cancel
Showing results for 
Search instead for 
Did you mean: 
cancel
141565
Views
39
Helpful
71
Comments
xthuijs
Cisco Employee
Cisco Employee

 

Introduction

In this article we'll discuss how to troubleshoot packet loss in the asr9000 and specifically understanding the NP drop counters, what they mean and what you can do to mitigate them.

This document will be an ongoing effort to improve troubleshooting for the asr9000, so if after reading this article things are still not clear, please do comment on the article and I'll have details added in case my current descriptions are not explanatory enough.

 

Core Issue

Before we are going to dive into the packet troubleshooting on the asr9000 it is important to undertand a bit of the architecture of the system.

Packets are received by the mac/phy, forwarded to the NPU who is processing the features, the NPU hands it over to the bridge who in turn delivers

the packets to the FIA (Fabric Interfacing Asic) upon which it is sent over the fabric (residing physically on the RSP) and handed over to the egress

Linecard who follows the FIA->bridge->NP-MAC path then.

The NP processes packets for both ingress and egress direction.

 

The Bridge is just a memory interface converter and is non blocking, unless you have QOS back pressure, there is generally no reason

for the Bridge to drop packets. This article will not focus on the bridge and FIA drop counters as they are generally not common.

 

Both NP, Bridge, FIA and Fabric are all priority aware and have separate queues for High and Low priority traffic as well as Unicast and Multicast. So when you look at the bridge you see UC/MC (unicast/mcast) as well as LP/HP queues.

A different article will focus on that.

 

Resolution

 

Here a graphical overview of how the individual asics are interconnecting with the different commands that point to the actual

asic of where we are viewing statistics from.

forwarding_architecture.jpg

In this picture Y is the slot where the linecard is inserted on and A identifies the individual interface.

 

The first place to verify drops is the interface. The regular show interface command is well known.

Drops reported on the interface are an aggregate of phyiscal layer drops as well as (certain!!) drops that were experienced in the NP.

 

When printing the show controller np counters command, you may not know which interface identifier connects to which, therefore use

the "show controller np ports all location 0/Y/CPU0" to find out which interface maps to which NP.

 

RP/0/RSP1/CPU0:A9K-BOTTOM#sh controllers np ports all loc 0/7/CPU0
Fri Mar  4 12:09:41.132 EST

 

                Node: 0/7/CPU0:
----------------------------------------------------------------

 

NP Bridge Fia                       Ports
-- ------ --- ---------------------------------------------------
0  0      0   TenGigE0/7/0/3
1  0      0   TenGigE0/7/0/2
2  1      0   TenGigE0/7/0/1
3  1      0   TenGigE0/7/0/0

 

In case you have an oversubscribed linecard, 2 TenGig interfaces can be connected to the same NP.

For 80G linecards such as the A9K-8T, you'll see 8 NP's listed.

For the 40 port 1GE, you'll see 10 1-Gig interfaces connecting to an NP, note that this is not oversubscription.

 

With that info you can now view the controller statistics and some information as to what the controller is doing.

Note that not all counters in this command constitute a problem! They are regular traffic counters also.

 

There is a simple rate mechanism that keeps the shadow counts from the previous command invokation.

Upon the next time the command is run delta's are taken and divided over the elapsed time to provide some rate counts.

It is important for more accurate rate counts instantly that you run this command twice the first time with a 2/3 second interval.

 

 

The output may look similar to this:

 

RP/0/RSP1/CPU0:A9K-BOTTOM#show controller np count np1 loc 0/7/CPU0
                Node: 0/7/CPU0:
----------------------------------------------------------------

Show global stats counters for NP1, revision v3

Read 30 non-zero NP counters:
Offset  Counter                                         FrameValue   Rate (pps)
-------------------------------------------------------------------------------
  22  PARSE_ENET_RECEIVE_CNT                                 51047           1
  23  PARSE_FABRIC_RECEIVE_CNT                               35826           0
  30  MODIFY_ENET_TRANSMIT_CNT                               36677           0
  31  PARSE_INGRESS_DROP_CNT                                     1           0
  34  RESOLVE_EGRESS_DROP_CNT                                  628           0
  40  PARSE_INGRESS_PUNT_CNT                                  3015           0
  41  PARSE_EGRESS_PUNT_CNT                                    222           0
....

