We ran a session tonight using 1.3.b76 with the server running on Mandriva 2010.0 under
java version "1.6.0_21"
Java(TM) SE Runtime Environment (build 1.6.0_21-b06)
Java HotSpot(TM) Server VM (build 17.0-b16, mixed mode)
Running with 4m stack and fixed at 1024m heap.
We were using the latest version of the Rumble framework. First of all clients seem to have a difficult time connecting properly on this version. Players would connect and show as connected but nothing they would type would appear on anyone else's chat window. Usually this would persist until the player rebooted their machine entirely. People can always connect, but they just end up constantly in this 'ghost' state.
Finally after about 2 hours of playing through a big fight on one map the server simply crashed the JVM completely.
I'd like to note that playing last week with 1.3.b75 produced nearly identical results, lots of ghost connections and JVM crash after about 3-4 hours.
In both cases we weren't doing anything unusual, the action stuck to a single map. There was heavy use of Rumble's v5 macros.
I'll try tomorrow to see if I can reproduce the ghost logins. The server crashes are pretty hard to reliably repeat as they seem to be coming up after a few hours only, but we'll run some sessions from the console and see what happens. Perhaps I can dig up a core dump of the JVM as well if anyone has an interest in analyzing it.
Dunno whether to blame the macro framework or the last 2 versions of Maptool, but something seems to be unstable these days. b70 was pretty much rock solid, at least it didn't seem to have network/core dump issues.
[1.3.b76] Linux networking issues and server crashes
Moderators: dorpond, trevor, Azhrei, giliath, jay, Mr.Ice, MapTool BugReport Manager
Forum rules
Posts that do not conform to the READ ME FIRST sticky thread are subject to deletion.
Posts that do not conform to the READ ME FIRST sticky thread are subject to deletion.
Re: [1.3.b76] Linux networking issues and server crashes
There should be a jconsole command as part of your JRE. It's a debugging console that can connect to the running JVM and ask it questions about the configuration.
You might try starting MapTool and connecting to it with jconsole. Then load a campaign and watch what changes in jconsole. Your goal is to become familiar with what the charts and graphs look like when there *isn't* a problem so you'll recognize an issue when there *is*!
Core dumps won't help much unless we have an exact duplicate of your environment as shared libraries will be allocating memory different, etc. But if there's a log.txt that would be useful. And you can read the notes in the XML files that are part of the ZIP for information on how to turn on debugging output. Note that turning on ALL of the debugging will produce a *huge* amount of log output, so there are some you may not want (like the macro execution) or you may want to reduce the amount of debugging output (visit apache.com and click the links for the Log4J project as that's what MapTool is using -- you can find additional information on how to configure the XML files).
HTH.
You might try starting MapTool and connecting to it with jconsole. Then load a campaign and watch what changes in jconsole. Your goal is to become familiar with what the charts and graphs look like when there *isn't* a problem so you'll recognize an issue when there *is*!
Core dumps won't help much unless we have an exact duplicate of your environment as shared libraries will be allocating memory different, etc. But if there's a log.txt that would be useful. And you can read the notes in the XML files that are part of the ZIP for information on how to turn on debugging output. Note that turning on ALL of the debugging will produce a *huge* amount of log output, so there are some you may not want (like the macro execution) or you may want to reduce the amount of debugging output (visit apache.com and click the links for the Log4J project as that's what MapTool is using -- you can find additional information on how to configure the XML files).
HTH.
Re: [1.3.b76] Linux networking issues and server crashes
Ah, just like my day job 
. You want some profiling with that? hehe. I'll see what I can come up with. It may take a week or two...
As for the people connecting but nobody can see what they type, this seems to be a really frequent occurrence. Pretty much 1 out of any 4 client logins it seems like. Also only seems to happen with the 3rd or 4th person connecting generally speaking. Usually 4 or 5 attempts will rectify it.
. You want some profiling with that? hehe. I'll see what I can come up with. It may take a week or two...As for the people connecting but nobody can see what they type, this seems to be a really frequent occurrence. Pretty much 1 out of any 4 client logins it seems like. Also only seems to happen with the 3rd or 4th person connecting generally speaking. Usually 4 or 5 attempts will rectify it.
Re: [1.3.b76] Linux networking issues and server crashes
Well, we know that the same MapTool JARs work on other operating systems. So that leaves the JVM or the shared libraries on the system. Assuming you grabbed the JRE from Sun/Oracle that won't be the problem either. Which leaves just one thing: the operating system platform.
All I can recommend without other debugging information is to make sure your libraries are all up to date. What is the command on Mandriva -- urpmi update? It's been so long I don't remember anymore...
All I can recommend without other debugging information is to make sure your libraries are all up to date. What is the command on Mandriva -- urpmi update? It's been so long I don't remember anymore...
Re: [1.3.b76] Linux networking issues and server crashes
Yeah, we are in a session tonight, it is just rough. I am all up to date. We ARE running MDK2010.0 x86_64. Dunno if anyone has any thoughts on 64 bit. JVM is standard sun 1.6.0_21 for 64 bit linux. I have run a LOT of java stuff on this machine, so I'd have to guess if it is a native library issue it is an odd one. Looks like we're going to call this week's session early, I had 3 out of 5 players not connecting reliably at least part of the time. Will crank up the L4J logging and some of us will test tomorrow. Forgot to do that this session.Azhrei wrote:Well, we know that the same MapTool JARs work on other operating systems. So that leaves the JVM or the shared libraries on the system. Assuming you grabbed the JRE from Sun/Oracle that won't be the problem either. Which leaves just one thing: the operating system platform.
All I can recommend without other debugging information is to make sure your libraries are all up to date. What is the command on Mandriva -- urpmi update? It's been so long I don't remember anymore...