[EMAIL PROTECTED] said:
} I bet you are right on this one.  As it stands, I pushed up my UDP
} timeout to something lame like 5hrs because ICQ was flapping.  It was
} only later that I learn that you can change ICQ's refresh directly.
} I'll try putting my UDP timeout to somthing like 80 seconds and see if
} that helps.  Do you think I should put the timeout even lower? 

It depends... he says helpfully :-)
If you are putting UDP through which needs long timeouts then you need to 
up them....   I have generally run them with short timeouts (2 minutes), 
but our gateways have it configurable....

}       1)      I've seen in the past that a "ipfwadm -M -l" will 
}               show a LOT of old, timing out MASQed connections 
}               (tcp timeout is 5 hrs).  Anyway, much like clearing 
}               IPFWADM's input or output rulesets, how can I 
}               delete all the MASQ entries as shown in 
}               "ipfwadm -M -l"?  Is there something I can cat into 
}               /proc/net/ip_masquerade to clear it out?  
} 
}               I thought that I could just change the UDP timeouts to 
}               1 second and let all the MASQ connections expire.  
}               Unfortunately, it seems that when you change the 
}               timeouts, it will only effect newly setup connections.
}               All the existing UDP MASQ connections still had their
}               original timeouts.

This is a hole in the current stuff - needs an additional API.  Wonder if 
its fixed in 2.2 ??   Maybe either a flush table or apply short timeout to 
table would suffice.


}       2)      When MASQ was dieing, I notice that the ip_masq_quake
}               module was "Used by 169".  This seems awefully high
}               to me but then again, I don't understand what that 
}               number means.  Anyway, When I got that no free udp 
}               ports available error, I tried to unload the quake
}               module only to see that it was "busy".  Is there a
}               way to FORCE modules to unload?  Would this require
}               a change the the rmmod tool?

The large number sounds right.  I don't know a means of a forced unload, 
and it would be better to have a means of asking a module to clean up and 
then using a normal unload (since some modules if removed while in use 
would result in a crash or other infelicity when new data arrived).

        Nigel.

} Thanks for your time Nigel (and anyone else that has some ideas!).

I've just started touting around your TrinityOS document as the answer to 
all people's questions, so thanks to you!


-- 
[ [EMAIL PROTECTED]   -  Systems Software Engineer ]
[ Tel : +44 113 207 6112                   Fax : +44 113 234 6065 ]
[      Real life is but a pale imitation of a Dilbert strip       ]
[ We're recruiting  http://www.theplanet.net/profile/recruit.htm  ]


---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
For daily digest info, email [EMAIL PROTECTED]

Reply via email to