#folia-help
1 messages · Page 40 of 1
We mention this a lot (and often) in here.
We even have a little infographic about it.
I know, players rtp on an 150k x 150k radius so I thought that would be the case for us
Should we not be using folia then?
If your players are too close together such that you're ending up with one large region, then you might want to use Paper instead.
Do you think paper can be effective at handling 300 players? (Vanilla)
Vanilla as in zero plugins?
Well we have viaversion and grimac
Send this?
the rtp is just spawn radius
I cant access a report right now but I can send one later
but the server runs fine with any less than like 18k chunks in a region
Why cant they split? Wouldnt it be possible to implement that?
i guess this wasnt donutsmp (isnt his grafana filled with goliaths?)
For 5k players u need more than making a community and techinal thingies, more data you need
holy
bros shocked 😄
Well, I am currently working on something similar. but I know I wont get that much players
1000 on its own is impressive
sharding is just a alternative way to do that
i mean, handling players
but no one cant get 1k players in single region
ofc not 😂
My server had problems running paper with 80 people
you should change things in game
I guess its because of the plugins
Plugins, hardware, network
poopy plugins, currently making my own
any of them may cause
yeah I am changing things
learning java is most important. At least folia with "a lot" players
I mean, medium knowledge is fine I guess
everyone can handle up to 150 players in single paper instance (overall)
if you want more, you need more knowledge
Directly proportional lmao
I guess it's a EU server?
🙂↔️
damn
Hello, its possible update folia 1.21.8 to folia 1.21.11 or i will have problems?
I have all plugins custom
1.21.11 is in alpha stage. it is possible but its not guaranteed to be stable for production
from what I know
What could happen?
it may corrupt the world file
O
Like in month in would be stable?
there is no folia 1.21.10
U
it got jumped from .8 to .11
why not
they started with the 26.1 snapshots
because there has not been enough time to find (and potentially fix) all the issues it could have
right
And make support folia snapchot 26.1 private its a mess?
(granted the fact that they didn't have to hotfix anything is a good sign)
Yrah
Yeah*
For now any errors then
Only waiting for bugs no?
Or why its isnt stable yet?
because there has not been enough time to find (and potentially fix) all the issues it could have
Okay
🥰
A yeah, i have a problem, with folia the rtp when tp, it stays you in the void like 8s
In coords very gigants
Thanks ❤️
Like 18929020
are you the dev of the plugin? otherwise it might be good for them to resolve this as they actually know what they are doing
For rtp its the only plugin no custom
Its betterrtp
I havent create a rtp plugin because i think it was the folia loading slow
(I have a good machine)
best to report this to them if it only happens when randomly teleporting and not normally, they can then provide the necessary info here if there really is an issue in folia itself
Better i will create my custom plugin
For rtp
For avoid this bug, first i will have to load the de region right?
And then the tp
that's more a #folia-dev question if you mean writing custom code but just use the teleportAsync API, it will do all this for you
Ty
- scheduling
Quick question, does Spark work with folia?
Yes, there's a Folia-specific branch/build for Spark.
I agree with this question. On my server, the vault block is absolutely not activated for completely new players - and it cannot be opened. Moreover, if you install the vault block from the creative, everything is fine with it. It seems that vault blocks break down over time
(Please don't post the same message to multiple channels)
Hello, has anyone run Folia under heavy load on a two-socket server or is there any info/best practices I can read? Currently I have one-socket server with 48 physical cores using ZGC in generational mode. Will there be performance drops compares to my current setup if I run on 24+24 (two processors with 24 cores each) server and assign, for example, entire 1st socket and part of 2nd (24+16) via cpuset in docker-compose and run it via numactl --interleave=all?
Folia doesn’t officially support NUMA, so we don’t really recommend it.
Thank you, even if I use a large two-socket server, for example, 48+48, and want to place two Folia servers on it and use cpuset and numactl --cpunodebind with --membind (pin instance to a specific socket and RAM from this socket) for each instance, is there still potential issues?
one Folia instance running on one socket will be fine, it's just that the region scheduler isn't numa-aware
Someone know plugin for stack mob and items or only mobs compatible with folia
What's the link to that?
Also, question regarding folia. I know it utilize more threads compare to PaperMC. Would single thread performance still be something to keep in mind of or would the multi-thread performance of the CPU be enough?
single thread is still very important
i made the experience that xeon CPUs are terrible for Folia even though they have lots of cores because they have worse single thread performance than my 11th gen i5.
(My xeons are xeon gold from 2017 and i run them with 12 memory channels)
Would a single thread point of 2100 be good enough if the multithread is 69448? Using the AMD EPYC 7742
i have no idea what a "thread point" is supposed to be.
if you talk about some benchmark score, then the number itself doesnt matter.
you can only really use those to roughly compare different CPUs with each other but you still need to have 1 CPU where you know the performance that you can compare against
i like to compare against my i5 as it got good single core performance, jsut as a baseline tho.
I dont have that many players yet so i cant say how well that one would go with +200 players.
(Also that i5 only has 6 cores so it doesnt qualify anyways but using just the single thread perfomance for comparison).
i would not recommend your CPU if you want a higher player count
my xeons (which struggle at 20 players in 1 region due to bad single thread performance) have roughly 50% of the i5's single thread performance
but if your players are very spread out like in a skyblock server then you still might be fine
Referring to the points here:
https://www.cpubenchmark.net/cpu.php?cpu=AMD+EPYC+7742&id=3547
what exactly are you planning to do on your server?
like what "gamemode"
how many players
Run an SMP with roughly around 150 players all at once
Mostly spread out in a 30,000 by 30,000 world
is it a "normal" SMP or is there some custom stuff that will bring people to distance themselves for most of the time?
The only custom stuff is custom world generation like Terra
then people will be close to each other
keep in mind the minimum distance between players has to be ~1k blocks, if they are closer than that then they will merge their regions.
maybe regular leaf/paper might be better for you
but it doesnt matter if you stick to Folia or use regular paper: your CPU probably wont be doing 150 players at once
Paper definitely isn't doing well hence why we're considering the EPYC machine since a user has it laying around
TPS is 8-10 and we've done as much optimization possible that we can that doesn't affect much of the gameplay
hence why we're looking at folia as a solution
what CPU does your current server have?
Ryzen 9 7900X
the EPYC will be a DOWNGRADE for you
Fully aware of that, but we were hoping that Folia would pull through with its multithreading
in a map size of 30k by 30k with 150 players there is a high chance that most players will be in a singular, large region.
you will most likely get similar performance to paper, or even worse
multicore on Folia is very nice, but it still requires decent single core.
A CPU for Folia is basicaly needs the same single core as paper but with more cores.
those extra cores just sit idle for paper but Folia can use them.
but the base single core performance still has to be decent
i hope you get my point, your multicore MONSTER of a CPU has ~67% of the single core of a BUDGET desktop CPU from ~4 years ago
that being said: the best way of knowing is testing.
for example your EPYC probably has the advantage of having more memory channels than the i5 and minecraft loves fast memory.
maybe you can get some extra performance by that fact alone.
i dont know the context of your server but maybe it is still worth a try.