 

What all these mnemonics mean is listed out in this table below:

 

 

1.  Common Global Counters

 

 

Counter NameCounter Description
IPV4_SANITY_ADDR_CHK_FAILURESIngress IPV4 address martian check failure 127.0.0.0/8 224.0.0.0/8, not 224.0.0.0/4 (225.0.0.0/8 and 239.0.0.0/8 don't increment error  count)   When source address is below, ASR9K forward the packet without any error. 0.0.0.0/8 10.0.0.0/8 128.0.0.0/16 169.254.0.0/16 172.16.0.0/12 191.255.0.0/16 192.0.0.0/24 223.255.255.0/24 240.0.0.0/4
DROP_IPV4_LENGTH_ERRORIngress IPV4 header length invalid or frame length incompatible with fragment length
DROP_IPV4_CHECKSUM_ERRORIngress IPV4 checksum invalid
IPV6_SANITY_ADDR_CHK_FAILURESIngress IPV6 frame  source or destination address invalid
DROP_IPV6_LENGTH_ERRORIngress IPV6 frame length invalid
MPLS_TTL_ERRORIngress MPLS frame outer label TTL is 0
MPLS_EXCEEDS_MTUMPLS TTL exceeded – punt frame to host
DROP_IPV4_NOT_ENABLEDIngress IPV4 frame and IPV4 not enabled in L3 ingress intf struct
DROP_IPV4_NEXT_HOP_DOWNDrop bit set in rx adjacency (ingress) or tx adjacency (egress) and ICMP punt disabled in intf struct (ingress/egress) for the interface –  drop bit in non-recursive adjacency or adjacency result
IPV4_PLU_DROP_PKTDrop bit set in leaf or non-recursive adjacency lookup miss (ingress/egress) and ICMP punt disabled in intf struct (ingress/egress) for the interface.
IPV4_RP_DEST_DROPIPV4 RP drop bit set in non-recursive adjacency or adjacency result and ICMP punt disabled in intf struct (ingress/egress) for the interface
IPV4_NULL_RTE_DROP_PKTNo IPV4 route (ingress/egress) and ICMP punt disabled in intf struct (ingress/egress) for the interface – null route set in leaf or recursive adjacency
DROP_IPV6_NOT_ENABLEDIngress IPV6 frame and IPV6 not enabled in L3 ingress intf struct
DROP_IPV6_NEXT_HOP_DOWNDrop bit set in rx adjacency (ingress) or tx adjacency (egress) and ICMP punt disabled in intf struct (ingress/egress) for the interface –  drop bit in non-recursive adjacency or adjacency result
DROP_IPV6_PLUDrop bit set in leaf or non-recursive adjacency lookup miss (ingress/egress) and ICMP punt disabled in intf struct (ingress/egress) for the interface.
IPV6_RP_DEST_DROPRP drop bit set in non-recursive adjacency or adjacency result and ICMP punt disabled in intf struct (ingress/egress) for the interface
IPV6_NULL_RTE_DROP_PKTNo IPV6 route (ingress/egress) and ICMP punt disabled in intf struct (ingress/egress) for the interface – null route set in leaf or recursive adjacency
MPLS_NOT_ENABLED_DROPIngress MPLS frame and MPLs not enabled in L3 ingress intf struct
MPLS_PLU_NO_MATCHIngress MPLS frame and outer label lookup miss and not a VCCV frame associated with a pseudo-wire
MPLS_PLU_DROP_PKTIngress MPLS frame and label lookup results return a NULL route, or have no forwarding control bits set,  OR one of the following: 1)     Next hop down – Drop bit set in rx adjacency (ingress) or tx adjacency (egress) and ICMP punt disabled in intf struct (ingress/egress) for the interface –  drop bit in non-recursive adjacency or adjacency result 2)     Leaf drop - Drop bit set in leaf or non-recursive adjacency lookup miss (ingress/egress) and ICMP punt disabled in intf struct (ingress/egress) for the interface. 3)     RP drop - RP drop bit set in non-recursive adjacency or adjacency result and ICMP punt disabled in intf struct (ingress/egress) for the interface 4)     No route - No route (ingress/egress) and ICMP punt disabled in intf struct (ingress/egress) for the interface – null route set in leaf or recursive adjacency 5)     Punt bit set in leaf results or recursive adjacency results – egress mpls punt not supported
MPLS_UNSUPP_PLU_LABEL_TYPEIngress MPLS frame outer label is 15 or less and not a router alert or null tag.
DEAGG_PKT_NON_IPV4Ingress MPLS frame outer label lookup control bits are L3VPN deagg. And the payload is not IPV4.

