Thanks for all of the replys :) On Fri, 9 May 2003, Hiddukel wrote:
> Unfortunately I can't be much help with the freeze, but on a side note > about the sighandler code, when you use a sighandler it is very rare > that you will get a core file as the code handles the signal and does a > copyover in your case or exits normally after logging something in other > cases. At first glance this may seem like a good idea, but in the great > scheme of things when you have a bug in your code that causes the mud to > crash it is much more desirable to have it dump a core so that you can > debug it rather than just having the server exit or do a copyover and > trying to log what may have happened. I have also found that for the > most part when you encounter a crash of the type you described with an > invalid character that the sighandler won't handle it correctly because > the copyover recovery will actually try to work with that same invalid > character. > The sighandler forks a new process so that it can dump a core, but you're right that it was most likely trying to use the same invalid character that is_granted_name crashed on. > As far as the freeze goes there could be a number of things not actually > associated with your mud. If your mud's CPU usage was at 0% when it > froze then it is quite possible that some other process on the machine > was looping and taking up all the processor time. If there are other > muds running on the same machine as your mud it is a very good > possibility that one of them looped and caused your mud to seem as > though it was frozen because there was no available processor time for > it. I know that none of this really helps in tracking down the cause, > but it is the best answer I can come up with about your problem. > I wish it was something that simple. I admin my own server, so unless one of my test ports are up (which they weren't), MoF is the only mud running on the machine. The only other things that really run on there are apache mysql, and sendmail, and those processes were running normally as well. That was one of the first things I checked for when it started.. ran so many netstats and tops and ps's that I practically had it all memorized before I finally decided that wasn't the problem :) That's why this is so mind boggling to me! Maybe it's just time to upgrade this server to the latest/greatest 4.x release of freebsd. > -----Original Message----- > From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On Behalf Of Jennifer > King > Sent: Thursday, May 08, 2003 23:32 > To: [email protected] > Subject: Mud process dying a slow death... > > Time for a question dealing with my main port for once, and not the new > one. And this one is totally frying me. > > Rom2.4b4 with b6 patches, 5+ years worth of changes and rewrites.. > recently installed a sighandler.. can't think of who wrote it offhand, > but > it's the one that checks to see if the mud has frozen and does the full > copyover thing and all that. The mud was running fine for 4.5 days > straight, then crashed when is_granted_name tried checking an invalid > character. The sighandler went into copyover.. and then totally crashed. > > Now, here's the big problem. After the mud was restarted again, I was > watching the process in 'top'.. the cpu % would drop to 0.00% and the > mud > would just freeze up there. Usually whenever it would get stuck in a > loop, > the process cpu would skyrocket, but here it is down at nothing, and the > mud is totally frozen. And it does this for at least 2-3 hours > straight.. > every couple minutes. > > Finally, it had another crash, the sighandler caught it, copyovered, and > now it seems to be running stable again. > > Unfortunatly, I have nothing in the logs saying why it froze, and of > course, no core files either.. so I don't know where to start looking. > I also have 2 zombied rom processes. > > Please, someone tell me you have some suggestions or ideas!! > > --- > RogueDragon @ A Merging of Fates MUD > telnet://mud.merging.org:5454 > icq: 2072355 (inactive atm), yim: roguedragon, aim: roguedragon69 > --- > Windows - Where do you want to go today? > Linux - Where do you want to go tomorrow? > FreeBSD - When are they going to catch up? > > -- > ROM mailing list > [email protected] > http://www.rom.org/cgi-bin/mailman/listinfo/rom > > > -- > ROM mailing list > [email protected] > http://www.rom.org/cgi-bin/mailman/listinfo/rom >
