Skip to content

Latest commit

 

History

History
36 lines (27 loc) · 1.8 KB

BRAINDUMP.md

File metadata and controls

36 lines (27 loc) · 1.8 KB

Unordered list of things to keep in mind

Open issues

What if BungeeCord goes down?

  • We lose all player <-> Partition <-> Server instance information
  • Do we want to persist this?

Players can still do /server

  • Disable this. Players don't get to choose their node

Every world seems to load the spawn chunk, no matter if keep-spawn-loaded is true or not

  • MultiverseCore plugin should be disabled as this loads spawn irrespective of the flag
  • When the first player connects World.getSpawn() is always retrieved, even if we don't need to select a new spawnspot ( See EntityPlayer.a(WorldServer) )

Teleporting should be coordinated by the proxy

  • Teleporting a player to a target location should never be decided by a node rather by the proxy
  • When teleporting, a message should be sent to the proxy requesting a teleporting
  • The proxy figures out which node the player needs to be sent to
  • It responds with either
    1. You can teleport because the coordinate is local to your node
    2. You can teleport AFTER setting the target region files read-only
  • Then the normal migration mechanism takes over

The proxy could keep track of loaded region files

  • If we have the nodes communicate their loaded/unloaded region files back to the proxy it could make sure we never load the same region files on two nodes.

There are a number of uses of worldserver.getSpawn()

  • Identify them, and neutralize or disable
  • We cannot have two servers loading the same chunks

Support for /locate requires us to increase MERGE_DISTANCE