Suggestions on a PPPoE server with mpd

[Julian Elischer]

Einstein Oliveira wrote:
> Julian Elischer wrote:
>> Einstein Oliveira wrote:
>>>> Einstein Oliveira wrote:
>>>>> Julian Elischer wrote:
>>>>>> Einstein Oliveira wrote:
>>>>>>> Hi all,
>>>>>>>
>>>>>>> In an attempt to change from pppoed + ppp to mpd I'm having some 
>>>>>>> strange behaviour with the later on the same host that usually 
>>>>>>> handles 100+ clients with the former.
>>>>>>>
>>>>>>> Basically it's a i386 FreeBSD 6.3 system with mpd-5.1 and 1 GB 
>>>>>>> Ram. Users can connect without any problem.
>>>>>>>
>>>>>>> After a few minutes a SECOND mpd process is initiated and this 
>>>>>>> one consumes all idle processing of the host.
>>>>>>>
>>>>>>> Whoever is connected before that still remains connected but no 
>>>>>>> new user can get in anymore. Also logging stops being generated.
>>>>>>>
>>>>>>> The only solution is a killall -KILL on mpd.
>>>>>>>
>>>>>>> On another host I have the same configuration working as expected 
>>>>>>> but with less than 50 clients.
>>>>>>>
>>>>>>> Both configurations are the same from mpd.conf.sample.
>>>>>>>
>>>>>>> So I'm looking for any suggestions that somebody has been using 
>>>>>>> to handle that many pppoe connections (100+) with mpd.
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>> Thanks,
>>>>>>>
>>>>>>
>>>>>>
>>>>>> people have handled many more than 100 I think..
>>>>>> however there is no reason that a 2nd should start up.
>>>>>> what is teh PPID of the 2nd?
>>>>>>
>>>>>> are you sure that it isn't being started up by whatever starts up 
>>>>>> the first?
>>>>>>
>>>>>>
>>>>>
>>>>> The second process has a bigger PID than the first and I'm using 
>>>>> the default /usr/local/etc/rc.d/mpd5 as startup script with 
>>>>> mpd_enable="YES" on /etc/rc.conf.
>>>>
>>>> what  about the PPID?  (Parent PID)  ps alx shows this.
>>>>
>>>>>
>>>
>>> here is the output of ps -alx:
>>>
>>>     0 12573     1   0  20  0  5424  3468 kserel Ss    ??    0:00.15
>>> /usr/local/sbin/mpd5 -p /va
>>>     0 12629 12573 283 131  0  5424  3452 -      R     ??    0:36.94
>>> /usr/local/sbin/mpd5 -p /va
>>>     0 12707 11713   0  96  0  1528   868 -      R+    p0    0:00.00
>>> grep mpd5
>>>
>>>
>> sorry I repeated myself..
>> hmmm kserel... it's threaded? I didn't know that..
>> does top -H show a bunch of threads?
> 
> 12720 root      96    0  8252K  5656K select   0:26  0.00% mpd5
> 12720 root      20    0  8252K  5656K kserel   0:26  0.00% mpd5
> 12720 root      20    0  8252K  5656K ksesig   0:26  0.00% mpd5
> 
>>
>> So the first copy of mpd has spawned another while threaded.
>> have you applied the patch for 6.3 in the 6.3 errata notifications?
>>
>> http://security.freebsd.org/advisories/FreeBSD-EN-08:01.libpthread.asc
>>
>>
> 
> Not yet. This host is using the RELENG_6_3 tag but I can do that if you 
> think it can help.

check to see if the patch is applied. If not it should be because it 
will screw any process that is threaded and does forks.
If that does not help we can look further, but first make sure this is 
done.




>