add longvideo_remove_failed_ecs.py & rov_sever_remove_failed_ecs.py

longvideo_remove_failed_ecs.py

@@ -0,0 +1,57 @@
+import requests
+import utils
+import longvideo_config
+import logging
+
+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 get_failed_sever_instances(slb_client, instance_ids):
+    """获取探活失败的机器"""
+    failed_instances = []
+    for instance_id in instance_ids:
+        ip_address = utils.get_ip_address(client=slb_client, instance_id=instance_id)
+        health_check_url = f"http://{ip_address}:8080/longvideoapi/test"
+        try:
+            http_code = requests.get(health_check_url).status_code
+        except:
+            failed_instances.append(instance_id)
+            logging.info(f"failed instances: {instance_id}/{ip_address}")
+            continue
+
+        if http_code == 200:
+            continue
+        else:
+            failed_instances.append(instance_id)
+            logging.info(f"failed instances: {instance_id}/{ip_address}")
+    return failed_instances
+
+
+def main():
+    """移除服务挂掉的机器"""
+    slb_client = utils.connect_client(access_key_id=longvideo_config.slb_client_params['access_key_id'],
+                                      access_key_secret=longvideo_config.slb_client_params['access_key_secret'],
+                                      region_id=longvideo_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=longvideo_config.slb_id_list[0])
+    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. 获取探活失败的机器
+    failed_instances = get_failed_sever_instances(slb_client=slb_client, instance_ids=online_instance_ids)
+    logging.info(f"failed instances count: {len(failed_instances)}")
+    logging.info(f"failed instance ids: {failed_instances}")
+
+    # 3. 摘流量
+    utils.set_instance_weight_process_with_slbs(client=slb_client,
+                                                slb_id_list=longvideo_config.slb_id_list,
+                                                instance_id_list=failed_instances,
+                                                weight_list=[(0, 15)])
+    logging.info(f"set weight with 0 finished.")
+
+
+if __name__ == '__main__':
+    main()

rov_sever_remove_failed_ecs.py

@@ -0,0 +1,56 @@
+import requests
+import utils
+import rov_server_config
+import logging
+
+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 get_failed_sever_instances(slb_client, instance_ids):
+    """获取探活失败的机器"""
+    failed_instances = []
+    for instance_id in instance_ids:
+        ip_address = utils.get_ip_address(client=slb_client, instance_id=instance_id)
+        health_check_url = f"http://{ip_address}:5001/healthcheck"
+        try:
+            http_code = requests.get(health_check_url).status_code
+        except:
+            failed_instances.append(instance_id)
+            logging.info(f"failed instances: {instance_id}/{ip_address}")
+            continue
+
+        if http_code == 200:
+            continue
+        else:
+            failed_instances.append(instance_id)
+            logging.info(f"failed instances: {instance_id}/{ip_address}")
+    return failed_instances
+
+
+def main():
+    """移除服务挂掉的机器"""
+    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. 获取探活失败的机器
+    failed_instances = get_failed_sever_instances(slb_client=slb_client, instance_ids=online_instance_ids)
+    logging.info(f"failed instances count: {len(failed_instances)}")
+    logging.info(f"failed instance ids: {failed_instances}")
+
+    # 3. 摘流量
+    utils.set_instance_weight_process(client=slb_client,
+                                      slb_id=rov_server_config.slb_id,
+                                      instance_id_list=failed_instances,
+                                      weight_list=[(0, 60)])
+    logging.info(f"set weight with 0 finished.")
+
+
+if __name__ == '__main__':
+    main()