Server Crash - MCPC+

[oryfoxer7410]

Haven't had any issues, running mcpc+ w/Tekkit 1.2.8e. Plugins - Essentials Spawn, Essentials group manager, essentials, prism (recommended mcpc 1.6.4 build). Server runs fine, resources looks good, but I keep seeing this error every 5-6 hours. At times, it will cause a crash. It just started out of nowhere, no recent changes either. 

 

http://pastebin.com/hYWGa3qX

 

 

 

http://pastebin.com/djHYrUYS

 

 

Wanted to see thoughts.

[plowmanplow]

These errors look a bit random. Unless you can repeat some action or setup that makes them happen consistently I would throw out a wild guess that you may be running low on PermGen or Heap. I've seen similar-ish errors that are resolved by increasing these. That is, of course, simply an anecdotal observation and the real source of the problem may be only peripherally realated. Give that a shot and see where it gets you. Can you put your startup line with arguments here?

[oryfoxer7410]

so, i got the same error again last night, crashed the server again.

 

http://pastebin.com/96fCf4kT

 

I have already set my permgen to 128, should i make it higher? Memory is good, CPU is good. I have a dedicated server, just running Tekkit with XEON X3-1270 & 32 GB of RAM. Running Ubuntu.

[oryfoxer7410]

java -XX:MaxPermSize=128m -Xmx30G -Xms25G -jar mcpc.jar

[plowmanplow]

Memory balance can sometimes be odd and non-intuitive. It is possible to allocate too much RAM to the server process. In the case of TekkitMain (and I'm just speculating here based on number of mods and such) it is possible that assigning about 94% of your total RAM to the server process may be too much. The OS needs a bit of memory as well and depending on how many concurrent users you have (you haven't specified) it may need a little more.

 

Some anecdotal data from one of my servers:

 

Physical RAM (well, virtual, but same thing in this context): 8G

RAM allocated to server process: 4G

Number of mods reported: 139 (big pack, lots of worldgen)

Concurrent players: <=8

PermSize: 128m

Free memory when no players connected: 2.2G (so OS is taking 1.8 and that escalates when folks are active)

 

There are a number of command line options you may be able to use which could improve your performance for GC and other Java parameters. This command line may be quite old and some of the options may have better alternatives now, but here's my command line which works well for me (open to suggestions).

/usr/local/java/bin/java -Xmx4096M -Xms4096M -XX:PermSize=128m -XX:MaxPermSize=128m -XX:+DisableExplicitGC -XX:+UseConcMarkSweepGC -XX:+UseParNewGC -XX:+UseNUMA -XX:+CMSParallelRemarkEnabled -XX:MaxGCPauseMillis=50 -XX:+UseAdaptiveGCBoundary -XX:-UseGCOverheadLimit -XX:+UseBiasedLocking -XX:SurvivorRatio=8 -XX:TargetSurvivorRatio=90 -XX:MaxTenuringThreshold=15 -XX:UseSSE=3 -XX:+UseLargePages -XX:+UseFastAccessorMethods -XX:+UseStringCache -XX:+UseCompressedOops -XX:+OptimizeStringConcat -XX:+AggressiveOpts -jar mcpcplus.jar nogui

[oryfoxer7410]

So, I added the arguments you are using to my startup script and also lowered my max and min RAM to 20 GB. This leaves 12 GB of the OS. This will serve as a test. If I am stable, I will gradually increase from there. If not, I will let you know. 

 

Thanks for the help. This just came out of nowhere, I really hope this is gone. 

[oryfoxer7410]

So, it crashed again, even with the arguments you provided. I am not sure what's causing the issue, but I removed Prism for now as  a test. What's hard to swallow is we have been stable for so long, then out of nowhere, these crashes start.... 

 

http://pastebin.com/9VUXDaGD

[plowmanplow]

How many people are logged in when this last one happened? What are they doing? The log is showing that you have almost 7000 chunks and 1600 entities across your dimensions. What is your view distance set to?

[oryfoxer7410]

view distance is currently at 10. Also, we had about 7 people logged on. We average around 10, but max around 15.

[oryfoxer7410]

As far as what they are doing, just normal stuff. The log shows a /home before... Nothing that stricks as a "smoking gun"

[oryfoxer7410]

I'll send you the complete log in a pm if you want.

[plowmanplow]

Something seems "off". I have to assume that someone (or multiple someones) are using chunk loaders to load massive numbers of chunks. Let's say there are 7 people and a server view distance of 10. That's 3087 chunks if there is no overlap at all. Add in a bit for spawn and you are around 3200 chunks. That's almost 4000 chunks which show as loaded beyond the ones for players unless I've grossly miscalculated something. That is a LOT of chunks.

[oryfoxer7410]

chunk loaders are in use on the server and have always been. Our goal is to have a default Tekkit Server. That's the idea and its worked fine, until the last day or so... 

 

I think I found the issue, once i removed prism, the issue seems to have gone away. I'm not seeing errors like I was before. 

 

But yea, we are a private server and we do allow chunk loaders, but even we a large world, we shouldn't be having crashes or these tick loops... 

 

So, I honestly think that Prism was causing the issue. I'll update you if I find out more. 

[plowmanplow]

I use CoreProtect on all of mine. Seems to work well.

[oryfoxer7410]

I'll check that out! Thanks!

[oryfoxer7410]

BTW, I had to heavily configure Prism to work well with Tekkit, Can I get a pastebin of your coreprotect config? I really don't want to have to go through the testing and fixing stage with another plugin... 

[oryfoxer7410]

Thanks so much plowman, CoreProtect is awesome. It just works, which is awesome, there was no configuration beyond my MySQL db setup... 

 

still running error free, so now confirmed it was Prism...