Alright thanks for the input!
as you can see for Folia it is really "it depends" xD
only reason why we're using the EPYC machine is because it just has a crap loads of RAM and with the current market right now, building a new machine is definitely not ideal
i mean you dont need that much RAM for a minecraft server with 150 players
Yeah, but the machine we have laying around was an epyc so we thought we can make use of it over simply sharing a large server with other servers on the 7900X
anyways thanks for the input
honestly might invest into a threadripper bundle since somehow that's cheaper than 64-128GB of DDR5 ram nowadays
FYI - since this is a Folia help channel and the Paper server, we'd really prefer if people didn't push folks to other forks in help channels.
is this flag good?
java -Xms140G -Xmx140G -XX:+AlwaysPreTouch -XX:+UseZGC -XX:+UseLargePages -XX:LargePageSizeInBytes=2M -Dterminal.jline=false -Dterminal.ansi=true -jar {{SERVER_JARFILE}}
Multiple cores still don't mean you've reached the end of your need for single-core performance.
Unfortunately, You need a lot of cores, each of which is strong in order to get best performance from folia & folia structure.
also use -XX:+UseCompactObjectHeaders with Java 25.
This may increase performance and as seen in benchmark tests, its reducing ram usage by 10-20%.
(It also may decrease performance)
^
So... test it. You can use it depending on how your performance goes. But there is a reality about decreasing ram usage.
any help for this ?
message.txt by @amber hare: https://pastes.dev/mrRUEy1Q1i
truncated logs are useless
this is Folia support, not Canvas.
message.txt by @remote sphinx: https://pastes.dev/dR24BwC6sm
As I just mentioned above, this channel is for Folia support, but not Canvas support
Even on Folia, this error shows up, and because of it I switched to Canvas.
You will need to be using folia if you want to receive any support here
If this occurs on Folia, you must provide Folia logs
ok th
ok i switched to folia
Post a log with the issue occurring on Folia, please.
ok now Folia and deleted plugin lagfixer
latest.log by @remote sphinx: https://pastes.dev/VLxUNr00C7
and
hello?
See the bot message above your 'hello?' message.
ok sorry
Who have spark for latest folia
Please elaborate
https://mclo.gs/euBXQCm Any idea? Folia 1.21.8-6
that looks to me like a plugin is doing something wrong? do you have a plugin list
Hey everyone, my server keeps getting crashed by others, is there any way I can fix it?
Here's my log https://mclo.gs/yn48rOo
I found that he's been using this command "/playerviewdistancecontroller:pvdc", but after a while the server crashes
I've made new discoveries on my Velocity; this player has been constantly spamming commands.
This is part of my log https://mclo.gs/wrA0aF0
heyo, quick question: what can cause global tps drops when the highest region is not above 25 mspt
Provide a spark report so we can take a look. Your question is too board and cannot really be answered. @simple berry
oh yeah, ill take a look
but i guess its too late now
it was like for 2 min
then it was alright again
Poorly written plugins possibly
also wasnt as bad this time, probably because there were less players
FTSFolia is a fork.
which is, u cannot get support from here.
also it will be unrelated with fork, your flags is not good for 32g heap.
he just explained his solution
i guess i was a bit serious 🙁
its more or less canvas
understandable
thought its maybe something like a "oh yeah maybe check x or y" situation
also may i ask for advice to improve my flags?
we are running on zen 5
maybe don't blindly copy random flags from somewhere for starters
so true ^
test the flags according to ur specs
flags may show nice or bad performance based on ur hardware.
but if you would want general advise, i wouldnt prefer g1gc on larger heaps.
most flags other than GC flags shouldn't be touched at all
the default values of flags are typically set to something that has shown to work well, sometimes these defaults are adjusted based on new findings, but generally it just doesn't make much difference. Otherwise the default wouldn't be what it is
and also many flags are set ergonomically based on the hardware, changing that will likely lead to worse results
java devs know more than me and urs of course all 95% of settings should not be changed 🤣
at least, override-
I mean, single thread performance still matters
each region is basically a small MC server in its own right
Just to be clear on this, that CPU is going to perform very poorly. It was fine enough nine years ago when it came out, but times change.
no, older xeons are terrible for Folia.
they might have many cores but those cores are slower than a budget desktop CPU from 5 years ago
so newer xeons wont be that bad?
I dont know.
Look at some benchmarks
xeons are generally not very good at single threaded performance
so, probably still bad
Hey, I’m running the server with the following JVM flags and around 150 players.
Do you think this setup is good and stable for this player count?
java -Xms48G -Xmx48G -XX:+UseZGC -XX:+ZGenerational -XX:MaxGCPauseMillis=10 \
-XX:+UnlockExperimentalVMOptions -XX:+UnlockDiagnosticVMOptions \
-XX:+AlwaysPreTouch -XX:+DisableExplicitGC \
-XX:ReservedCodeCacheSize=1G -XX:+PerfDisableSharedMem \
-XX:+UseStringDeduplication -XX:+OptimizeStringConcat \
-XX:+OmitStackTraceInFastThrow -XX:+TrustFinalNonStaticFields \
-XX:+UseFastUnorderedTimeStamps -XX:+UseVectorCmov \
-XX:+UseAES -XX:+UseAESIntrinsics \
-XX:NmethodSweepActivity=1 -XX:+UseCodeCacheFlushing \
-XX:+SegmentedCodeCache -XX:+TieredCompilation -XX:TieredStopAtLevel=4 \
-DPaper.IgnoreJavaVersion=true -Dfile.encoding=UTF-8 \
-Xlog:async --add-modules jdk.incubator.vector \
-jar server.jar nogui
We don’t support offline mode servers, as you have been told already.
i am running it on a intoel core i-3
maybe i use 64gigs max and 4 gigs min
Folia tuning is complicated. Please don't give advice if just kinda throwing out guesses.
Each server has its own specific needs. General help things (like Aikars, etc.) may not be truly helpful, but they've gained a foothold in the community because they've been tested many times.
this is obviously for folia itself, same things valid for paper too.
A bunch of those JVM flags are just aikars ones which probably aren't needed, the rest are either defaults, irrelevant or come at the cost of performance
fr
does folia make sense for a playerbase which is spread out like this? pl3xmap link
imo, no, your server has little players and Paper can handle it just fine. Using your current players and where they are, that's at best 4 regions
even if you get MORE players later, the map is too small and if players keep generally equal distance between them, it will just be one giant region.
okay, ty for the info 😄
is network-compression-threshold=-1 work on folia?
i put it -1 and still getting hell nah of cpu usage 😦
- #folia-dev for dev stuff 2. that event isn't fired on folia
Well im on folia software, sorry i shouldve thought about that, ill move my messages.
Me used velocity+folia too
Should probably update folia to 1.21.8
but you're outdated
damn
Who already test
EDF and WORK_STEALING
kid named filesystem / hard drive errors:
Does someone have the Same Problem that Sometimes Ping Spikes very high on folia 1.21.8 latest because on paper that Problem ist gone
And it also Spikes when i turn on Like a bonemeal Farm Sometimes
check out your system use.
maybe you overload the CPU or RAM or networking?
Is there a Folia-compatible plugin that allows viewing players' inventories and Ender Chests? (Supports offline users)
Is it perhaps API-wise impossible?
we do not provide support for offline mode
Is he talking about offline mode or when a player is simply disconnected?
does it happen to only some and not all player?
sme lie kthe img? n their pign range for those that have shit ping is aroud nthe same latency?
Fair point lol.
Im at a point where whenever I see "offline user" my brain goes to autocomplete to say that.