 

2.  NP debug and drop counters

 

  • Bulk      Forwarding Counts
  • Inject/Punt      type counts
  • Main      Drop counts
  • QOS/ACL      drop counts

 

 

Counter NameCounter Description
PARSE_ENET_RECEIVE_CNTTotal frames received from Ethernet on an ingress NP
PARSE_FABRIC_RECEIVE_CNTTotal frames received by an egress NP from the fabric – Note that this currently includes loopback frames
MODIFY_FABRIC_TRANSMIT_CNTTotal frames sent to fabric by an ingress NP
MODIFY_ENET_TRANSMIT_CNTTotal frames sent to Ethernet by an egress NP
PARSE_LOOPBACK_RECEIVE_CNTTotal loopback frames received by an NP – Note that these are egress loopback frames only and do not include ingress loopback frames such as Interflex
PARSE_INTERFLEX_RECEIVE_CNTTotal frames received by an NP from the interflex loopback port – these are ingress loopback frames
MODIFY_FRAMES_PADDED_CNTTotal number of frames sent to fabric that were undersized and required padding to 60 bytes.
MODIFY_RX_SPAN_CNTIngress frames sent to fabric on a SPAN link – these frames are replicated
MODIFY_TX_SPAN_CNTEgress frames sent to Ethernet on a SPAN link – these frames are replicated
MODIFY_LC_PUNT_CNTTotal frames punted to LC host (all punts are handled by the modify engine)
MODIFY_RP_PUNT_CNTTotal frames punted to RP host (all punts are handled by the modify engine)
MODIFY_PUNT_DROP_CNTTotal punt packets dropped by ingress modify engine due to punt policing, OR due to punt reason lookup fail.
MODIFY_LC_PUNT_EXCD_CNTPunt frames to LC dropped due to punt policing
MODIFY_RP_PUNT_EXCD_CNTPunt frames to RP dropped due to punt policing
PARSE_INGRESS_DROP_CNTTotal packets dropped by ingress NP in the parse engine
PARSE_EGRESS_DROP_CNTTotal packets dropped by egress NP in the parse engine
RESOLVE_INGRESS_DROP_CNTTotal packets dropped by ingress NP in the resolve engine
RESOLVE_EGRESS_DROP_CNTTotal packets dropped by egress NP in the resolve engine
MODIFY_INGRESS_DROP_CNTTotal packets dropped by ingress NP in the modify engine
MODIFY_EGRESS_DROP_CNTTotal packets dropped by ingress NP in the modify engine
PARSE_INGRESS_PUNT_CNTTotal packets flagged for punting by the ingress NP in the parse engine
PARSE_EGRESS_PUNT_CNTTotal packets flagged for punting by the egress NP in the parse engine
RESOLVE_INGRESS_L3_PUNT_CNTTotal Layer3 packets flagged for punting  by an ingress NP in the resolve engine due to lookup results or other cases such as ICMP unreachable
RESOLVE_INGRESS_IFIB_PUNT_CNTTotal Layer3 packets flagged for punting  due to an ingress IFIB lookup match
RESOLVE_INGRESS_L2_PUNT_CNTTotal Layer2 packets flagged for punting  by the egress NP in the resolve engine due to cases such as IGMP/DHCP snooping or punting of ingress L2 protocol such as LAG or MSTP or CFM
RESOLVE_EGRESS_L3_PUNT_CNTTotal Layer3 packets flagged for punting  by an ingress NP in the resolve engine due to lookup results or other cases such as ICMP unreachable
RESOLVE_EGRESS_L2_PUNT_CNTTotal Layer2 packets flagged for punting  by the egress NP in the resolve engine due to cases such as IGMP/DHCP snooping or punting of ingress L2 protocol such as MSTP or CFM
PARSE_LC_INJECT_TO_FAB_CNTFrames injected from LC host that are intended to go directly to fabric to an egress LC without modification or protocol handling (inject header is stripped by microcode)
PARSE_LC_INJECT_TO_PORT_CNTFrames injected from  LC host that are intened to go directly to an Ethernet port without modification or protocol handling  (inject header is stripped by microcode)
PARSE_FAB_INJECT_IPV4_CNTTotal number of IPV4 inject frames received by the egress NP from the fabric (sent from RP)
PARSE_FAB_INJECT_PRTE_CNTTotal number of layer 3 preroute inject frames received by the egress NP from the fabric (sent from RP)
PARSE_FAB_INJECT_UNKN_CNTTotal number of unknown inject frames received by the egress NP from the fabric (sent from RP).  These are punted to the LC host.
PARSE_FAB_INJECT_SNOOP_CNTTotal number of IGMP/DHCP snoop inject frames received by the egress NP from the fabric (sent from RP).  For DHCP, these are always re-injects from the RP.  For IGMP, these can be re-injects or originated injects from the RP.
PARSE_FAB_INJECT_MPLS_CNTTotal number of mpls inject frames received by the egress NP from the fabric (sent from RP).
PARSE_FAB_INJECT_IPV6_CNTTotal number of IPV6 inject frames received by the egress NP from the fabric (sent from RP)
INJECT_EGR_PARSE_PRRT_NEXT_HOPInject preroute frames received from the fabric that are next hop type.
INJECT_EGR_PARSE_PRRT_PITInject preroute frames received from the fabric that are PIT type .
L3_EGR_RSV_PRRT_PUNT_TO_LC_CPUInject preroute frames that have a PIT table lookup miss.  These are punted to the LC host.
L2_INJECT_ING_PARSE_INWARDTotal number of ingress CFM injects received from the LC host that are intended to be sent to the fabric.    These can either be bridged or pre-routed to a specific NP.
CFM_L2_ING_PARSE_BRIDGEDTotal number of ingress CFM injects received from the LC host that are intended to be bridged and sent to the fabric.  The bridge lookups are done for these inject frames resulting in unicast or flood traffic.
L2_INJECT_ING_PARSE_PREROUTETotal number of ingress CFM injects received from the LC host that are intended to be pre-routed to a specific output port on an egress NP. bridged and sent to the fabric.  The bridge lookups are not done for these inject frames.
