How to use the server with replicas and load balancer?

[robertofalk]

Description

Because of the in memory session id, when there are more than one replica (and having a load balancer in front of the server), requests from the same client might be handled by different replicas, which doesn't work, since not all clients are aware of sessions all the time.

I saw that K8s might manage sessions using session affinity but I was wondering if on the library level, someone is using something like redis to sync the session between replicas?

Version Information

FastMCP version:                                                       2.3.0
MCP version:                                                           1.8.0
Python version:                                                       3.13.2
Platform:                      Linux-6.10.14-linuxkit-aarch64-with-glibc2.36
FastMCP root path: /workspaces/mcp-server/.venv/lib/python3.13/site-packages

[zzstoatzz]

let's move this to a discussion for now until we get some more opinions expressed on what should change in fastmcp