Monitoring Backend Instances

This section describes the status check of backend instances in detail.

The route_offline_thread in Web Server 7.0, performs a status check of backend instances. It looks up for all the alive backend instances by sending OPTIONS HTTP request. If the server responds, it is confirmed that the backend instance is alive. After the initial phase, the thread performs status check of offline backend instances every 60 seconds (time duration is not configurable). The route_offline_thread now tries to connect to offline instance and when the connection is successful, it sends OPTIONS HTTP request. If the instance responds, once it recovers from a hang, the route_offline_thread immediately identifies it as online. Thus, the thread does not impose any performance impact to the system and to the backend instances.

After a backend instance is identified as online, if it shuts down or crashes, the route_offline_thread will not be able to track. When such an instance is chosen by the Web Server reverse proxy parameter, for serving a request, it results in read or send failure, indicating that the connection to the instance is broken.

You can define a response timeout value using http-client-config ObjectType function in obj.conf file.

ObjectType fn="http-client-config" timeout="400"

The default timeout value is 300 seconds.

Once the response timeout value is defined, the reverse proxy parameter closes the connection to the backend instance if the connection hangs for more than 400 seconds and identifies it as offline.

When a reverse proxy parameter tries to establish a connection with a backend instance which is hanging or too busy, it waits for a connection response for a maximum period of five seconds and then identifies the instance as offline.