(2.3.0 + Paper Plugin) Multiplayer: Chunks loading slowly but LODs loading fine?
was trying out the DH Support plugin in a Paper server last night, distant LODs actually work great but there seems to be some weird bottleneck around the vanilla chunks loading in.
If we walk towards a previously-unexplored direction we quickly hit a boundary where the vanilla chunks don't render (seems like the blocks are still there, we can highlight and break them, but they're all invisible). After about 5-10 mins of hanging around the area those chunks eventually pop in. Looks like this:
9 Replies
The DH lods themselves load in very quickly though. We had at max 3 people with DH 2.3 on the server at a time, saw this issue at lod distance 128 up to 1024
My client as an example (Fabric 1.21.1)
Server is Paper 1.21.1-127 with DH Support (0.6.1), one other plugin for player graves, and the world has a few light terrain gen datapacks like terralith and structory
If we walk towards a previously-unexplored directionupdate after more testing - this actually still happens with chunks we've already been to, just had it happen with our home base chunks (they were invisible for 5+ minutes)
oof, ouch that’s concerning
your server may be at an abysmal tps
i’m not very what what you can do threading-wise @Jckf is there any way to reduce the amount of threads DHS uses to generate lods?
Yes! :) There is a configuration option to set the number of worker threads. You can also reduce the number of parallel requests per client to lower server load
Default is 4 threads and 20 requests
ahh so it's potentially just a server load issue? the computer running the server is sitting around 70% cpu usage while we're running around testing this (ryzen 5 7600), but that pc is also running one of the clients so they might be fighting for resources
we tried cutting both the worker threads and parallel requests in half, still seeing the issue tho it might be slightly better
/tps shows
20 20 20from my tests tps stays normal, it's the chunk loading that dies under load
we had an old computer lying around that we moved the server to
it's working great now with default settings! probably was an issue with the server fighting for resources on that other computer that was also running a client 🙃
nvm... still happening with dedicated server 😢
try setting a low bandwidth limit on clients or a rate limit on server