The MPLS WG Archive

Cell Relay Retreat>MPLS WG Archive>month:2004-Apr> msg00051



[Date Prev][Date Next][Thread Prev][Thread Next]  
  [Date Index][Thread Index][Author Index][Subject Index]

MPLS Ping question

  • From: "David Allan" <dallan@nortelnetworks.com>
  • Date: Thu, 8 Apr 2004 12:46:33 -0400
  • Cc: mpls@UU.NET

IMHO the address should always be as precise as possible in a traceroute, so
the one used would be independent of the presence of the downstream mapping
TLV. If there is an interface address it should be used...

I would agree that some clarification in the draft would be useful, hope the
authors agree..

Dave

> -----Original Message-----
> From: Igor Achkinazi [mailto:achkinazi@hotmail.com] 
> Sent: Thursday, April 08, 2004 12:38 PM
> To: Allan, David [CAR:NS00:EXCH]
> Cc: mpls@UU.NET
> Subject: RE: MPLS Ping question
> 
> 
> Dave,
> 
> Section "4.4. Sending an MPLS Echo Reply" of 
> draft-ietf-mpls-lsp-ping-05.txt 
> states: "The source IP address is a routable address of the replier".
> 
> It should say something like this: "In case Downstream Mapping TLV is 
> present in incoming MPLS Echo Request, the source IP address 
> must be equal 
> to IP address of an interface from that TLV, which belongs to 
> the replier. 
> If Downstream Mapping TLV is not present in MPLS Echo 
> Request, the source IP 
> address is any routable address of the replier."
> 
> Such clarification will be much better than including of 
> router loopback 
> address into MPLS Ping packet.
> 
> Thank you,
> Igor
> 
> 
> 
> 
> >From: "David Allan" <dallan@nortelnetworks.com>
> >To: "'Igor Achkinazi'" <achkinazi@hotmail.com>
> >CC: mpls@UU.NET
> >Subject: RE: MPLS Ping question
> >Date: Thu, 8 Apr 2004 12:14:36 -0400
> >
> >What you are describing is an implementation problem, the 
> ping request 
> >should be processed (and reply generated) at interface Intf3 
> so even if 
> >the Intf4/Intf2 link is used, the reply source address 
> should be Intf3.
> >
> >IMO there are other problems introduced if the router 
> loopback address 
> >is included in lieu of numbered interfaces. It would hide subtle 
> >problems (such as wrong link of a link bundle between two LSRs being 
> >used).
> >
> >Dave
> >
> > > -----Original Message-----
> > > From: Igor Achkinazi [mailto:achkinazi@hotmail.com]
> > > Sent: Thursday, April 08, 2004 11:21 AM
> > > To: Allan, David [CAR:NS00:EXCH]
> > > Cc: mpls@UU.NET
> > > Subject: RE: MPLS Ping question
> > >
> > >
> > > Dave,
> > >
> > > I understand that in described case a source IP address 
> of MPLS Echo 
> > > Reply packet will not match Downstream Mapping TLV of previous
> > > probe. There is no
> > > easy way to find which IP address belongs to which router, so the
> > > implementation can compare only IP address of reply packet
> > > with IP addresses
> > > from Downstream Mapping TLV of previous step of the traceroute.
> > >
> > > In general case though, IP address of Echo Reply does not need to 
> > > match IP addresses from Downstream Mapping TLV of previous probe,
> > > because there can
> > > be parallel links where one is used for MPLS traffic and
> > > another is used for
> > > IP traffic (and MPLS Echo Reply is IP packet, not MPLS):
> > >
> > >     Intf1<----MPLS--->Intf3
> > > LSR A                     LSR B
> > >     Intf2<----IP----->Intf4
> > >
> > > where LSR B is a downstream for LSR A.
> > >
> > > In this case Downstream Mapping TLV will have Intf3 IP 
> address, but 
> > > IP address from Echo Reply will be equal to Intf4 IP address.
> > >
> > > May I suggest including Router ID into MPLS Ping packet and into 
> > > Downstream Mapping TLV, one per downstream interface, so this 
> > > confusion will be unambiguously resolved? This way we will always 
> > > know which router is sending
> > > MPLS Echo Reply and list of downstream routers in respect 
> to previous
> > > traceroute probe.
> > >
> > > Thank you,
> > > Igor
> > >
> > >
> > >
> > > >From: "David Allan" <dallan@nortelnetworks.com>
> > > >To: "'Igor Achkinazi'" <achkinazi@hotmail.com>
> > > >CC: "'mpls@UU.NET'" <mpls@UU.NET>
> > > >Subject: RE: MPLS Ping question
> > > >Date: Thu, 8 Apr 2004 10:37:42 -0400
> > > >
> > > >Hi Igor:
> > > >
> > > >The short answer is downstream mapping is not part of the 
> > > >processing rules for the request termination (only processes FEC 
> > > >stack and received label stack). The presence of the Downstream 
> > > >Mapping TLV in the request is effectively a flag saying "send me 
> > > >one in the reply" only.
> > > >