L2_INJECT_ING_PARSE_EGRESSTotal number of ingress CFM injects received from the LC host that are intended to be sent to a specific output port on an egress NP.
INJ_UNKNOWN_TYPE_DROPUnknown ingress inject received from the LC host – these are dropped.
STATS_STATIC_PARSE_INTR_AGE_CNTTotal Number of Parse Interrupt frames associated with learned MAC address aging.
STATS_STATIC_PARSE_INTR_STATS_CNTTotal Number of Parse Interrupt frames associated with statistics gathering.
PARSE_VIDMON_FLOW_AGING_SCAN_CNTTotal Number of Parse Interrupt frames associated with video monitoring flow scanning.
UNSUPPORTED_INTERRUPT_FRAMEUnsupported interrupt frame type received on NP.  Supported types are for aging and statistics only.
RESOLVE_BAD_EGR_PRS_RSV_MSG_DROP_CNTInvalid parse engine to resolve engine message received for an ingress frame – this is likely a software implementation error.
RESOLVE_BAD_ING_PRS_RSV_MSG_DROP_CNTInvalid parse engine to resolve engine message received for an egress frame – this is likely a software implementation error.
EGR_UNKNOWN_PAK_TYPEA frame received from the fabric by an egress NP has an unknown NPH header type – these are dropped.    This is likely a software implementation error.
EGR_LB_UNKNOWN_PAK_TYPEA frame received from the egress loopback queue by an egress NP has an unknown NPH header type – these are dropped.    This is likely a software implementation error.
XAUI_TRAINING_PKT_DISCARDThese are diag packets sent to the egress NP at startup from the fabric FPGA.   These are ignored and dropped by software.
PRS_DEBUGTRACE_EVENTThis counter increments whenever a preconfigured debug trace event matches on an ingress or egress frame.   Currently the debug trace is only configurable for all drop events.
IN_INTF STRUCT_TCAM_MISSIngress frame with a pre-parse TCAM lookup miss – these are dropped.  This is likely due to a TCAM setup problem.
IN_INTF STRUCT_NO_ENTRYIngress frame with a pre-parse TCAM index that is invalid – these are dropped.  This is likely due to a TCAM setup problem
IN_INTF STRUCT_DOWNIngress frame with a intf struct entry with down status – these are dropped, with the exeception of ingress LACP frames on a subinterface.
OUT_INTF STRUCT_INTF STRUCT_DOWNEgress Layer3 frame with an egress INTF STRUCT that has line status down, and ICMP punt is disabled in the intf struct.
IN_L2_INTF STRUCT_DROPIngress Layer2  frame on a subinterface with the L2 intf struct drop bit set.
ING_VLAN_FILTER_DISCARDA vlan tagged packet has a intf struct index that does points to a default layer3 intf struct – these are dropped.  Only untagged pkts can be setup on a default layer3 intf struct.
IN_FAST_DISCA ingress frame was dropped due to the early discard feature.   This can occur at high ingress traffic rate based on the configuration.
UNKNOWN_L2_ON_L3_DISCARDUnknown Layer2 frame (non-IPV4/V6/MPLS) on a Layer3 interface.     These are dropped.   Known layer2 protocols are punted on a layer3 interface such as ARP/LACP/CDP/ISIS/etc.
L3_NOT_MYMACA layer3 unicast IPV4/IPV6 frame does not have the destination mac address of the router, or any mac address defined in the VRRP table.  These are dropped.
IPV4_UNICAST_RPF_DROP_PKTWith RPF enabled, an ingress IPV4 frame has a source IP address that is not permitted, or not permitted on the ingress interface – these are dropped.
RESOLVE_L3_ING_PUNT_IFIB_DROPIngress frame has TTL of 0, or discard BFD frame if BFD not enabled
MPLS_VPN_SEC_CHECK_FAILAn ingress MPLS does not have the proper VPN security code, and IPV4 is not using the global routing table.
MPLS_EGR_PLU_NO_MATCHMPLS frame outer label lookup on egress NP does not match – these are dropped.
MPLS_EGR_PLU_DROP_PKTMPLS frame outer label lookup on egress NP has a result with the drop bit set.
RESOLVE_EGR_LAG_NOT_LOCAL_DROP_CNTAn egress frame hash to a bundle interface has a bundle member that is not local to the NP – these are dropped.  This can be due to a bad hash, or a bad bundle member index.   It can also occur normally for floods or multicast – non-local members are dropped for each copy that is a bundle output.
RESOLVE_L3_ING_ERROR_DROP_CNTL3 ingress protocol handling in the resolve engine falls through to an invalid case – these are dropped.  This would occur due to a software implementation error.
RESOLVE_L3_EGR_ERROR_DROP_CNTL3 egress protocol handling in the resolve engine falls through to an invalid case – these are dropped.  This would occur due to a software implementation error.
ING_QOS_INVALID_FORMATAn ingress packet has a QOS format value in the ingress intf struct that is not valid.   Only checked if QOS is enabled in the ingress intf struct.   Valid formats are 20/3/2/1/0.
RESOLVE_L3_ING_ACL_DENY_DROP_CNTTotal number of ingress Layer3 frames dropped due to an ACL deny policy.
RESOLVE_L3_EGR_ACL_DENY_DROP_CNTTotal number of egress Layer3 frames dropped due to an ACL deny policy.
RESOLVE_L2_ING_ACL_DENY_DROP_CNTTotal number of ingress Layer2 frames dropped due to an ACL deny policy.
RESOLVE_L2_EGR_ACL_DENY_DROP_CNTTotal number of egress Layer2 frames dropped due to an ACL deny policy.
RESOLVE_L2L3_QOS_C_DROP_CNTTotal number of ingress/egress frames dropped to a child policing policy.
RESOLVE_L2L3_QOS_P_DROP_CNTTotal number of ingress/egress frames dropped to a parent policing policy.

