In the function named Monitor._check_with_retry, if _check_once failed, the monitor will call self._topology.reset_pool(address) to reset the server's pool. But, it doesn't reset the monitor's pool. I think the monitor need reset both its own pool and server's pool.

I have a circumstance. Our System Administrator provide a host for us to get multiple mongos IP address, our application need to choose one mongos to connect. In order to auto reconnect to another mongos after a mongos goes down. I need to write a new Pool class, the Pool class need to choose a new mongos in the _init_ function and reset_pool function. But, after I finished the Pool class. I found that when a mongos goes down , the monitor reset the server's pool every time the monitor call self._check_with_retry. Because the monitor didn't reset its own pool.

I write a new Monitor class to reset its own pool, but the _check_with_retry function is not a public function. So, it's not guaranted that the function will not be removed or changed.

So, I think the code that reset the monitor's pool need add into the Monitor class.