API wise its defenetly not a challenge
small update on this, we reevaluated our flags to this:
java --add-modules=jdk.incubator.vector \
-XX:+UnlockExperimentalVMOptions -XX:+UnlockDiagnosticVMOptions \
-XX:+UseNUMA -XX:+UseNUMAInterleaving -XX:-IntelJccErratumMitigation \
-XX:+EnableVectorSupport -XX:+UseVectorCmov -XX:+UseVectorStubs -XX:+AlignVector \
-XX:+UseFastStosb -XX:+UseCharacterCompareIntrinsics \
-XX:+UseFastUnorderedTimeStamps -XX:+UseStringDeduplication -XX:+PerfDisableSharedMem \
-XX:+AlwaysPreTouch -XX:+AlwaysPreTouchStacks \
-XX:+UseLargePages -XX:LargePageSizeInBytes=2M \
-XX:+UseCriticalCompilerThreadPriority -XX:+UseThreadPriorities -XX:+UseCriticalJavaThreadPriority \
-XX:+DisableExplicitGC -XX:MaxGCPauseMillis=100 \
-XX:G1NewSizePercent=40 -XX:G1MaxNewSizePercent=50 \
-XX:G1HeapRegionSize=32M -XX:G1ReservePercent=15 -XX:InitiatingHeapOccupancyPercent=20 \
-XX:G1RSetUpdatingPauseTimePercent=5 -XX:G1MixedGCCountTarget=4 -XX:G1MixedGCLiveThresholdPercent=90 \
-XX:SurvivorRatio=32 -XX:MaxTenuringThreshold=1 \
-XX:ParallelGCThreads=24 -XX:ConcGCThreads=8 \
-XX:+AlwaysActAsServerClassMachine -XX:ReservedCodeCacheSize=400M -XX:NonNMethodCodeHeapSize=12M \
-XX:ProfiledCodeHeapSize=194M -XX:NonProfiledCodeHeapSize=194M -XX:NmethodSweepActivity=1 \
-XX:G1SATBBufferEnqueueingThresholdPercent=30 -XX:G1ConcMarkStepDurationMillis=5 -XX:G1HeapWastePercent=5 \
-Djdk.nio.maxCachedBufferSize=262144 -Dusing.aikars.flags=https://mcflags.emc.gs -Daikars.new.flags=true \
-jar server.jar
and it seems to be much better
(also uses hugepages now)
That literally looks like you asked an LLM for every possible flag in java
I doubt half of those actually do anything that matters
good job at setting most of the flags to their default values
i did this on the host java -XX:+UnlockDiagnosticVMOptions -XX:+UnlockExperimentalVMOptions -XX:+PrintFlagsFinal -version
then checked which are NOT set 'to my liking' by default
no i dont set defaults
Stick to aikars flag we provided if you are clearly clueless
Half of the flags you set are just java defaults
-XX:+UseNUMA -XX:+UseNUMAInterleaving most likely does nothing as Folia is not NUMA aware
maybe the GC ones (some else did them)
yeah ik
maybe in the future
Whoever did them, stop trusting them
that said, using large pages can indeed make sense
You should probably also be actually using Folia if you're asking for help here, not some random fork.
yes i understand, just thought an update might be interesting
no further questions on forks
lol deleting your spark won't help
90% of here is default
have fun
in the future the new scheduler will support numa aware
yee i saw
right now, it will not help.
but in the future, thanks to leaf, we can use this and it may help performance
pretty sure java already applies that numa flag if it detects more than 1 socket
Game_type also appears to only have 1 cpu so numa will do nothing lol
I don't know if this will help, but I noticed that Ptero's Eggs for Java 21 doesn't support NUMA. I had to rebuild a Dockerfile, adding numactl libnuma1 so that Folia could see that there were multiple sockets
As just said above - Folia is not currently NUMA aware.
So it's not officially supported. You may encounter a lot of problems with multiple CPUs right now.
Actually I was wrong, it is as of 2 days ago in build 3 but you have to swap schedulers and use a startup flag.
The latest commit allows the use of NUMA (but at our own risk)
well it has 2 ccds
only one imc, but it should help reduce l3 cache misses
well misses are the wrong word prob
transfers
you have to
make ur own image
you probably need glibc and other libraries loaded on image
It's already created and working for me; I've put it on my Git repository for Java 25 and 21
-D%1$s.NumaScheduling=<true/false> - Enable NUMA aware scheduling if supported. -moonrise
for graalvm?
yes
I don't know if I can share it here, otherwise go to my GitHub profile, it's the very first one in the repository tab
i have to look smh on skullians
time to look oracle have a library like this
letss goooo there it is!
also, different topic. Did u look at it to io_uring?
it can make a attack surface but if u allow 2 rules it should be okay for prod servers
"you need to implement io uring to server.jar"
I've never heard of io_uring
like epoll, nio etc
void how do i know if i overloaded my networking because cpu and ram are fine
sometimes to everyone sometimes to only a few
use a tool like btop which shows you the network usage in real time and compare that to a speedtest on the same machine
if the speedtest tells you a max of 100mbits and btop shows a usage close to 100mbits then you are probably at the limit
yea i already checked that it isnt the problem but could bungeecord also be the problem ?
i have no experience with proxy setups but i do know that people really do not recommend bungee here
what do they recommend
velocity
do you know the diffrents
it's made by paper
no but this is not a folia specific topic
overall better #velocity-help
Do the end rings exist on folia/paper servers?
No
is this fine
no
you dont have enough ram for process.
After 9gib, your server cant allocate more and probably os kill the server at 104gb allocation
Also
In the newer builds (1.21.11) there’s a config flag for it
Haven't tested this, but in theory that's the config.
But that's specifically at far distances. Check the Mojira issue for more info.
End rings only exist at far distances due to 32 bit overflow
And paper fixes it by default by casting them to floats
Yes, but it's been "fixed" for a long time, so people sometimes think they're supposed to occur earlier.
Yeah you have to go really far lol
Hi guys how to setting folia 1.21.8 ?
Builds are available on the wesbsite.
Hey There! I was wondering why folia 1.21.9 and 10 are missing or if i am just looking in the wron places, i cant seem to find these
There aren't builds for them. Leaf was busy so he just did 1.21.11.
You generally want to be on the most recent version of Folia anyway, so no real need for 1.21.9 or 1.21.10
okay thanks
Hey guys! I want to make a Minecraft minigame server with various custom duels gamemodes.
Paper servers have a single tick loop which I think is really bad because lag from a minigame can affect other minigames. (I want to be able to add experimental gamemodes without being scared to lag the enitre server)
I would like to ask just how bad you think it is to use Folia for this although I only need separate dimensions to tick separately. I am especially interested in the effect on resource consumption, as I will write my own plugins.
Thank you in advance!
You likely want to actually make a network with multiple Paper Servers + Velocity.
The problem with this is that I've heard that a JVM has a lot of overhead and it would be impractical to have a JVM for each match. I believe Hypixel has not all but several matches running on each Paper instance
Hypixel probably has everything on its own server instance, but they also have their own forks of Spigot/Paper.
Oh, I did not know
they have dev blog where they explain their minecraft choices
they hardforked spigot 1.7.10 back then it's called hygot their own jar
That's cool
If you're aiming for maximum density you'd have multiple server instances still but also multiple minigame instances per server
Folia doesn't allow loading/unloading worlds at runtime so wouldn't be a good fit for that
So I can't make new dimensions?
