import logging import time import docker import rov_server_config import utils from rov_server_update import server_health_check 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') def server_restart(slb_client, instance_id, image_name): try: logging.info(f"instance:{instance_id}") # 1. 获取ip ip_address = utils.get_ip_address(client=slb_client, instance_id=instance_id) logging.info(f"ip address:{instance_id}/{ip_address}") # 2. 摘流量 utils.set_instance_weight_process(client=slb_client, slb_id=rov_server_config.slb_id, instance_id_list=[instance_id], weight_list=[(0, 60)]) logging.info(f"set weight with 0 finished.") # 3. 重启容器 docker_client = docker.DockerClient(base_url=f'tcp://{ip_address}:2375', timeout=60) image_id = docker_client.containers.get(image_name) image_id.restart() logging.info(f"docker restart finished.") time.sleep(5) # 4. 探活 server_health_check(slb_client, instance_id) time.sleep(30) logging.info(f"health check finished.") # 5. 挂流量 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=[instance_id], weight_list=add_weight_list) logging.info(f"server restart finished, instance: {instance_id}/{ip_address}") except Exception as e: logging.error(f"server restart fail, instance: {instance_id}") logging.error(e) def main(): try: 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']) # 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. 逐台重启 image_name = 'rov-server' for i, instance_id in enumerate(online_instance_ids): server_restart(slb_client=slb_client, instance_id=instance_id, image_name=image_name) logging.info(f"重启进度: {i+1}/{online_instance_count}") logging.info(f"server restart finished!") except Exception as e: logging.error(e) if __name__ == '__main__': main()