View Issue Details
|ID||Project||Category||View Status||Date Submitted||Last Update|
|0003238||unreal||ircd||public||2007-02-16 04:21||2007-02-16 08:18|
|Platform||i686, x86_64||OS||Linux||OS Version||2.4.27, 2.6.18|
|Fixed in Version||3.2.6|
|Summary||0003238: rehash causes ircd to open new fd for modules|
|Description||At each rehash, ircd re-opens each module, causing the number of open files to increase and eventually reach whatever limit was set (1024 here), causing the ircd to refuse any further connections.|
|Steps To Reproduce||Rehash a running server|
|Additional Information||Reproduced on 4 different ircd's. Two on a 2.6.18 x86_64, one on a 2.4.27 i686, and one on a 2.6.10 i686.|
|Tags||No tags attached.|
|3rd party modules|
||hmm, so if you rehash a LOT of times it will eventually reach the limit ?|
Yeah, a lot. I came across an ircd that had rehashed 507 times. This network wanted a list of active channels in the motd updated every hour.
Yes. Hourly rehash. Yuck.
||lol, well, they should use /rehash -motd to do it, but thats besides the point, this is definetly a big issue|
Does "/rehash -motd" reload all modules? I don't think so.
Edit: I didn't see the post above...
Strange nobody noticed this before, but I can confirm it on my machine.
The odd thing is I've heard of people that rehash their ircd every 5 minutes (yeah crazy cronjob), hmm :P.
Perhaps this is new in 3.2.6?
||Definitely agree that "/rehash -motd" should be used instead. However in this particular case, rehash was achieved via kill -HUP, since it was a cron job.|
int file_exists(char* file)
fd = fopen(file, "r");
Fixed in CVS (32* and 33*):
- Fixed file descriptor leakage on rehash. This resulted in [number of modules loaded]
file descriptors being leaked upon every /REHASH.
So if you, for example, had 3 modules loaded and rehashed 30 times, it would cause
the ircd to consume 90 useless file descriptors (which often means 90 less file
descriptors being available to clients).
Was indeed 3.2.6 only, cause was:
"- loadmodule now reports proper errors when the actual file can't be found, instead of blaming
it on the temp file, reported in 0003015.
|2007-02-16 04:21||Le_Coyote||New Issue|
|2007-02-16 05:35||djGrrr||Note Added: 0013220|
|2007-02-16 05:49||Le_Coyote||Note Added: 0013221|
|2007-02-16 06:37||djGrrr||Note Added: 0013222|
|2007-02-16 07:09||Grunt||Note Added: 0013223|
|2007-02-16 07:10||Grunt||Note Edited: 0013223|
|2007-02-16 07:57||syzop||Status||new => confirmed|
|2007-02-16 08:01||syzop||Note Added: 0013224|
|2007-02-16 08:04||Le_Coyote||Note Added: 0013225|
|2007-02-16 08:04||syzop||Note Added: 0013226|
|2007-02-16 08:18||syzop||Status||confirmed => resolved|
|2007-02-16 08:18||syzop||Fixed in Version||=> 3.2.6|
|2007-02-16 08:18||syzop||Resolution||open => fixed|
|2007-02-16 08:18||syzop||Assigned To||=> syzop|
|2007-02-16 08:18||syzop||Note Added: 0013227|