You can make them with datapacks but they're created on startup and then fixed
You'd want to check Hangar/etc
Might not exist then. Generally if you are large enough to rely on Folia you can usually afford a dev to write or modify plugins for it, there aren't that many public plugins supporting it natively
yet 
We really don't want this kind of random side comment in help channels.
what different between grid-exponent: 4 and 6
6 would make the regions larger.
hello, is the allow-nether: false option removed from the server.properties in folia?
i think i found it, let me try rq
/config/paper-global.yml
under misc: there is enable-nether, you should set it to false
yes, worked. thanks. 🙂
Is there a way in folia to see what plugin causes high mspt when a player joins? Spark profiler --ticks-over doesnt work on folia
There is a spark plugin for folia
I mean, there is no singular tick thread, so an option like that makes little sense
Only real option would be to hook up a dedicated profiler
But high mspt that i saw on spark affects all regions on folia or how does that work exactly? This is my first time trying to work with folia
Regions tick individually
One region generally won’t impact other regions from ticking unless you don’t have enough threads for ticking
Yea but its about onPlayerJoin events
I have very low cpu usage but high mspt, 16 cores 32 threads, should i switch from paper to folia? its for smp
How do i whitelist
currently this my setup
server.properties contents
...
enforce-whitelist=true
...
whitelist.json contents are
[]
yet i'm able to join my server
are you opped?
i already ran
deop on my username
enabling the whitelist and enforcing the whitelist are 2 different options, maybe you have it disabled?
how does one enable it?
Same as Paper/vanilla - whitelist=true
and this line goes in what file?
I can't find much info on the docs
It's a vanilla setting, not a Paper setting. Try searching for Minecraft configs.
oh
it got enabled by whitelist on
command
I think u should check which plugin handle teleport
like huskhome plugin before they was on paper after that they supported folia but still keep the same logic code like paper
so when player accept or do something with teleport, huskhome make an it own thread to accept tpa
that what happened to my side and there 20-30 players sent me an tpa and i have tpauto on and server crashed
Hello. Its there spark for folia 1.21.11?
I cant find
Because i have some problems of tps and i cant see what its the problem without spark
Is there anyone who knows how to resolve this issue? I am currently using the Citizens plugin.
I suggest asking the citizens folks for help with citizens
This not goes for 1.21.11
It have errors
message.txt by @warm crown: https://pastes.dev/12idA42RP1
From here
That is highly unfortunate
So how i can see why tps
its going very bad
:c
Its going like 4tps XD
In paper goes all well its only with folia
Any other info? Most likely some plugin is bad that you have
Can someone confirm this
i cant confirm because spark doest goes in folia 1.21.11
Get a folia server with an ocean only world and get a bunch of nautiluses and see if it lags
Have you reported the issues to the spark folks? 🤔
(Wrong chat, srry)
do yall think its a good idea to update to folia 1.21.11
or wait till its no longer experimental
You are the experiment*
does this mean 1.21.8 is unsupported now?
For Folia, yeah.
my players told me that attribute swapping allows a player to combine the damage of an axe with the range of a spear.
is that possible on Folia too and if yes, is there a fix for that?
paper-global.yml
update-equipment-on-player-actions: false -> this one allows mace swap
update-equipment-on-player-actions: true -> this one fixes mace swap
is there some sort of setting for connection throttling?
last time i started my server all the people got hit with "Connection throttled" and only 1 player every 5 seconds could join.
this is 100% not intended.
(I had an event set with a time and on that time i just did /whitelist off so it was ~20 people trying to join at once)
paper-global.yml
max-joins-per-tick: 99999
bukkit.yml
connection-throttle: -1
guys how tf do these dev build compatible plugins for folia 1.21.11 please help
Do you have question about Folia? Cn you share more in detail what exactly that you need help with?
ye mb for not saying from the start what do u guys do in the maven pom file
abt some api or idk
folia api for 1.21.11
can i send the code herer do is it allowed
imma jjus send
message.txt by @buoyant river: https://pastes.dev/f7D2ISgNl1
Report from players: ender pearl issue: ender pearl stasis chamber set up in the End and then player goes to the overworld, the pearl in the End is gone when they go back to the End. Same with if the ender pearl is set up in the overworld and then player travels to the nether, the ender pearl is gone when they go back to the overworld. This issue seems to appear after we updated from Folia 1.21.8 to Folia 1.21.11 (Build #8)
Hello, any idea how to fix this warn with folia 1.21.11? (it spam my console)
this seems like a spark issue, best to report it to them
i just noticed that the download page for older builds of Folia still states that 1.21.8 is supported, is this an error?
Guys, is datapack finally stable for all kind datapacks?
There are no plans of supporting datapacks with command functions
Damn, I guess structure is good enough
Those sorts of datapacks tend to be pretty slow, and for Folia you're already likely to be writing your own stuff anyway.
Any idea when will Folia support all vanilla features like Paper?
All vanilla features? Probably never.
It's not really designed to replace Paper, it has different uses. So some stuff just isn't likely to work because of the nature of how its built.
damn
i set these settings but people still cant join the server and get the "Connection throttled" message
3 players get in right after a restart but even after 2 minutes have passed no other player can join
Do this? ^
Yeah it should
where can i find the download?
it is weird, after a few more minutes a handful of people could join
but still not all
and the ones who can join always join very close to each other in terms of time delay between them.
they always come in groups of like 3 players
i dont know if it helps but i made another report with 29 players online. (Players can join with an average of like 30 seconds delay per player)
https://spark.lucko.me/M3TXBBJrJO
this is not -1
bruh
ok yeah i fixed it now it works, thank you.
ima go into a corner and cry 😭
Spark reports to the rescue!
Typos in config files do happen sometimes, so don't beat yourself up too much.
I tried to help out here since I'm playing on their server, but I wasn't able to because I've never used Spark. For future reference, how do you find config files?
If you're just a player and not an admin, you won't have access to those files.
Txslx doesn't seem to be an admin here, either.
in the configurations tab on spark
You might need to press the I button on the top row to get it to show that info
I meant an admin on the Minecraft server.
Oh, I see what you mean. I thought you meant view the actual config files. Duh, sorry.
Yeah the tab in the spark report.
I think there's a video (or howto?) on the spark website about how to read a spark report and what to look for.
That is what I was missing. Using an unfamiliar software is very difficult when all the buttons are just icons.
Their docs are decent.
I need WorldGuard for Folia but WorldGuard for folia doesn't seem to work
Here it says only on 1.21.10
But there's no 1.21.10 version for Folia
Just try it on a verson nomal worldguard supports
You could try the EngineHub Discord.
Or just try it anyway.
Enginehub most likly wont help
I have no idea what changes this fork makes so no idea if it's safe to use or not.
Some people fork a plugin and then just modify the plugin config to "support Folia" but make zero other changes. So caveat emptor.
Yeah but those don't work, the task scheduler is completely different
Which causes WorldGuard to crash
We've been using it ourselves for a very long time, can vouch for it, but again, if they want it, they gotta take the "risk" themselves
Yeah
I love it when people deliberately make it harder by not including a .jar file
Yeah that's fair. Just not something we (as Paper) can recommend.
If you're using Folia then building plugins is the least of your problems.
If you're using Folia, you should have a basic understanding of compiling plugins
I do
Should be zero issues to compile stuff.
