[c-nsp] Problem with dscp packets marking on 76th platform.

[Teslenko]

Hello All,

We try to introduce Qos in ours IP/MPLS backbone network,
constructed on routers 7600th series

All 76-s' are P or PE devices should accept from outside  MPLS or IP
traffic.
On PE devices we mark packages and we want, that DSCP was transferred
transparently within MPLS domain. But we have problem.

We use IOS  v12.2 (33) SRC1 now.


Testing passed on CISCO7609-S with linear card WS-X6708-10GE


==================================
			              The test #1
==================================

P device CISCO7609-S ingress port on linear card WS-X6708-10GE,
egress port on linear card WS-X6724-SFP.
Device PE2 the last in a chain, PHP option enable  by default

 The scheme of traffic's movement looks as  follows

 CE--> SW-1> (ingress) PE1 (egress)--> P--> (ingress) PE2

1.1 Interfaces  of the P  device are configured as follows

interface TenGigabitEthernet (WS-X6708-10GE)
 dampening
 mtu 4470
 ip address yy.yy.yy.yy 255.255.255.252
 carrier-delay msec 0
 mpls traffic-eng tunnels
 mpls ip
 hold-queue 1000 in
 ip rsvp bandwidth
end
!
interface GigabitEthernet (WS-X6724-SFP)
 dampening
 mtu 4470
 ip address xx.xx.xx.xx 255.255.255.252
 carrier-delay msec 0
 mpls traffic-eng tunnels
 mpls ip
 hold-queue 1000 in
 ip rsvp bandwidth
end


1.2. Marking of traffic occur on ingress interface of PE1

>>
>> policy-map test-in-dscp-set
>> class class-default
>> set dscp 39

2.3. Stock-taking dscp labels occur on ingress interface of PE2.

>>
------------------------------------------------- listing-------------------
>> ping from CE
>>	Type escape sequence to abort.
>>	Sending 100, 100-byte ICMP Echos to 10.10.10.5, timeout is 2 seconds:
>>	Packet sent with a source address of 10.10.10.1
>>	!!!!!
>>	Success rate is 100 percent (100/100), round-trip min/avg/max = 1/4/9 ms
>>
>> PE2#sh policy-map interf Gi0/1.662 in class match-test-dscp
>> GigabitEthernet0/1.662
>>
>> Service-policy input: Customer-test-In
>>
>> Class-map: match-test-dscp (match-any)
>> 0 packets, 0 bytes
>> 30 second offered rate 0 bps
>> Match: ip dscp 39
>> 0 packets, 0 bytes
>> 30 second rate 0 bps
>>----------------- end of listing----------------------------------------

As appears from an example marking do not occur


1.4 MPLS trace looks as follows

PE1#trace mpls ipv4 213.xxx.xxx.4 255.255.255.255
Tracing MPLS Label Switched Path to 213.xxx.xxx.4/32, timeout is 2 seconds
 Type escape sequence to abort.
   0 213.xxx.xxx.202 MRU 4470 [Labels: 50 Exp: 0]
 L 1 213.xxx.xxx.201 MRU 4474 [Labels: implicit-null Exp: 0] 169 ms
 ! 2 213.xxx.xxx.18 4 ms

PE1 encapsulate MPLS header to a package with value of the label = 50
and a field
Exp=0

==================================
			              The test #2
==================================

The scheme of traffic's movement looks as  follows

 CE--> SW-1> (ingress) PE1 (egress) --->(ingress)PE2

2.1 Device PE2 the last in the chain, it have PHP option enable by default

MPLS trace looks as follows
PE1#trace mpls ipv4 213.xxx.xxx.4 255.255.255.255
Tracing MPLS Label Switched Path to 213.xxx.xxx.4/32, timeout is 2 seconds
 Type escape sequence to abort.
   0 213.xxx.xxx.19 MRU 4470 [Labels: implicit-null Exp: 0]
 ! 1 213.xxx.xxx.18 4 ms

PE1 don't encapsulate MPLS header to a package.
Result:

>> PE2#sh policy-map interf Gi0/1.662 in class match-test-dscp
>> GigabitEthernet0/1.662
>>
>> Service-policy input: Customer-test-In
>>
>> Class-map: match-test-dscp (match-any)
>> 100 packets, 0 bytes
>> 30 second offered rate 0 bps
>> Match: ip dscp 39
>> 100 packets, 0 bytes
>> 30 second rate 0 bps

So DSCP label comes on PE2

2.2 Device PE2 the last in a chain, option PHP switched off

PE2 (config) # mpls ldp explicit-null
PE2 (config) #

So PE1 encapsulate MPLS header to a package,
And as result packages again comes without DSCP label.

Result.
When MPLS label was encapsulated to header, as result DCSP label was
cleared.

Does anybody know decision for this problem?