Date: Thu, 7 Dec 2000 13:30:08 -0500 From: Bill Vermillion <bill@bilver.wjv.com> To: freebsd-isp@freebsd.org Subject: Re: Strange fetchmail/freebsd/verio problem Message-ID: <20001207133007.A1846@wjv.com> In-Reply-To: <20001207112056.H1710@staff.msen.com>; from wayne@staff.msen.com on Thu, Dec 07, 2000 at 11:20:56AM -0500 References: <200012071518.KAA58009@mail.wanlogistics.net> <20001207112056.H1710@staff.msen.com>
next in thread | previous in thread | raw e-mail | index | archive | help
On Thu, Dec 07, 2000 at 11:20:56AM -0500, Michael R. Wayne thus spoke: > On Thu, Dec 07, 2000 at 10:17:59AM -0500, bv@wjv.com wrote: Got 3 mails on this - I'll use this to reply. In my groggy state last night - starter problem, lost my cell phone, just got a dead computer back with new m'board, and then that, made it a BAD day and I was not thinking as clearly as I normally do. > > I maintain some servers on a Level3 backbone. While waiting > > for our DS3 and the ATM to come up so I can get ADSL service > > from 'myself', I've been using my old dialup account on Verio. > > working. > > Anybody have any idea, or what I can turn on to trace this problem. > Well, I suspect either you have a packet loss problem, a maintenance > problem, or a fragmentation problem. > Where does traceroute (both directions) say the problem starts? > It's not impossible that someone is doing maintenance and you are > getting burned, call everyone involved that you buy from. Do the > packets travel the same path (in and out) normally and when the > problem starts? Same path (partway) to your other machine as to > Verio ? I did traceroute last night, but I'm going to recheck again. Hm. Got dropped one on the traceroute from this machine [the Verio dialup] the target machine. I didn't see that last night when I checked. That was as the Washington Verio to L3 Washington. From this machine it's verion in Orlando > Atlanta > McLean Va > Level 3 Wa.DC > Atlanta > Orlando To get my 'stuff' I run an ssh to a machine in Cocoa FL I do some maintenance on. A tracroute to that one is C&W in Cocoa -> Pompano Beach -> Dallas -> Level3 Dallas ->Atlanta -> Orlando The ssh route to Cocoa is Verio Orlando -> Atlanta -> McLean -> Philadelphia -> C&W West Org, Nj, -> C&W Atlanta, -> L3 Atlanta, -> Orlando. > I suspect that it is NOT a FreeBSD issue. I KNOW it is not a FreeBSD issue, but since all the stuff I maintain is FreeBSD based I figured this would be THE place to ask to point me to the proper tools I may have overlooked to debug this. > Uh, stupid question, you DO have every single Ethernet interface > (including L3) set to proper interface speed/duplex and not set to > "Auto", right? I've seen problems like this with mismathces before, > they only show up under load. It's doesnt' appear to L3. But from some other suggestion I tried pinging with setting the packet sizes. I had just pinged before and got no more than 5% packet loss. Not good but I'd seen worse. So I set packet size to 128 and got 36% loss through Verio. Ooh. Tried 500 bytes - packet loss 66%. Go for broke. I tried 1000 byte packets and saw 96% loss. Guess that shows where the problem is. Washington area always seems to be a problem. I'm assuming that's where it is because the loss on the traceroute, and that the routes that work don't use the Verio Wa to L3 link. On the C&W link I got no packet loss until I tried 1000 byte packets. Lost ONE packet with that overgrown size. So I tried the right tool, but without the right parameter. > OT: how has L3 been to deal with and what are the prices like? Well we did some co-location scouting for an old Unix database client of mine before they decided to become a dot-com startup. We had heard of Level 3 but not seen any of it until that trip. One walk inside and we were sold. Uunet looked like poor in comparion [this is the Herndon VA area]. Agis was just moving in - but I liked their layout better than Uunet. After seeing L3 there was no contest. Basically L3 is the 3rd network built by James Crowe. He sold his MFS - Metropolitan Fibre system to the LDDS/Wordlcom/UUNet group, and after a few months there he decided it was time to build a network 'the right way'. Our prices are great. However L3 has stopped colo facilities except in one or two cities. Current pricing is far higher than our price. But they are pricing it differntly so a low-bandwidth user would come out cheaper than our monthly rate, but if they got to what we've contracted for they would be higher. So if we get all our six racks filled we can put racks in the space that colo.com has - about 1/4 mile away - and still be on the L3 network at our pricing - but rack prices are higher. We have to use two telephone companies as the metro area is divided between Sprint and Bell South. We will probably go over to optical as the port charges for DS3, OC3 and OC12 differ from each other by only about $50/month. The interface cards are more expensive, but after the second link the savings on the IDU's make it pay for itself. There is going to be a glut of colo space in this area. An old Costo warehouse is being rebuilt - Colo.com has part of it. Duran <Dugan?> out of Texas is convering and old Penny's store and adding a 3rd floor - all colo/clec space - about 1 block from Bell main. The new AT&T facility is coming up. Time Warner is just coming up. It's sort of frightening. We're just a small ISP [with no dial-ups] and we're building a co-op for a customer, and handling special connectivity. Sort of custom communications. Thanks everyone for those things I had overlooked. Bill -- Bill Vermillion - bv @ wjv . com To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-isp" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20001207133007.A1846>