Thanks for showing me the hidden gem that is freshsmp
Just a f u to the developer
np, Root likes to promote Folia
And i thank root for open sourcing it all
almost all of their plugins have folia support and they seem to be using the plugins themselves on prod
since they use Canvas (fork of Folia)
For a while (like just now while) i was using a jank build of worldedit and worldguard that i made combining 2 forks into 1
Just discovered canvas, it's amazing!
A reminder that this is a help channel - please take side convo to somewhere else, thanks!
That's why i didnt say any more
it doesn't load on startup, what version of Folia are you rocking?
I'm using 1.21.8
Oh I see
Question, does Spark work automatically on Folia or is there a fork of it that supports it?'
There's an official spark build for Folia, it doesn't automatically install.
Search this channel for info on where to find it.
Are you also using FreshSMP their WorldEdit?
https://github.com/FreshSMP/FastAsyncWorldEdit-Folia
Blazingly fast world manipulation for artists, builders and everyone else: https://modrinth.com/plugin/fastasyncworldedit - FreshSMP/FastAsyncWorldEdit-Folia
i am, but it has bugs
use it at your own risk
I wonder why Folia developers didn't just keep the same API calls as Paper and not just change the logic to function in the new multi threaded environment
Would you mind sharing the jarfile with me in DMs? It's been building for 7 minutes and is at 30%
We don't really recommend this. Txslx seems like a relatively trustworthy person but that could be a random jar that contains malware.
He's the owner of LifestealSMP
Also, again, if you can't build a thing locally then you're going to have a hard time with Folia.
I doubt that.
I can build it on my Ryzen 9 9950X machine but I'm not bothered to do so
If you can't be bothered to build something then people are not likely to be bothered to assist you with stuff.
Anyway - please don't ask people in here to send you jars.
You're a great person, you've been a huge help in assisting me so far, I really appreciate you.
Your sarcasm is not appreciated. People have done their best to help you here. You haven't done your best to receive that help.
I genuinely hope you get things working and have a great weekend.
The fact you think it's sarcasm tells me you know very well how bad of a job you've been doing here.
I wish you the best of days
I've muted you for a day. Sorry. Hopefully this will help you in the future.
Does anyone know of a stable alternative to essentialsx that supports folia?
I would probably just recommend EssentialsX
Custom fork with folia? hmm
there is a fork available with Folia support maintained by someone, you wouldn't need to make your own
Do you have the link? pls
I don't, but surely Google has one
Euphoria's github page has a link to it, forgot the atual link
I joined to see if anyone has the site link for that region tester, I saw someone use it to plan out how big a world border needs to be so I wanted to try it myself.
I don't think it's public. Just a project someone made for themselves.
Perhaps they'll link you if they see this message, though.
But there's a bunch of random math distribution tools online. You want like 1000 blicks from each player. So you could maybe find a tool that'll simulate that
Ah fair enough, thank you ❤️
I run Folia on a Xeon system with two CPUs. I read in the chat that Folia 1.21.11 can now use NUMA? Do I need to adjust anything in the Java flags for this, or does Folia recognize it automatically?
i talked with rootbeer on this, he have questions in his head when he made this
overall all plugins which is ported to folia may not give you the experience/smh you wanted.
someones may, but someones may not.
You’re replying to someone who has left the server.
welp, didnt saw they left
Yep, just a heads up.
see the commit message here https://github.com/PaperMC/Folia/commit/eee7128bc810195ad758ea5ace1b72c600896d3b
check the numa support in your cpu with this: numactl --hardware
if you have multiple nodes more than 0, (at least 2) you can use it.
(for example, 9950x has two physical CPU chiplets (CCDs), they both connect to a single I/O Die (IOD) that manages the memory)
Thanks for the information. I will check that on my system.
also lemme give you the output of command
available: 2 nodes (0-1)
node 0 cpus: 0 1 2 3 8 9 10 11
node 0 size: 32128 MB
node 0 free: 12450 MB
node 1 cpus: 4 5 6 7 12 13 14 15
node 1 size: 32256 MB
node 1 free: 28100 MB
node distances:
node 0 1
0: 10 21
1: 21 10
(as example*)
we may try -XX:ZFakeNUMA=(0-16) but it needs a fastdebug compiled jvm (debug version of jvm) (because this flag is not in product.)
for especially on testing purposes
My system appears to have two nodes. What could be the disadvantages of the new scheduler? Instability?
Node 0 free is 5.8gib and Node 1 free is 766mb 💀
yeah, you have 2 node, you can give a try
that doesn't make any sense if you don't want to debug how zgc allocates regions
the tests are limited, i mean not everyone have multiple nodes
thats why i said test purposes
the fakenuma shouldnt be used in prod
it's not for testing applications
for debugging zgc, yes. Not for debugging the folia numa-aware scheduler
(compromised account, already banned)
hey, i'm wondering if it's worth getting folia for my buddy's server.
it doesn't have a lot of players, roughly like 5-15 online at a time, but i'm noticing how most of the times, tps seem to drop without RAM or CPU usage going past even 50%.
But looking at the main thread usage, it's almost always running at 100%, which is probably caused by a lot of massive cactus farms and stuff like that.
The players are usually spread apart like 20k-30k blocks and in mutiple dimensions (mostly overworld and end), but I'm not sure if that's enough for folia to actually do its thing and be more effective than paper already is
Your best bet would be to just try it
Given all plugins are compatible (if any)
But in theory it should perform better
It’s possible. Typically we’d say “send a spark report to figure out what’s causing the problem” because it’s possible that you’d switch to Folia and have the same issue(s).
wait let me create one
I just do spark profiler open to get the link I assume?
https://spark.lucko.me/CHXg4miLiG
This is 1.5h of spark profiling, hope it's not too much
let me make one with spark profiler start --timeout 11 real quick
Your issue is just that this CPU is not suitable to run a Minecraft server
How are you hosting the server? Most reputable budget server hosts should give you better performance
20% of your tick is ticking villagers, but you only have 10 of them
here's another 11s report at a higher load: https://spark.lucko.me/X30Bcu6FGn
currently a VPS
8gb ram and 4 vcpus running ubuntu
running on very low budged I think, not sure
i'm guessing it's this option on hetzner
so do you think folia could reduce/fix the lag, or would he need to get a completely different server
It’s only so much to can paddle your bicycle to make it go faster. You need an actual car in this case.
hm alright. Is there any hosts you can recommend that would handle the server fine?
preferably under $25 and US east coast
To put this a different way that isn't a boat analogy - Folia works "faster" by taking more advantage of a faster CPU than Paper would.
So in order to really get a benefit from Folia you still want a faster CPU. Folia is aimed at higher performance stuff. It doesn't really "squeeze" more speed out of slower CPUs.
And really, most of the folks who say their software does are being misleading. You can make stuff a bit faster but it means making compromises like disabling stuff or making various tweaks.
So that's usually where we tell folks to start. Get a good host, tweak and config your Paper server. Once you hit that limit then maybe Folia would be a good option (if your server is geared toward it).
hmm I thought folia would maybe be faster by using the rest of the total CPU, as currently the load is just 100% on the main thread and all the others are pretty much unused
You still have a pretty crappy CPU
Folia wants more threads and high clock speed/fast CPU.
70% of your server tick being spent on only 20 villager entities is quite worrying (though, it is only an 11 second profile)
hm alright yeah
Yeah there's some stuff like this in that report that you can focus on. Tweak the server, get a better CPU, etc. You shouldn't need to move off of Paper with this setup. You'd be limiting your plugin capabilities (and other stuff) if you switch off Paper.
is there any decent server he can get with this though?
