[j-nsp] Layer3 incompletes increasing

[Josef Buchsteiner]

Wednesday, May 11, 2005, 5:53:32 PM, you wrote:

  
AK> On Wed, 11 May 2005 17:41:37 +0200, Niels Bakker wrote:
AK>  [..]
 >> None of those are happening on my LAN.  I've sent JTAC pcap files from a
 >> traffic analyser right between the switch and the router but they don't
 >> seem to be able to figure it out either.

AK>  Incidentally I see the L3 incompletes on a certain IX port
AK>  in Amsterdam as well. So that makes me even more clueless,
AK>  as I understand I cannot catch them on the Juniper as the
AK>  throwing- away is taking place before I can capture them.

AK>  But then, it is definitely between 6.3R2 and 6.3-20050114.0
AK>  so it cannot be that hard to trace down...?!

     We had a symptom where we reported framing errors due to runt
     frames and L3 incomplete counters were decreasing. This was
     fixed in 6.3R3+

     Ethernet interfaces are operating in cut-through mode and
     therefor a DA reject in the PIC can result in L3 incomplete
     and therefor the assumption was that each DA reject will end up
     in L3 incomplete which may not be true anymore for all cases so
     this has been removed.

     By doing this change you will see now much more L3 incomplete
     now since we do not subtract for each DA reject the L3 incomplete
     counter

     I know it is not nice but the best we could do and explains why
     you see a higher rate of L3 incomplete.

     Josef






 >> What I've found out so far is that nobody in the world really knows when
 >> or why the "L3 Incompletes" counter increases; that it's 8-bits in the
 >> hardware; but also that it and the "DA Rejects" counter step in close
 >> harmony (at least in my case), which I'm still investigating.

AK>  Yes, the 'DA Rejects' are increasing as well, I noticed that
AK>  as well.

AK>  Alexander

AK>  _______________________________________________
AK>  juniper-nsp mailing list juniper-nsp at puck.nether.net
AK> http://puck.nether.net/mailman/listinfo/juniper-nsp