Server clustering

[Krzmbrzl]

Context

Servers that have a lot of concurrent users and/or are critical and should therefore be made redundant in terms of a single server process potentially failing for one reason or another.

Description

There should be a possibility to have multiple Mumble server process instances to form a server cluster that to the outside appears as a single logical server. Internally, the different server instances (which may also sit on entirely different hardware) should to some sort of automatic load balancing and potentially also optimizing which users are handled on which server in order to minimize that user's latency.

Mumble component

Server

OS-specific?

No

Additional information

No response

[andreasstieger]

You should probably look into transparent failover (faster recover or high availability) and sharding (per room) first.