In no particular order, lagless.gg, Prismnodes, bloomhost, lilipad.gg, you can check them out and with good budget of $25 you can get ryzen 9+ with all of them if switching is what you decide to do.
alr I'll check them out
thank you
All of them have ping checker for locations, you can do your own research but those hosts are all pretty decent.
are the VPS options from some of them also good, or should he just go with the minecraft-only options?
Generally you don't want a general purpose VPS for Minecraft.
The Minecraft-specific host options tend to be a better bet.
alr thanks
frankly I dont quite understand how this issue occurs.
what happens here:
- a player seemingly sent a register packet containing a lot of plugin channels to register on (or multiple of them)
- the player gets kicked after registerin 128 channels because the maximum amount of channels is exceeded
- the tick thread that was responsible for doing that packet ticking freezes.. forever? The line from the stacktrace of the tickServer method does refer to the code that ends up ticking the plugin channel packet. but the stacktrace doesnt contain any information besides that
😿
It’s stuck in the while loop
The question would be why, id expect disconnection to clean that up, probably want to see if it can be reproduced without plugins, etc
I mean wouldnt that require hasPackets() to continously return true? and that only happens when the packet queue has been popped
also wouldnt that be visible in the stacktrace atleast some of the time
*also maybe interestingly, ive just seen someone else, running a fork of folia, with the same issue, but in their stacktrace the TickThread ownership check and the subsequent arraylist contains check is also visible.
Between these two CPUs, how massive would the difference be when using Folia?
Is it worth paying more for the one in the first photo?
Thank you in advance
Depends on how many heavy regions you'll have
Between these two CPUs, the difference on Folia is noticeable, but it’s not just raw performance, it’s about how Folia scales with CPU characteristics.
- The Epyc (64c/128t, lower clock) offers much higher parallel capacity, which helps when you have a lot of regions running at the same time (large worlds, many players spread out).
- The Ryzen (16c/32t, very high clock) has much stronger per-core performance, which is great for heavy workloads inside individual regions, but it has less total parallelism.
So whether it’s “worth paying more” for the first CPU really depends on your workload:
- If you’re running huge worlds with lots of concurrent players, the EPYC can scale better.
- If your player count is lower or more centralized, the Ryzen can perform just as well or even better with proper tuning.
ahh chatgpt explain
anyways
In general, the EPYC would be able to handle significantly less activity per region than the Ryzen, but way more regions
While testing, I found something really interesting.
I’ve been testing large worlds on Folia — for example, a 100k × 100k map with ~1000 players — and observed the following:
- On low-clock CPUs (2–3 GHz) with very high core/thread counts (96–192 threads): → Lowering the grid-exponent helps distribute load more evenly across regions, allowing the server to handle many players without region-level lag
- On high-clock CPUs (5–6 GHz, e.g. Ryzen) with fewer cores (~16 threads): → Increasing the grid-exponent reduces the number of active regions, which helps prevent excessive region scheduling from consuming CPU time
so :
- low clock + many cores → lower grid-exponent (e.g. 2–4, depending on workload)
- high clock + fewer cores → higher grid-exponent (e.g. ~5–7)
In my case, with a high-clock CPU, the default grid is 4 had ~250 active regions at ~300 players.
After increasing the grid-exponent to 7, the number of active regions dropped to ~80–100, which saved a significant amount of CPU and allowed room for more players
Are you testing this with bots?
But basically that’s what KP said. If you have more cores you’d get more regions, but less activity per region.
is folia likely to break many public datapacks the same way it does w plugins?
It'll break any datapack that uses command functions
And on Folia you shouldn't be using datapacks anyway.
Real players
You have a Folia instance with 1,000 real players and you're able to swap physical hardware and configs a bunch to be able to test this?
no no just 300 players
Oh, your message said 1,000 players.
So in your testing did you swap hardware and configs multiple times with the 300 players?
Or was that an example also?
Ah okay, I misunderstood your message.
Apparently on my server a specific mechanic has worked differently for years and I didn't notice. It's the mechanic around item-merging. The vanilla mechanic is that when there's a moving item and a stationary item and they are in merging range, the moving item will merge into the stationary item. On my server it was the other way around. The stationary item would merge into the moving item. People have apparently built loads of farms around this principle and now those farms are broken. The thing is, I don't even know how I "fixed" it to make it go back to vanilla
But I want to unfix it now
Is there an option in paper, spigot, bukkit, configs that I don't know about?
Also I'm on folia
Which is why I asked here
see reply in #paper-help since you crossposted.
Is sand duping in an end portal not a feature within Folia?
the same options that exist for duping in paper exist in folia too
Those are already set to true
if you let a block fall from up high through the portal it works, but a piston sand duper doesnt seem to work
Hi! I want to switch to folia but multiverse core isn't supporting that software. What shold i use instead of that?
theres https://modrinth.com/plugin/worlds-1 but frankly all of them rely on unsupported API
so - has a chance of making stuff not work
is there a known bug that /setworldspawn does not work correct on folia? it does not change the wordspawn if you move it to other cords
have you tried without any plugins and datapacks? (It works fine on my server but it is not on the most recent version)
yeah latest version
hey! are there any dupes active in the latest paper/folia version?
We don't really discuss dupes or exploits publicly. Anything that we find gets patched, is the best answer I can give.
alr! tysm
They are mostly caused by poorly made plugin as well so I'd start with that.
They are not recommended.
Is there a way to adjust Folia region grid size?
to some degree yes, but if your players are 500 blocks or less, there is no solution.
is that a chat gpt answer thats ultimately not even really true?
@latent wave please do not reply with AI generated answer, normally I wont delete the message but the one sent perviously is just flatout incorrect so I will remove it.
MoreFoWorld
Try the plugin foliaflow and use the same config settings for paper server with sand dupers
Try the plugin „Tweaks“ it’s working with it pretty good
That plugin is archived and not getting support anymore
im using it for 1.21.11 and its working fine with the latest version of it
Hi everyone,
I’m having an issue with a player on my server and I was hoping to get some help.
One player can’t connect to the server, and when they try, it only shows the message "UUID of player."
I checked the player list, and it shows that they’re always online, but when I teleport to them, I can’t see their entity.
I’ve tried using the kick command to remove the player, but no matter how many times I repeat it, it doesn’t work.
There are no errors in the server logs. I asked around elsewhere, and people suggested it might be an issue with the scheduler.
Post a log of them joining and leaving/being kicked?
I have seen reports of similar behaviour in other discords
I'm getting the same on DonutSMP.
this rarely happens and it's due to teleportAsync, idk why but some chunks don't seem to complete the load which makes them never actually get added on the region
so there is a bottleneck about teleportAsync cant reach the speed?
I mean, bottleneck has a pretty explict meaning
chunks not completing a load wouldn't be a bottleneck, that would be a pretty bleh bug to deal with
my poor 9950x 💔 i need help with netty, does anyone have zstd for velocity and folia?
what do you want to you use zstd for?
to make server not using too much cpu for netty?
Start by posting a spark report to look at CPU usage.
wdym? i have my own metric to monitor my whole full machine and unifiedmetric 💔
If you're asking for help here and it's regarding CPU usage, we'd want to see a spark report to start.
oh no im just asking if anyone have zstd patch for folia and velocity so my server can use it instead zlib
That won’t help—the Minecraft client does not support Zstandard. We can compress world regions as Zstandard, not the protocol (unless you mod the client OR Mojang supports it officially).
