[c-nsp] MPLS qos on 6500/sup720 - setting EXP when it shouldn't?

[Manu Chao]

Can you please send "show mls qos detailed"?

On Wed, Nov 17, 2010 at 1:02 PM, Phil Mayers <p.mayers at imperial.ac.uk>wrote:

> We're about to start the complex task of using QoS on our MPLS core network
> (the main reason is that we're seeing drops on downstream LAN switches with
> tiny buffers, and want the upstream core to egress the "important" packets
> first, hopefully mitigating this)
>
> Upon investigation I'm seeing traffic flowing across the network with MPLS
> EXP != 0. This is odd, because we have QoS enabled (for CoPP) but with all
> ports set to untrusted - my understanding is that the internal DSCP (and
> thus transmit CoS/EXP) should therefore be zero.
>
> For example:
>
> #sh mls qos last module 8
>
>  ----- Module [8] -----
> QoS last packet policing information:
>    ---------------------------------------------------------------------
>    Packet was transmitted
>    Packet L3 Prot: 0, packet length: 300, dont_plc: yes
>    Input COS: 0, TOS/DSCP: 0x0/0
>    Output TOS/DSCP: 0x0/0[unchanged]   Output COS: 0[unchanged]
>    Output MPLS EXP (if outgoing packet is MPLS): 3
>    ---------------------------------------------------------------------
>    Aggregate policer index: Input - 0(none), Output - 0(none)
>    thr_hi_ip: 0x0   leak_hi_ip: 0x3FF drop_ena_ag_ip: no
>    thr_lo_ip: 0x0   leak_lo_ip: 0x3FF
>    thr_hi_op: 0x0   leak_hi_op: 0x3FF drop_ena_ag_op: no
>    thr_lo_op: 0x0   leak_lo_op: 0x3FF
>    ---------------------------------------------------------------------
>    Microflow policer index: Input - 0(none), Output - 0(none)
>    ---------------------------------------------------------------------
>    Netflow policer: nf_hit: yes, nf_addr: 0x109D9, snap-shot matches
>    NT&NS: l3_prot: 6(0), 155.198.77.100.0xDA5D ==> 155.198.63.71.0x01BD
>    NT&NS: vlan_number: 4D
>    cr_tstamp: 0x2EBBE8, ls_tstamp_nf: 0xBB39BE, tos_nf: 0x0, use_nf_thr: no
>    thrshld_nf: 0x0, leak_nf: 0x0, drop_ena_nf: no,
>    bkt_nf: 0x0, pkt_cnt_nf: 0x0, thr_excd_cnt_nf: 0x0
>    ---------------------------------------------------------------------
>    acl_redirect: no, acl_rdt_rpf: no, acl_drop: no, police_drop: no
> no_route: no
>    err_pkt: no, l3_flow: yes, acl_bridge: no, rpf_fail: no, hw_error: no
> ip_dont_sc: no
>    to_router: no, recirculated: no, to_be_recirculated:no
>    apply_qos_ip: no, apply_qos_op: no, drop_ip: no drop_op: no
>    plc_level2: no, plc_level3: no, tos_i: 0x0 vmap_data: 0x0
>
> As you can see, the original packet has CoS and DSCP == 0. Vlan 0x4D is
> Vlan 77, which is tagged on a single switchport:
>
> interface GigabitEthernet8/23
>  description xxx
>  switchport
>  switchport trunk encapsulation do
>  switchport trunk native vlan 2005
>  switchport trunk allowed vlan 77,xxx
>  switchport mode trunk
>  switchport nonegotiate
>  logging event spanning-tree statu
>  storm-control broadcast level 0.1
>  no cdp enable
>
> #sh queueing interface g8/23
> Interface GigabitEthernet8/23 queueing strategy:  Weighted Round-Robin
>  Port QoS is enabled
> Trust boundary disabled
>
>  Port is untrusted
>  Extend trust state: not trusted [COS = 0]
>  Default COS is 0
>
> ...as you can see, no QoS config on this port. I'm baffled where these
> non-zero EXP are coming from; I'm seeing a mix of values (1-7) although most
> traffic is zero.
>
> Any ideas?
> _______________________________________________
> cisco-nsp mailing list  cisco-nsp at puck.nether.net
> https://puck.nether.net/mailman/listinfo/cisco-nsp
> archive at http://puck.nether.net/pipermail/cisco-nsp/
>