other routing anomalies
For the last week the *ONLY* way I can hold a connection to tigger at jvnc.net has been to log into a net access machine in philly and telnet from there to tigger rather than telnet directly from my house. Both connections go througth the CIX to get me 20 miles up the road in New jersey!! When I get packets from this connection it is like a 1200 baud connect. The infuriating thing is that after giving me 1200 baud service for a minute or two, it will stop for any where from up to 45 seconds to 3 minutes before moving a single character!!!!! I have complained to jvnc. They say they are laboring mightily to solve the problem and that a LOT of their users are affected. Not much luck so far ...still 300 baud service. BUT a traceroute from tigger to a commercial web site in wash DC run by PSI goes through MCI's MAE EAST ++ router which is where *MY* packets oughtta go rather than to California and back. Anyone have any idea why? The person I asked at JVNC today didn't. Here are the two traces done within the last 30 minutes. JVNC To my router via cix traceroute to 205.164.155.1 (205.164.155.3), 30 hops max, 40 byte packets 1 nomad-eth0 (128.121.50.50) 2 ms 2 ms 2 ms 2 liberty-hssi1/0-1 (130.94.40.250) 11 ms 10 ms 10 ms 3 waters-ser1/6 (130.94.45.249) 77 ms * 61 ms 4 ness-ser4 (130.94.17.250) 89 ms 126 ms 116 ms 5 goldengate-ser8 (130.94.15.65) 204 ms 186 ms 163 ms 6 cix-jvnc.west.cix.net (149.20.6.1) 194 ms 152 ms 149 ms 7 * hlc.west.cix.net (149.20.64.32) 109 ms 139 ms 8 hlc-east.gw.hlc.net (205.214.32.1) 128 ms * 159 ms 9 * mae-east.netaxs.com (192.41.177.87) 176 ms 172 ms 10 wynd-mae-east-gw.netaxs.com (207.8.160.5) 374 ms 198 ms 194 ms 11 wynd-mfs/k2ne/goldner-we0.netaxs.com (207.8.186.88) 226 ms 127 ms 160 ms 12 k2ne-wynd-t1.tty0.netaxs.com (206.161.91.14) 133 ms 184 ms 158 ms 13 router-ewing.netaxs.com (205.164.155.1) 167 ms 221 ms 199 ms Traceroute JVNC to comcast web site via mae east: tigger-gcook> traceroute www.comcast.com traceroute to www.comcast.inter.net (38.15.13.172), 30 hops max, 40 byte packets 1 nomad-eth0 (128.121.50.50) 2 ms 4 ms 3 ms 2 liberty-hssi1/0-1 (130.94.40.250) 37 ms 15 ms 13 ms 3 border2-hssi2-0.NewYork.mci.net (204.70.45.9) 23 ms 14 ms 24 ms 4 core1-fddi-1.NewYork.mci.net (204.70.3.17) 13 ms 12 ms 16 ms 5 core1-hssi-2.Washington.mci.net (204.70.1.5) 20 ms 19 ms 19 ms 6 borderx1-fddi0-0.Washington.mci.net (204.70.2.4) 18 ms 22 ms 18 ms 7 mae-east-plusplus.Washington.mci.net (204.70.74.102) 26 ms 19 ms 21 ms 8 * mae-east.psi.net (192.41.177.245) 35 ms 32 ms 9 38.1.2.14 (38.1.2.14) 21 ms 37 ms 28 ms 10 www5a-gw.inter.net (38.15.1.40) 36 ms 25 ms 52 ms 11 www5a.inter.net (38.15.13.2) 31 ms * 28 ms ********************************************************************* Gordon Cook, Editor & Publisher Subscriptions: Individ-ascii $85 The COOK Report on Internet Individ. hard copy $150 431 Greenway Ave, Ewing, NJ 08618 USA Small Corp & Gov't $200 (609) 882-2572 Corporate $350 Internet: cook@cookreport.com Corporate Site Lic. $650 http://pobox.com/cook/ for new COOK Report Glossary of Internet terms *********************************************************************
I don't know anything about how jvnc.net runs their business, but a few weeks ago, they decided to start advertising some of my /19's, along with a bunch of other AGIS routes, effectively taking me and a lot of AGIS customers off-line for several hours. Maybe they're still having routing problems... On Tue, 16 Apr 1996, Gordon Cook wrote:
For the last week the *ONLY* way I can hold a connection to tigger at jvnc.net has been to log into a net access machine in philly and telnet from there to tigger rather than telnet directly from my house. Both connections go througth the CIX to get me 20 miles up the road in New jersey!! When I get packets from this connection it is like a 1200 baud [..] I have complained to jvnc. They say they are laboring mightily to solve the problem and that a LOT of their users are affected. Not much luck so far ...still 300 baud service. BUT a traceroute from tigger to a commercial web site in wash DC run by PSI goes through MCI's MAE EAST ++ router which is where *MY* packets oughtta go rather than to California and back. Anyone have any idea why? The person I asked at JVNC today didn't. Here are the two traces done within the last 30 minutes.
