Help me understand the connection pool

[Kampfkarren]

Hi, I have on a Docker Swarm a node that fires out GETs to a node on the swarm, like http://my_service.local. There's a lot of nodes in the swarm that have that host, and the intent was that firing out ::get would pick a different one every time. Docker uses a round-robin DNS to make this happen.

However, when I reused a client, it would use the same node over and over again for a repeated amount of time. This stumped me for several days, and it was solved once I made a fresh client every time.

I would like to know this:

1. What exactly was happening there?
2. Is there anything I can configure on Client to make that not happen? It is important that different nodes are picked every time.
3. If I can, is that even better than making a new Client every time?

Much appreciated!