search

PaperMC / Paper

The most widely used, high performance Minecraft server that aims to fix gameplay and mechanics inconsistencies
https://papermc.io/
Other
9.96k stars 2.31k forks source link

end is severly glitched .build 46 . END unplayable!!!!!! #5996

Closed Aditya14011 closed 2 years ago

Aditya14011 commented 3 years ago

Expected behavior

ability to normally enter the end and load chunks

Observed/Actual behavior

as soon as a entered end . it lagged and crashed. the few spawn chunks loaded but nothing else did the logs r in the paste.gg link https://paste.gg/p/anonymous/9460ad9f28654ea3af6f416a16bf759a

Steps/models to reproduce

make endportal enter end

i tried it on several servers with paper build 46 1.17 it crashed every time after not responding for 10 sec

Plugin list

no plugins just the mojang caves data pack

( i dont think that should affect anything as the server ran smoothly before we entered end. the nether worked fine too)

Paper version

This server is running Paper version git-Paper-46 (MC: 1.17) (Implementing API version 1.17-R0.1-SNAPSHOT) (Git: 9e07703)

Agreements

Other

No response

jackburkhardt commented 3 years ago

I can't reproduce this at all, even with the caves and cliffs datapack. When you start the server back up, are you able to enter the end again at all? Or will it always crash no matter what?

In addition, the thread dump you posted says in call caps not to report to paper and that it's not a bug or a crash, your server is just not responding. If all of your testing happened on Aternos, it could be something on their end causing your server to lag out. Someone can feel free to correct me if this is actually a bug though.

embeddedt commented 3 years ago

I have always found on my (relatively weak) personal machine that generating the end in both 1.16 and 1.17 for the first time often causes a very noticeable lag spike. If Aternos was sufficiently loaded at the time that you were testing this, it's likely that the lag spike became so long that the server watchdog simply stopped the server.

The solution is probably to just pregenerate a 200-block radius in the end at a time when Aternos is less busy.

Chew commented 3 years ago

Not entirely sure how end generation is so expensive considering it's nothing... why there's structure lag when they're no structures anywhere near the spawn is beyond me

TauCubed commented 3 years ago

Not entirely sure how end generation is so expensive considering it's nothing... why there's structure lag when they're no structures anywhere near the spawn is beyond me

Aren't the end pillars structures?

Chew commented 3 years ago

Was specifically referring to randomly generated ones (End cities), which only spawn on the end islands. The pillars are the exact same position in every world nowadays I believe?

kennytv commented 3 years ago

We still have been unable to reproduce this, so it would be great if you could give us a timings report on the setup you get this issue (so we can get the configuration setup and some other info timings provides). Also, does this actually hard lockup your server, or does your server tick/log normal messages again at some point after the first 10 second watchdog warning appears (do you see entities moving, can you drop an item or hear place sounds, aka does it "just" lag a lot or fully crash immediately)?

embeddedt commented 3 years ago 
[09:25:01 INFO]: This server is running Paper version git-Paper-46 (MC: 1.17) (Implementing API version 1.17-R0.1-SNAPSHOT) (Git: 9e07703)
/seed
Seed: [-3014090999035849675]
/tp @s -14 74 64
/setblock ~ ~ ~ end_portal
Changed the block at -14, 73, 64

The issue is not 100% reproducible. I ran this sequence of commands the first time and the lag occured. The second time it only lagged for about 2 seconds.

I do find world generation in 1.17 to be much slower than 1.16, but I am using the preview datapack so that might be expected.

When the issue occurs, all ticking seems to freeze and chunks do not load in the End for at least 15 seconds. My server does begin working normally again after that.

I'm trying to reproduce it with Timings running and will attach a report once I do so.

embeddedt commented 3 years ago

https://timings.aikar.co/?id=1206f4184dae4917bf91cfc7052a8a85

I'm guessing the "Sync Chunk Load" indicates I've done something very wrong with the configuration. :laughing: I'll investigate that further.

EDIT: sync-chunk-writes is set to false in server.properties. However, I've noticed enable: true is missing from async-chunks in paper.yml. It is there on my 1.16 server. Do I need to add that?

JakeTeeEm commented 3 years ago

I'm having a similar issue on my own server. Whenever you step through an exit portal, the server either crashes, or leaves you there in "waiting for chunk" forever. Had to either restart the server, or even get an alt on to teleport myself out of the glitched area.

MiniDigger commented 3 years ago

[11:51:33] [Server thread/INFO]: Scanning for legacy world dragon fight...

can you try to see if you can repro this if you disable the config option for this?

embeddedt commented 3 years ago

I set scan-for-legacy-ender-dragon: false and I get identical behavior (minus that log message).

e-im commented 3 years ago

Are folks still able to reproduce this on current versions? I haven't been able to on either old or new builds including with the caves and cliffs datapack.