I don't know anything about how jvnc.net runs their business, but a few weeks ago, they decided to start advertising some of my /19's, along with a bunch of other AGIS routes, effectively taking me and a lot of AGIS customers off-line for several hours. Maybe they're still having routing problems...
Ugh, I replied privately to Gordon; didn't realize he sent it to nanog :) JVNC said to him in an earlier message that they were currently defaulting into MCI and the reason that they were sending outgoing data to netaxs through the CIX was because *any* route that showed up at the CIX would be taken since they were just defaulting into MCI. I don't know why they were doing it - perhaps they were hit too hard by MCI backbone problems (which should??? be resolved shortly w/ OC3). We almost had to shut off the backup connectivity we get by having our routes announced @ the CIX because it's a poor route to choose when MAE-East is functioning fine. Anyway, I don't know what kind of routes JVNC uses or what kind of routing they're doing, but I'm hoping they take full routes from MCI again shortly. Avi
Avi wrote Ugh, I replied privately to Gordon; didn't realize he sent it to nanog :) JVNC said to him in an earlier message that they were currently defaulting into MCI....... COOK: let me clear up some uncomfortable ambiguity in avi's use of the pronoun "it"...... my original post to which he replied was to nanog..... I did NOT send his private reply to nanog. In response to Randy: I *KNOW* nanog should be avoided for this sort of thing and I have BELIEVE me I have been trying to get this worked out through JVNC. I have called them everyday beginning a weekago today to try to get a resolution to the problem. they have been courteous, but the answer has alwsays been we are working on it it should be done by the end of the day. unfortunately it doesn't get done. My post last night has resulted in some fresh information that I hope will get the problem fixed. So in that sense going public here could provide the solution.. I will try to avoid any continuation of this thread. BTW the problems as of right now is definitely NOT fixed. :-( ********************************************************************* Gordon Cook, Editor & Publisher Subscriptions: Individ-ascii $85 The COOK Report on Internet Individ. hard copy $150 431 Greenway Ave, Ewing, NJ 08618 USA Small Corp & Gov't $200 (609) 882-2572 Corporate $350 Internet: cook@cookreport.com Corporate Site Lic. $650 http://pobox.com/cook/ for new COOK Report Glossary of Internet terms *********************************************************************
Avi Freedman writes:
Ugh, I replied privately to Gordon; didn't realize he sent it to nanog :)
JVNC said to him in an earlier message that they were currently defaulting into MCI and the reason that they were sending outgoing data to netaxs through the CIX was because *any* route that showed up at the CIX would be taken since they were just defaulting into MCI.
I think that Gordon assumed that I was a GES/JvNCnet employee. In any case, either you or Gordon understood wrong. JvNCnet takes partial routes from CIX. If they took full routes from CIX and only default from MCI, all traffic would go to CIX. What you want is to get the AS for netaxs (and/or hlc?) put into the as path filter list that JvNCnet uses in their router that peers with CIX. Suggest you contact them directly if you care.
I don't know why they were doing it - perhaps they were hit too hard by MCI backbone problems (which should??? be resolved shortly w/ OC3).
Can you say 16M AGS+? And of course not updating the filters to reflect reality as more people have peered at CIX.
Anyway, I don't know what kind of routes JVNC uses or what kind of routing they're doing, but I'm hoping they take full routes from MCI again shortly.
I don't think they ever have. -Steve
Avi Freedman writes:
I think that Gordon assumed that I was a GES/JvNCnet employee. In any case, either you or Gordon understood wrong. JvNCnet takes partial routes from CIX. If they took full routes from CIX and only default from MCI, all traffic would go to CIX. What you want is to get the AS for netaxs (and/or hlc?) put into the as path filter list that JvNCnet uses in their router that peers with CIX. Suggest you contact them directly if you care.
Yes, we want to get them to just not hear _4969_ at the CIX...
I don't know why they were doing it - perhaps they were hit too hard by MCI backbone problems (which should??? be resolved shortly w/ OC3).
Can you say 16M AGS+? And of course not updating the filters to reflect reality as more people have peered at CIX.
Anyway, I don't know what kind of routes JVNC uses or what kind of routing they're doing, but I'm hoping they take full routes from MCI again shortly.
I don't think they ever have.
-Steve
Anyway, sorry this got propagated out into public by Gordon... Avi
how about submitting your problem report to your provider as opposed to nanog? this is a forum for network providers, not users, to try to sort out our problems. don't mean to pick on you specifically, but you should know better. randy
participants (5)
-
Avi Freedman -
Gordon Cook -
Pete Kruckenberg -
randy＠psg.com -
Steven L. Johnson