Date: Sun, 13 Feb 2005 08:53:20 +0300 From: "Andrew P." <infofarmer@mail.ru> To: Dan Nelson <dnelson@allantgroup.com> Cc: FreeBSD-Questions <freebsd-questions@freebsd.org> Subject: Re: Concealing short disconnects Message-ID: <420EEB50.7010304@mail.ru> In-Reply-To: <20050212074138.GD49626@dan.emsphone.com> References: <420D7EE3.5000305@mail.ru> <20050212040327.GA49626@dan.emsphone.com> <420D8177.30600@mail.ru> <20050212074138.GD49626@dan.emsphone.com>
next in thread | previous in thread | raw e-mail | index | archive | help
Dan Nelson wrote: > In the last episode (Feb 12), Andrew P. said: > >>Dan Nelson wrote: >> >>>In the last episode (Feb 12), Andrew P. said: >>> >>>>I have a few machines behind my FreeBSD box. The box connects to >>>>ISP via ppp (PPPoE protocol). It's all working very nicely, but the >>>>ISP is a pain - it disconnects every 24 hours. I can reconnect in >>>>just a moment - so the diconnect is usually less than a second >>>>long, but many applications, like ICQ/MSN and games "feel" the >>>>disconnect. The matter is that these applications can handle fairly >>>>large packet loss (e.g. Counter-Strike can cope with at least >>>>15-second long 100% packet loss), but AFAIK it's in the nature of >>>>the TCP/UDP that a disconnect is a disconnect. >>>> >>>>As I know that FreeBSD is full of magic, is there any way to >>>>conceal these reconnects as short moments of 100% packet loss? I am >>>>ashamed to know very little about protocols' technicalities, but >>>>I'll look into any sources you advise. >>> >>>Check to see if your IP number changes when you reconnect. If it >>>does, there's nothing you really can do; the remote system you were >>>talking to knew you only by your old IP, and those packets coming to >>>them from this other IP are unrelated. >> >>It changes only once in about a week. Let's say it doesn't change >>at all. What then? > > > I'm still suspicious :) The two most common causes for connection > resets are IP address changes and NAT resets. /usr/sbin/ppp keeps its > NAT table across disconnects as long as the process itself stays > running, so I don't think that's the cause. If you have root access to > a remote system, try running tcpdump on it and your local machine while > running something like top over ssh, and watch what happens when your > connection drops and reconnects. > No, there's really nothing to be suspicious about :) The IP doesn't change (well, in the process of IPCP it virtually does, first to 10.0.0.1/0 and then back to the assigned one - but that doesn't count, does it), the ppp process stays, but TCP/UDP streams are somehow interrupted. Don't worry anyway. Disconnects happen in 5-6 in the morning, when all the users are sleeping and the only one sleepless surfer is unlucky me, trying to seamlessly upgrade self-made internet connection sharing box from 4.10 to 5.3. BTW, if only anyone happens to know: I asked list before, but got no reply. When ISP actually assigns new IP address, I occasionally get double IPs on the tun0 interface (the old one and the new one simultaneously). Everything's working fine, but the dyndns updater can't recognize the IP change. Is there a way to fix this glitch/ feature? I've really manned and googled for it - without succes. Best wishes, Andrew P.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?420EEB50.7010304>