We were seeing something similar yesterday (and yup - it's still there today). We also noted that if you turn on the record-route options, then your problem will disappear (I suspect packet options may cause delivery to be ok - though while the RR option helps, DF doesnt, but "timestamp" will). We're not seeing packet loss in any other scenario. I noticed, suspiciously, that there is always a very "hockey toothed" (every other packet) pattern in the lossage (load balancing onto a botched circuit?). Things I've played with: * size 50% (tried 800, 1200, 2500) * DF 50% * RR 100% * pattern seems to be generally 50% * timestamp 100% * TOS 50% We've opened up a case with cisco ... seems pretty bizarre, but could it be the ATM switch? As long as we're not using the source address of the ATM network - things are fine (e.g. sourcing off of the loopback, going one hop back, etc). - Tom On Fri, 6 Oct 2000, Richard Inhand wrote:
Date: Fri, 6 Oct 2000 04:50:10 -0700 (PDT) From: Richard Inhand <chapinhand@yahoo.com> To: nanog@merit.edu Subject: ATM Question
I'm currently looking into a weird problem at an ATM peering point on the West Coast, when I ping from my peer router to any of my peers I get some but not total packet loss, when I ping from one router back or use the loopback interface as source things are fine. Whereas, with all my other ATM peering point routers, for example at MAE-EAST I can ping my peers cleanly from the same router. This isn't (to my knowledge) affecting traffic just wondering why this strange phenomenon should occur only in the one place.
Any body else had any similar problems?
Richard Hand
__________________________________________________ Do You Yahoo!? Yahoo! Photos - 35mm Quality Prints, Now Get 15 Free! http://photos.yahoo.com/