From: Con Spathas (ccie19226@googlemail.com)
Date: Fri Nov 21 2008 - 06:49:34 ARST
Good point Pavel although I'm not entirely convinced it's CoS that I'm
seeing as I'm not marking that anywhere.
R1 is setting the DSCP to EF to begin with but once inside the SP it's all
MPLS EXP.
I think Antonio hit the nail on the head in my case with the bug report he
found. A little disconcerting though to be told not to use 'debug' as the
workaround.
Although the marking does seem to work for me even when I have the debug on
as I observed on the next hop router in the SP cloud - which is different to
what Rich is seeing.
My testing was done on 3640 running 12.4(16) code. I haven't really tried
any other code at this point.
I think being aware of this on lab day is better than seeing it for the
first time and possibly wasting time trying to debug something that is
actually working.
On Thu, Nov 20, 2008 at 8:03 PM, Pavel Bykov <slidersv@gmail.com> wrote:
> Since CEF is the mechanism that is working with tags, I expect FIB/LFIB to
> do the tag bidding.
> When you debug, then it has to be process switched, so that might be
> responsible.
>
> What I was thinking of though, is the exact output of debug command. Is it
> really EXP or CoS? Because since you are using VLANs, it could just refer to
> cos markings.
> Are you setting CoS as well? Can you try setting CoS?
>
>
> On Thu, Nov 20, 2008 at 5:20 PM, Rich Collins <nilsi2002@gmail.com> wrote:
>
>> Hello,
>>
>> I just tried a very similar scenario to yours (Uniform mode) with
>> dynamips.
>>
>> If I turned on "debug mpls packet"on P1 (where the match and rewrite of
>> the
>> EXP occurs) then the policy maps/service policies wouldn't work. It
>> confused me because I had seen some increment in the counters but that was
>> from the time before I had activated the debug. So it worked properly
>> after
>> turning off this debug and observing further along in the P core such as
>> on
>> your R4 (enabling debug mpls packet here).
>>
>> Is there some caveat when using "debug mpls packet"?
>>
>> -Rich
>>
>>
>>
>> On Mon, Nov 17, 2008 at 11:40 AM, Con Spathas <ccie19226@googlemail.com
>> >wrote:
>>
>> > Gday,
>> >
>> > I think I'm missing something quite fundamental here but I can't seem to
>> > find the reason why...
>> >
>> > I have the following setup:
>> >
>> > R1 (CE) - R2 (PE) - R3 (P) - R4 (P) - R5 (PE) - R6 (CE)
>> >
>> > R2 is matching ICMP inbound from R1 and marking it MPLS EXP 4
>> > R3 is matching MPLS EXP 4 from R2 and remarking it to MPLS EXP1
>> >
>> > I see that is happening when I debug on R4 (the COS=4 on the XMIT is the
>> > VPN
>> > label) which is what I expect as R5 is currently sending implicit null
>> to
>> > R4.
>> > If I configure R5 to advertise explicit null in LDP, the debug output
>> > changes correctly and the xmit CoS = 1. Anyhow that's not my issue -
>> just
>> > background info.
>> >
>> > R4
>> > *Mar 1 02:40:51.859: MPLS: Et0/0.34: recvd: CoS=1, TTL=253,
>> Label(s)=19/21
>> > *Mar 1 02:40:51.863: MPLS: Et0/0.45: xmit: CoS=4, TTL=252, Label(s)=21
>> >
>> > The bit that is eluding my understanding is the output on R3.
>> >
>> > R3
>> > *Mar 1 00:10:59.075: MPLS: Et0/0.23: recvd: CoS=4, TTL=254,
>> Label(s)=17/21
>> > *Mar 1 00:10:59.079: MPLS: Et0/0.34: xmit: CoS=4, TTL=253,
>> Label(s)=19/21
>> >
>> > I can't work out why the xmit is showing the CoS=4 - even though R4 does
>> > receive it as CoS=1 ? I expected on R3 to see CoS=1 on the XMIT.
>> >
>> > This is the policy map in have inbound on R3's connection to R2.
>> >
>> > class-map match-all EXP4
>> > match mpls experimental topmost 4
>> > !
>> > policy-map IN
>> > class EXP4
>> > set mpls experimental topmost 1
>> >
>> > The policy-map output is incrementing counters correctly. So whilst the
>> > re-marking is working - I don't quite understand why the debug on R3
>> isn't
>> > consistent with what I subsequently see on R4.
>> >
>> > Any insight would be great. Thanks.
>> >
>> > Con.
>> >
>> >
>> > Blogs and organic groups at http://www.ccie.net
>> >
>> > _______________________________________________________________________
>> > Subscription information may be found at:
>> > http://www.groupstudy.com/list/CCIELab.html
>>
>>
>> Blogs and organic groups at http://www.ccie.net
>>
>> _______________________________________________________________________
>> Subscription information may be found at:
>> http://www.groupstudy.com/list/CCIELab.html
>>
>>
>>
>>
>>
>>
>>
>>
>
>
> --
> Pavel Bykov
> -------------------------------------------------
> Stop the braindumps!
> http://www.stopbraindumps.com/
Blogs and organic groups at http://www.ccie.net
This archive was generated by hypermail 2.1.4 : Mon Dec 01 2008 - 08:18:31 ARST