> > > >The error code in the reply will indicate correct 
> forwarding as the 
> > > >label and FEC in the FEC stack will align, it is up to the
> > > originator
> > > >of the ping to figure out that what actually happened does
> > > not concur
> > > >with the downstream mapping TLV from the previous step of the 
> > > >traceroute (IP address of reply will not match expected 
> downstream 
> > > >router). At that point to be truly rigorous, you'd need to
> > > back TTL up
> > > >one and see if downstream mapping had changed.
> > > >
> > > >cheers
> > > >Dave
> > > >
> > > > > -----Original Message-----
> > > > > From: Igor Achkinazi [mailto:achkinazi@hotmail.com]
> > > > > Sent: Wednesday, April 07, 2004 1:01 PM
> > > > > To: mpls@UU.NET
> > > > > Subject: RE: MPLS Ping question
> > > > >
> > > > >
> > > > > To avoid confusion I want to rephrase my question and clarify 
> > > > > the scenario I have a question about.
> > > > >
> > > > > Let's think about path established using LDP, and a network 
> > > > > administrator sending MPLS trace-route Echo request from 
> > > > > ingress. This MPLS Ping packet ends up at some LSR A in the 
> > > > > middle of MPLS cloud:
> > > > >
> > > > >                  B ----------> (after topology change)
> > > > >                 /
> > > > > -------------> A
> > > > >                 \
> > > > >                  C ----------> (before topology change)
> > > > >
> > > > > When MPLS trace-route Echo request (not regular MPLS
> > > ping) comes to
> > > > > LSR A before topology changes, LSR A reports C as a 
> downstream 
> > > > > router in Downstream Mapping TLV. Immediately after that
> > > IGP reroute
> > > > > causes LDP to
> > > > > reconfigure it's downstream information on LSR A. 
> When ingress 
> > > > > sends a new MPLS Echo request with TTL incremented LSR A 
> > > > > forwards it to B, but MPLS Echo
> > > > > request has old Downstream Mapping TLV reported by A before,
> > > > > which has label
> > > > > and IP address from C. So B will report FEC/Label mismatch.
> > > > >
> > > > > I cannot find how the draft covers such scenario
> > > > >
> > > > > Thank you,
> > > > > Igor Achkinazi
> > > > >
> > > > >
> > > > >
> > > > >
> > > > > >From: "Igor Achkinazi" <achkinazi@hotmail.com>
> > > > > >To: mpls@UU.NET
> > > > > >Subject: MPLS Ping question
> > > > > >Date: Mon, 05 Apr 2004 18:42:02 +0000
> > > > > >
> > > > > >Hi,
> > > > > >
> > > > > >There were several discussions about MPLS Ping already, but
> > > > > I could not
> > > > > >find the answer to following question about trace-route
> > > > > mode: assume that
> > > > > >just before reroute some midpoint LSR X responds with
> > > its current
> > > > > >downstream information (label, interface, hashing
> > > > > addresses), so MPLS Ping
> > > > > >initiator will build a packet meant to go to the old
> > > > > downstream in respect
> > > > > >to LSR X. However at the time LSR X gets new MPLS Ping
> > > > > packet (with TTL
> > > > > >incremented by one), its hardware will be configured with a
> > > > > new forwarding
> > > > > >information and the packet will be forwarded to a new
> > > > > downstream router
> > > > > >that will complain about FEC/Label mismatch.
> > > > > >
> > > > > >Since trace-route fails because of topology change, 
> it should 
> > > > > >be repeated again. But network operator sending MPLS Ping
> > > from ingress
> > > > > router might not
> > > > > >know that some LSR along the way just did a reroute, so he
> > > > > would assume
> > > > > >that the LSP is broken. How the operator can differentiate
> > > > > between topology
> > > > > >changes and broken LSP, since LDP topology changes are not
> > > > > visible to him?
> > > > > >
> > > > > >Thank you,
> > > > > >- Igor Achkinazi
> > > > > >
> > > > >
> > > > > 
> ________________________________________________________________
> > > > > _
> > > > > MSN Toolbar provides one-click access to Hotmail from any
> > > Web page -
> > > > > FREE download! 
> > > > > http://toolbar.msn.com/go/onm00200413ave/direct/01/
> > > > >
> > > > >
> > >
> > > _________________________________________________________________
> > > Watch LIVE baseball games on your computer with MLB.TV, included 
> > > with MSN Premium!
> > > http://join.msn.com/?page=features/mlb&pgmarket=en-us/go/onm00
> >200439ave/direct/01/
> >
> 
> _________________________________________________________________
> FREE pop-up blocking with the new MSN Toolbar - get it now! 
> http://toolbar.msn.com/go/onm00200415ave/direct/01/
> 
>