Currently we have set up BungeeCord-local locks to see if the players are connected or are able to connect. The local state that tells if a player can join, called GlobalLock, is supposed to be synchronized amongst all instances of BungeeSemaphore by means of pub/sub communications through Redis.
This makes scaling in/out difficult, especially when BungeeCord processes are managed through orchestration tools such as k8s.
The aim of this issue is to consider moving all the states to a single Redis instance, and to see its consequences.
kory33
commented
3 years ago
Currently we have set up BungeeCord-local locks to see if the players are connected or are able to connect. The local state that tells if a player can join, called
GlobalLock, is supposed to be synchronized amongst all instances of BungeeSemaphore by means of pub/sub communications through Redis.This makes scaling in/out difficult, especially when BungeeCord processes are managed through orchestration tools such as k8s.
The aim of this issue is to consider moving all the states to a single Redis instance, and to see its consequences.