(I know it’s possible. I’ve actually experimentally done such a patch with a Fabric mod.)
If you want to lower Velocity networking CPU usage related to compression you can lower the zlib compression level (set to 6 by default) and raise the compression threshold (set to 1024 by default).
That would require a non-standard patch to both Velocity and Paper/Folia.
does 8 schedulers running all the time lag the server that much?
Depends on what they do
that running all the time in 8 worlds
You want #folia-dev for dev questions.
mb
No worries
reuseport and io_uring may help atp lol.
but needs implementations by user.
Hello, I have been testing copper golems functionnalities recently and saw that if I go away the copper golem starts to be reallyyyyy slow (6 or 7 times slower than in vanilla). Is there a stupid reason I missed that this might not work ? (I am on a clean version of folia 1.21.11(build 10) without plugins, and of course, the chunk I put the copper golem in, is chunk loaded) I guess it has to do with a sort of activation range, but the wiki says that copper golems are excluded from this
entity activation range maybe?
"Unlike most other passive mobs, copper golems do not suspend their behavior when more than 32 blocks from a player. " straight from minecraft wiki, I thought about it also at first
I am talking about the spigot/paper setting named like that, not some vanilla mechanic with similar behaviour
also I am confused by the "slowing down" part, it still "works" just incredibly slower
that would point towards the entity activation range, as that has a periodic wakeup of the entities every few ticks built-in
my bad x)
excluding/separating copper golems from the rest like vanilla does it could make sense though
that's what I was going to say, as it stands I have (in my knowledge) no way of keeping them activated other than maxing entity activation range for I think "misc" category ? But if I do this it's a nightmare performance wise
I wouldn't call it a "nightmare", especially not on folia xD
best to open a feature request on Paper for that tbh (if there isn't one already)
message.txt by @vapid lynx: https://pastes.dev/zql02koQiE
Does paper's anti xray work on folia?
yes
What is the utilisation 400% thing dictated by? Like which config option. I'm wondering how to increase the max capacity
the # of threads configured
the "threads" option in paper-global.yml
So currently this -1 value causes it to use only 4 threads? I have a 7950x3d with 16 cores and 32 threads. Whats the max physical upper limit on threads I'm able to use? Is it 32 or 16? And what's a sane limit? Like if it only counts cores, should I go with like 12 for the threads value
grid-exponent: 4
scheduler: CRS
threads: -1
Either way it seems that 4 is far too low when I have 16 cores
Ngl that doesn't make much very clear
What CPU are you running? And what does grid-exponent being 3 vs 4 really mean
there is no "official best way" of doing it.
i just roughly follow that guide and eyeball some stuff
Can you answer whether threads in paper-global.yml refers to a CPU core or a CPU thread?
grid-exponent basicaly controls how big each region is
threads refers to cpu threads
So 3 means the regions are smaller? And so you have more regions?
yes, with 3 i get multiple regions in a scenario where i only got a single one with 4
Damn so I can put it up to like 24 and it will be within the bounds of my 16c-32t cpu?
that is the save approach yea.
you can do that for starters and if you run out of tick threads again and notice some CPU threads still being idle then you can increase it
like i said, start somewhere save and then eyeball it if you need more
having 24 instead of 4 will already be a very great improvement
Yerp
but you should also take into account that other tasks require multiple threats too, like netty IO or GC
Surprised I've been doing this well with just 4 threads in the first place
yep
and reserve at least 1 thread for the OS itself
you got a CPU with very good single core.
so it can handle a lot on single core performance alone.
its just that having more threads takes advantage of the whole cpu
Got it originally for single-threaded performance
i would say it is a very good CPU for Folia
I think so too
theres still a few ticking behaviours in folia that can cause these exceptions, when an entity in one tick region attempts to access entities of another tick region. like this one
How good is the server health looking now with the updated thread config?
Good, I'd say. Considering that our multi-region ticking thing was only taking advantage of 4 threads instead of 20
And we've had such improvement already from 4.
Awesome! Would be interested to see the utilisation stats when we have that many ppl on a again.
is there any way to decrease region merge distance?
the closest thing we got is the grid-exponent setting in paper-global.yml
reducing that number reduces the region size (Ofcourse only if people are not too close to each other).
On a server a while back I only got 1 region with that being set to 4 but when I set it to 3 I got multiple regions
Hi everyone!
I'm trying to use multiple worlds on Folia, but I'm having issues: Multiverse doesn't work, and the Worlds plugin is breaking the chat.
@sly pagoda don’t cross post
sorry
does anyone know how to force the MultiThreaded Tick Regions to have a Max amount of chunks in them, cause on our server the players are all about in radius of 3k blocks around the spawn and there are like 2-3 regions so the TPS are at rock bottom with like a max of 11
You cannot force that. If players are too close, they will need to be in the same region.
How far must players be that they are not in the same region anymore?
Around 1000
If players pass by close to each other causing regions to merge, they'll resist unmerging, so it could be more in that case
It highly depends on the configuration
It relies on shifting coordinates to do the merging
Technically with your own hands on the deck you can get that to be a few chunks
Just gotta have the courage and mess with it
Not sure if its a bug or setting but on folia we have a very commonly used raid farm but the raiders seem to not go through the portal, My guess its the portal cooldown which is either higher as a setting of paper or a bug of folia. or something else.
Probably worth investigating what is their despawning behavior in multiplayer, it's a little peculiar
Raider despawning
Raiders that are part of a raid do not contribute to the mob cap and do not despawn naturally until the raid is over or they get far enough away from the raid center. It is also required that the chunks they are in are reloaded (for example by getting out of render distance and coming back) or they don't disappear even if the player is >128 blocks away.
Maybe it's an instance of
case is, I am on ther other side of the portal but they still dont go through
I load both sides (With my alt)
either I built it wrong (Dont think so, I was using the schematic) or its a bug or setting
Entity has a cool down before they can enter portal again, how long has those been there?
dont know, but how long is the cooldown?
I saw people in the comments on youtube that it works in 1.21.10 but cant find out why this doesnt work
Since I dont know if they are using fabric, paper or folia, where I am right now on folia
Is there a way to reduce that? To like 250 blocks?
Where can i configer that?
global config
Do you know what do i have to change?
To what?
2
And what does this do? Do the regions get smaler? And what about 1?
Default is 4 right?
idk
i dont think u can do 1.
try it
but the mem overhead will be crazy
cpu too
But how with grid exponent 2 in the config it seems very big
are you sure your players are spread out enough?
ok and how du i do this?
they are to close thats the problem
Folia isnt an ideal solution for you if your player are just too close. There is no way around it @north light
then Folia is just not the solution you need
but is there no way to reduce the 1k to like 500 blocks of player distance
yo, I'm new to writing plugins for Folia and I'm currently unsure whether to choose Kotlin or Java.
do you know either of them?
I know both.
the general go to is java since basically the entire eco system is java only
so i would recommend using that
maven is awful and gradle is awesome
thats all i am going to say
Okay, thanks guys!
Nope. ~1000 is the smallest it will go.
If the players are too close you should probably use Paper.
Do you know why?
Two region cannot touch each other. The space between two players has to be certain distance over as buffer.
(Most non-technical explanation i can come up with)
What Eternity said, basically. It's how Folia was designed.
If the players are too close they're just in one region.
