Problem 100% CPU then crash, only certain chunks

[Piercy]

• OS: Centos
• VPS with oplink
• Direwolf20 (not sure what version, where do i find the numbers? updated it on 4 June 2013 to try fix these issues)
• No Additional Mods added
• 4.8Ghz 8 core, 4gb ram 80gb HDD
• Less than 10 players, whitelisted private server

Hi All,
I am having an issue with my Direwolf server crashing. When the server crashes it hits 100% cpu on one of the cores and doesnt show any signs of coming back to life, just sits at 99-100%. The crash seems to occur when a player access a certain area in the world. We can reproducibly go to this area of land and crash the server. I wondered if it was an entity issue but scanning with region fixer didnt show anything in particular (111k entities total, but did not indicate any chunks). The issue happens with both mcpc+ and ftbserver jar files.

I installed clearlagg but that again doesnt indicate anything.

Here are the most recent crash logs, I couldnt find anything useful in them, but maybe you can. Any help is highly appreciated i really want to get this solved. Also, are there any good tools to try help identify issues like this?

Thanks,
Piercy

Logs:
http://pastebin.com/4Xxta9ah
http://pastebin.com/r1SLNdQQ

 

[dariusc93]

The issue here look like its related to the latest version of MCPC, unless it can be confirmed by someone using pure vanilla forge mods. You may want to remove the forcechunks.dat and see if that helps any. Also make sure your using the latest version of direwolf because i know their was a issue with one of the old versions of which had issues with chickenchunks. If a player enters in before the crash, get their location and investigate. Remove any chunk loaders and xycraft blocks there, and if you need to remove the chunk or region manually and work on it from there. You may also want to update your java startup script so that it can use more than one core. As for the entities issue, use worldguard /stoplag to halt all server activites to remove all entities (warning it remove all entites - mobs and items). and then do /stoplag -c . You may want to get a java profiler (or install Tickthreading) and see what dims are affected the most and by what. You may want to set up a profiler on the server like VisualVM (if i said that correct) and use that to see what processes/class is being used the most of the server.

 

[Piercy]

The issue here look like its related to the latest version of MCPC, unless it can be confirmed by someone using pure vanilla forge mods.

The only contradiction on this is that it happens on the standard ftbserver.jar. However, it could be different issues on both jar's and just coincidence.

You may want to remove the forcechunks.dat and see if that helps any.

Will do, i presume this is what stored the chunk loaders / loaded chunk info?

Also make sure your using the latest version of direwolf because i know their was a issue with one of the old versions of which had issues with chickenchunks. If a player enters in before the crash, get their location and investigate. Remove any chunk loaders and xycraft blocks there, and if you need to remove the chunk or region manually and work on it from there.

Will see how I get on, but the manually removing stuff sounds like a decent way to investigate it. As for versioning is there an official place to find version numbers and stuff? At the moment im just "grabbing the latest" with no idea what version that is.

You may also want to update your java startup script so that it can use more than one core. As for the entities issue, use worldguard /stoplag to halt all server activites to remove all entities (warning it remove all entites - mobs and items). and then do /stoplag -c . You may want to get a java profiler (or install Tickthreading) and see what dims are affected the most and by what. You may want to set up a profiler on the server like VisualVM (if i said that correct) and use that to see what processes/class is being used the most of the server.

Ill check out the startup script thing, i thought it was already using multiple cores. I have installed Tickthreading, and it does recover from the crash which is very handy.

Thanks for your detailed answer, ill do some digging and see what i come up with.