_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
    :return:
    """
    # 1. 购买机器并启动
    ess_instance_ids = utils.create_multiple_instances(
        amount=ess_count,
        client=create_client,
        **rov_server_config.instance_config,
    )
    time.sleep(60)

    # 2. 发送启动脚本到机器上
    utils.send_file_to_ecs(client=create_client, instance_id_list=ess_instance_ids, **rov_server_config.start_sh)
    logging.info(f"send start shell file finished, instances: {ess_instance_ids}")
    # 3. 启动服务
    server_start_sh = os.path.join(rov_server_config.start_sh['target_dir'], rov_server_config.start_sh['name'])
    server_start_commend = f"sh {server_start_sh} {version}"
    utils.run_command(client=create_client, instance_ids=ess_instance_ids, command=server_start_commend)
    # 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
        [(slb_client, instance_id) for instance_id in ess_instance_ids]
    ]
    await asyncio.wait(tasks)
    logging.info(f"health instances count: {len(health_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(60)
        utils.add_backend_servers(client=slb_client, slb_id=rov_server_config.slb_id, instances=health_instances)
        health_instance_ids = [instance_id for instance_id, _ in health_instances]
        add_weight_list = [(10, 30), (20, 20), (40, 10), (60, 10), (80, 10), (100, 10)]
        utils.set_instance_weight_process(client=slb_client,
                                          slb_id=rov_server_config.slb_id,
                                          instance_id_list=health_instance_ids,
                                          weight_list=add_weight_list)
        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(client, instance_id, container_name):
    """
    移除旧容器并删除旧镜像
    :param client: 客户端连接
    :param instance_id: instanceId type-string
    :param container_name: 容器名称 type-string
    :return:
    """
    ip_address = utils.get_ip_address(client=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:
            container_id = client.containers.get(container_name)
            container_id.remove(force=True)
            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(create_client, slb_client, instance_ids, max_workers, version):
    """
    线上机器更新
    :param create_client:
    :param slb_client: slb客户端连接
    :param instance_ids: instanceId type-list
    :param max_workers:
    :param version: 版本标记
    :return:
    """
    instance_ids_group = [instance_ids]
    update_finished_count = 0
    for instance_id_list in instance_ids_group:
        logging.info(f"update instances: {instance_id_list}")
        # 1. 摘流量
        utils.set_instance_weight_process(client=slb_client,
                                          slb_id=rov_server_config.slb_id,
                                          instance_id_list=instance_id_list,
                                          weight_list=[(0, 60)])
        logging.info(f"set weight with 0 finished, 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 = 'rov-server'
        loop = asyncio.get_running_loop()
        executor = ThreadPoolExecutor(max_workers=max_workers)
        tasks = [
            loop.run_in_executor(executor, remove_container_image, *args) for args in
            [(slb_client, instance_id, container_name) 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
        utils.send_file_to_ecs(client=create_client, instance_id_list=instance_id_list, **rov_server_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(rov_server_config.start_sh['target_dir'], rov_server_config.start_sh['name'])
        server_start_commend = f"sh {server_start_sh} {version}"
        utils.run_command(client=create_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
            [(slb_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(60)
            utils.add_backend_servers(client=slb_client, slb_id=rov_server_config.slb_id, instances=health_instances)
            health_instance_ids = [instance_id for instance_id, _ in health_instances]
            add_weight_list = [(10, 30), (20, 20), (40, 10), (60, 10), (80, 10), (100, 10)]
            utils.set_instance_weight_process(client=slb_client,
                                              slb_id=rov_server_config.slb_id,
                                              instance_id_list=health_instance_ids,
                                              weight_list=add_weight_list)
            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(create_client, slb_client, instance_ids):
    """
    停止并释放机器
    :param create_client:
    :param slb_client:
    :param instance_ids: instanceId type-list
    :return: None
    """
    # 1. 摘流量
    utils.set_instance_weight_process(client=slb_client,
                                      slb_id=rov_server_config.slb_id,
                                      instance_id_list=instance_ids,
                                      weight_list=[(0, 60)])
    logging.info(f"set weight = 0 finished, instances: {instance_ids}")
    time.sleep(10)
    # 2.移除slb
    utils.remove_backend_servers(client=slb_client, slb_id=rov_server_config.slb_id, instances=instance_ids)
    # 3. 停止机器
    utils.stop_instances(client=create_client, instance_ids=instance_ids)
    logging.info(f"instances stop finished, instances: {instance_ids}")
    # 4. 判断机器运行状态是否为Stopped
    while True:
        response = utils.get_instances_status(client=create_client, instance_ids=instance_ids)
        if response.get('Code') is None:
            instances_list = response.get('InstanceStatuses').get('InstanceStatus')
            # logging.info(instances_list)
            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(10)
        else:
            logging.error(response)
            sys.exit()
    # 5. 释放机器
    response = utils.release_instances(client=create_client, instance_ids=stopped_instances)
    if 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]
        num = int(sys.argv[2])
        slb_client = utils.connect_client(access_key_id=rov_server_config.slb_client_params['access_key_id'],
                                          access_key_secret=rov_server_config.slb_client_params['access_key_secret'],
                                          region_id=rov_server_config.slb_client_params['region_id'])
        create_client = utils.connect_client(access_key_id=rov_server_config.create_client_params['access_key_id'],
                                             access_key_secret=rov_server_config.create_client_params['access_key_secret'],
                                             region_id=rov_server_config.create_client_params['region_id'])

        # 1. 获取slb下所有机器
        online_instance_ids = utils.get_instance_ids(client=slb_client, slb_id=rov_server_config.slb_id)
        online_instance_count = len(online_instance_ids)
        logging.info(f"online instance count: {online_instance_count}.")
        logging.info(f"online instance ids: {online_instance_ids}")
        # 2. 对其中n台机器进行更新
        update_instance_ids = online_instance_ids[:num]

        # 3. 扩容机器并启动新服务 扩容数量：灰度发布机器数量
        logging.info(f"ess instances start ...")
        logging.info(f"ess instance count: {num}")
        asyncio.run(ess_instance(create_client=create_client, slb_client=slb_client,
                                 ess_count=num, max_workers=2, version=version))
        logging.info(f"ess instances end!")

        # 4. 原有机器进行更新
        logging.info(f"update online instances start ...")
        asyncio.run(update_instance(create_client=create_client, slb_client=slb_client,
                                    instance_ids=update_instance_ids, max_workers=2, version=version))
        logging.info(f"update online instances end!")

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


if __name__ == '__main__':
    main()