Descriptive Problem Summary:
For some odd reason whenever my servers are on a BETA version they just go haywire after a while. There's no bottlenecking or anything that is being experienced. However, when I perform top and/or htop I can see that some games are capping out on their CPU. The server controller (master) hits 99% and stays there. Sometimes the childs also hit their CPU caps, whatever it is allocated to be. Why it does this, I have no clue. However, it only happens to BETA. Well, scratch that. It's been happening to Stable recently as well. But, it happens 80% less on Stable. Stable is... literally stable. Like it should be. But, BETA has it's issues with staying up for a long time.
Numbered Steps to Reproduce Problem:
Leave a game up for days without shutting it down. It may happen after an hour of uptime.
Does the problem occur:
Every time? Or how often? Once or more every 3 days. It does not last 3 days without doing it.
In other games? Yes. Multiple games are hosted on the same server box and they will do it occasionally as well.
On other computers? Yes, I've had other server boxes on the BETA version do this as well. My Server #1 was on BETA but had to be removed since a lot of games were just out of the blue capping out the CPU. I instead designated an individual server for the Web Client version.
When does the problem NOT occur?
When running a stable version. Note, this has been occuring before 507 as well. As long as it's beta, it's unstable.
Workarounds:
Run a stable version or just shutdown and start up the game again. A reboot doesn't fix it.
---
Next time the problem occurs, since I literally just reset the DreamDaemon for the problem instance, I will take a screenshot and run a SIGUSR2 and SIGSEGV since it needs to be shutdown anyways. I'll paste those in a second post for this thread.
ID:1737321
 Dec 11 2014, 9:30 pm
|
|||||||||||||
| |||||||||||||
|
Dec 12 2014, 7:05 am
|
|
Lummox JR
  |
We'll definitely need more specific info on this. I don't see any reason the beta should be less stable than 506, since 507 has seen a number of significant fixes. However this is obviously not an issue for all games, and is likely specific to SS13 or even a particular branch of it.
|
 Backtrace for BYOND 507.1267 on Linux: | |
Haven't experienced this with my SS13 server, uptime around 1-2 weeks on Linux.
I'm assuming Xirre has checked his own code to see if it wasn't his own problem..? | |
Laser50 wrote:
Haven't experienced this with my SS13 server, uptime around 1-2 weeks on Linux. I honestly don't feel as if it has anything to do with SS13. I've reviewed several pieces of my code multipel times. The only fancy things in my code are shell() calls, javascript for file management, and datums for information handling. Other than that, it's the same thing everyone else is doing. The basics in BYOND. Ifs, fors, Stat(), etc. :) Uhm... also, let it be known that it's also games that are freshly made that are effected as well. One known to crash would be Ralf1324's game, Sekia. It does it even on Stable. Beta just makes it happen 10x more frequent. | |
Well, ignore this and the one above unless you feel they hold some clue to something else. What's going on here is that shell() is calling unzip on the linux system and looking for files of type .sav, .txt, and .log. Which is fine for it to say "caution". What comes after it is what is buggy. The server is in DEBUG mode... Seeing as how there are a lot of "Process XXX dead!" messages, which means the game was shutdown through the server or through the game itself or... it just crashed. This also means that the game was constantly started up with shell() calling DreamDaemon. When that happens, it also calls shell() to limit DreamDaemon's CPU. Then, it goes to getServer_CPU on the new list of servers and that proc calls getCPU_Usage on whatever iteration it is on. It also checks if it exists. Which is what confuses me as to why it would crash.
I'll try to find a bug report for one actually pertaining to this error. It's just so odd because this is recent. The previous high CPU-consumption errors didn't hang the game. It just displayed a high CPU outside of the game while the inside, world.cpu, said it was at 1% or 0% all the time.  caution: filename not matched: *.sav | |
Doesn't seem like tonight is a good night. These bug reports may not be helpful, but, I wouldn't know that 100% as I've never looked deep in to them. Nor do I think I can.
Here's 3 more culprits on a stable version (I forgot to screenshot one, that's the Shell Server Utility):  Tomeno's server for "Console", by Exadv1 runtime error: type mismatch: "\"extern 0 0 chat \[nick]@" += /list (/list) Shinx's game, Warrior Cats. Sadly, I couldn't grab the error report in time. Shell Server Utility -- Does it just say crashing due to illegal operation everytime I sigsegv or try to sigusr2 when it's hanging? Process 14437 dead! | |