search

MinecraftPortCentral / Cauldron-Issues

Issue tracker for Cauldron - http://cauldron.minecraftforge.net
13 stars 6 forks source link

60-120 sec lag spikes (seeking for experienced opinion, prob. non cauldron related) #112

Open Slind14 opened 9 years ago

Slind14 commented 9 years ago

Hey,

I don't think this is related to cauldron but if possible I would like to get your opinion on this blood. We have the issue with the Modsauce and a custom Modpack, the server freezes at random times for about 60-120 seconds. The server thread is sleeping/waiting during this period. Thread-dumps show that the File IO Thread is blocking. Unfortunately I have no idea what could cause this for such a long time, especially as it is far away from any autosave/worldsave.

I would be glad about any ideas and possible further investigation possibilities on this topic.

https://gist.github.com/Slind14/a60a259749d5e140a05b#file-gistfile1-txt-L1000

InsaneJ commented 9 years ago

This sounds similar to the issue we're experiencing: https://github.com/MinecraftPortCentral/Cauldron-Issues/issues/107

jules103 commented 9 years ago

do you guys want to check something for me does your custom server has EE3 (equivalent exchange)?

jules103 commented 9 years ago

i removed EE3 from my modsauce server, i ran opis 4times one of them it showed a really high number of ping, i just tried it the freeze seems fixed for me now.

jules103 commented 9 years ago

let me know if it worked for you guys!

Slind14 commented 9 years ago

No EE3 on the custom pack.

InsaneJ commented 9 years ago

No EE3 on our server, only TFC.

bloodmc commented 9 years ago

@Slind14 Run WarmRoast https://github.com/sk89q/warmroast then provide screenshots.

Slind14 commented 9 years ago

@bloodmc I did run visual vm and during multiple 1-2 minutes lag spikes I did create multiple thread dumps. They all look the same. It is not the server thread as it is waiting. One of these full thread dumps is posted above.

bloodmc commented 9 years ago

@Slind14 at net.minecraft.world.chunk.storage.AnvilChunkLoader.func_75821_a(AnvilChunkLoader.java:287)

Your server is locking up writing NBT data to your region files. Are you using an SSD?

Slind14 commented 9 years ago

Thanks this confirms what I figured, due to some of the timeouts being during a world unload, the thing I don't understand is that I can run 4 1.6 Servers with 4 times the players on one machine and two 1.7 server time out on another machine. (all worlds are pre generated) So the thing I'm wondering is why does it take that bloody long with these packs. I did profile the hard drives of both boxes and they reach about the same write/read IO.