Yeah i know i figured that out… merge at round about 800 and the merge split at 1200…
But if the server view distance is at 8 chunks woud 400 Merge in and 650 Merge out work?
Please see what ocelot said. This is a technical limitation.
It has to be higher due to worldgen
~900 is the lowest it can possibly get.
If players are closer, you want to use Paper.
Does this improve the Performance with 50 players?
You can do 50 people on Paper easily with slight gameplay nerfs.
problem still persists on folia. Otherside plugin is not updated and not folia compatible. How do I fix my farm tho?! 😄
Hey guys, anyone knows anything about this?
Players are bugging when going through Nether portals while inside a boat (sometimes with a villager). The screen turns completely black, like falling into the void, and after disconnecting the player can’t rejoin because the server thinks they’re still online.
[21:33:35 ERROR]: [ca.spottedleaf.moonrise.common.util.TickThread] Thread Region Scheduler Thread #0 failed main thread check: Cannot update owner state asynchronously
when they do that, i get:
Entity threw exception at world_nether:5.298300709645618,72.0,5.881312189102742
java.lang.IllegalStateException: World mismatch: expected world but got world_nether
Which version / build?
Post a log
.4 is no longer supported
I semi recall something about that bug on the GitHub, no idea what state it was in
Yeah would update. Lots of changes there.
how do i make sure my plugin works well with newer versions of the server?
Test it?
should i keep a copy of the old server
You should always have backups
Hey, I have two little questions. I heard that folia even does improvements with less than 16 cores is that true?
Atm I have rented 2 vps one with 6 cores and 3,7ghz CPU and one with 4 cores and 5,7ghz CPU would it be worth it to host folia on one of them than just use paper on the 5,7ghz CPU?
It is not reccommended but nothing is stopping you from trying
If you have 6 threads on that VPS, i would say dont even bother.
If you have 12 then its still too little but might work out if your server is not that big
Folia basicaly wants the same single core performance as paper, just more cores of it at once.
So you really do not want to chose a slower cpu just because it got more cores.
I got some xeon gold cpus with plenty of cores but they suck for folia because their single core is super bad
1.21.11, build #10
Sorry for the delayed reply.
My server data mysteriously disappeared, and I can't find any logs regarding deletion. Therefore, I can't find the original logs.
The news of these players leaving and disconnecting from the client may be due to my use of velocity.
A read timeout is displayed on Velocity, and then the client shows a connection timeout.
There is nothing else. After some investigation, I guess it might be caused by a plugin, because this problem did not occur before I installed this plugin.
Is it better to pre-generate world without players on Folia or Paper? If Folia, how can I set up config to utilize all cores for parallel generating?
paper
Do it without players and plugins on Paper
I have a question regarding the problem with ender pearls vanishing when you change dimension. I looked it up and found the Github post(Fix Folia #421/#423) but I didn't quite understand where to find the information that needs to be changed to implement the proposed fix.
There’s no fix merged. You could fork folia and modify it however you’d like, but that’s outside the scope of this channel.
I have a folia server that doesnt seem to be adding new regions for seperate areas of the map
it only creates regions for overworld and nether
Are your players too close together?
Players need be 1000+ blocks away from each other.
yes, i have two chunks of players 3000 blocks from each other
Send a spark report?
will do
Has anyone else experienced a situation where one region has low TPS, while the rest don't? This region has practically nothing, 0 players. Looks like generating chunks
We don’t support offline mode servers (as you have been told).
How can I change wandering trader trades on Folia?
"change" in what way?
Remove vanilla trades and add custom trades
you would need a plugin to add custom trades, or wait until 26.1 releases where they become data driven and you can do it with a datapack
I could not find plugin for these purposes that supports Folia, unfortunately
Folia is more experimental so you need to be prepared to make custom plugins if you want to use it
"I" already made two plugins to block journeymap+xaero and for team management, but I'm locked out until Friday :D
I guess wandering traders will be disabled for now...
how do you block a client side mod?
I think those two mods have some special feature where they can be disabled by the server.
But it is build into the mods
so it still rely on client trust?
seems to be easy to bypass
Yeah there is a """"hack""""" version of xaeros which removes that.
The official xaeros page even explicitly calls that version out
But please dont rely too much on AI, one or two simple features are no problem but anything extra ontop of that and you will end up with the worst code ever.
Yeah the resource is ARR but people who use hack ain’t going to care lol
It also detect using of xaeroplus somehow (just tested)
UPD. Looks like it took implementation from Minimap Control plugin
some mods make it very easy (xaero disables certain features when you send it a chat message and also opens plugin message channels), but theres also currently an exploit that allows you to figure out what translation keys (=keybinds) are registered on a client which allows you to figure out if the client is using certain mods. but obviously all of that still relies on client trust and can be bypassed
but in any case this is not really relevant to this channel
hey guys! do know whats may be causing this?
did you check the spark?
what should i check?
something unusual, like a plugin with high cpu usage
looks like an issue that i had
spark doesnt show my plugins use, idk y
its just with that region?
or changes over time
2 different regions
and check if that happens from the moment the server starts or maybe some point in the time
wait. Spark is avalible on folia? the command disappeared when switching from paper to folia
You need the plugin
https://spark.lucko.me/0AVfuwjUxb
seems to be empty
resolved the lag issue, one of my staff set randtickrate at 9
any other JAVA args i should add:
https://spark.lucko.me/1z8IjUTRIO
the latest commit of folia fixed this bug
Does initial-enabled-packs=vanilla,minecart_improvements doesn't work on Folia 1.21.8? I don't see /gamerule minecartMaxSpeed command
is your world already exist before the datapack is enabled or after?
my world was already pre-generated, I forgot to enable this datapack
You have to manually add them in then
@little grotto see this #paper-help message be warned to do backup first
Has anyone seen random player-specific timeouts only on Folia, while Paper works fine behind Velocity?
The players affected usually disconnect when close to mobs or other players, and both Velocity and Folia only show a generic timed out. We already tested without plugins and even disabled Anti-Xray mode 2 (helped some players, not all). Could this be related to entity updates, networking, MTU/packet fragmentation, or Velocity keepalive/compression, or even something involving Geyser/Floodgate? Any insight would help a lot.
The setup is a Ryzen 9 7950X host with 128GB RAM, running Velocity (8GB) in front of a Paper lobby 1.21.10 (4GB) and a Folia 1.21.8 (28GB)survival backend. The issue only happens on the Folia server, the same players stay stable on the lobby. We also tested with DDoS protection disabled and mirrored the setup on a test server, but the behavior stays inconsistent (some players improve, others don’t).
on folia specifically, ive seen ping spikes, sometimes at the same time as IO spikes, never timeouts tho
Is there a way to divide a Folia server into shards? So that multiple physical servers can host the same world.
No, that's a different kind of thing altogether. There was a project for doing that called MultiPaper but afaik it had some problems and was even harder to write plugins for (distributed computing is basically the hardest kind of programming there is)
Also depend on why your “same world” means in this context. You can use the same seed on all your sub servers and host them individually by divides the word into smaller sections with borders.

How to achieve that?
With a bunch of custom software, basically
Ask the guy above you /s
But yeah DonutSMP has the current best impl of sharding with Folia imo.
Honestly it's too crazy
That's why I quit
Endedup on a psycho place
Hey that's why I quit Bukkit too 😄
Sometimes u gotta pick your value and health
know a fix