 

 

Layer3 Multicast related counts

These include:

 

  • IP      mcast forwarding counts
  • Video      Monitoring counts
  • Multicast      Fast Reroute counts

 

Counter NameCounter Description
IPV4_MCAST_NOT_ENABLEDIngress IPV4 frames dropped due to ipv4 multicast not enabled.   The ipv4 multicast enable check may be ignored for non-routable IGMP frames. Also, the ipv4 multicast enable check does not apply to well known multicast frames.
IPV6_MCAST_UNSUPPORTED_DROPIngress IPV6 multicast frame and IPV6 multicast not supported in the current release
RESOLVE_IPM4_ING_RTE_DROP_CNTIngress IPV4 multicast frame route lookup fails with no match.
RESOLVE_IPM4_EGR_RTE_DROP_CNTEgress IPV4 multicast frame route lookup fails with no match.
RESOLVE_IPM4_NO_OLIST_DROP_CNTEgress IPV4 multicast frame has zero olist members – no copies are sent to output ports.
RESOLVE_IPM4_OLIST_DROP_CNTEgress IPV4 multicast frame has olist drop bit set  (no copies are sent to output ports), OR egress IPV4 multicast frame olist lookup fails.   Olist lookup failure could be due to a bad copy number, or a missing entry for that copy.
RESOLVE_IPM4_FILTER_DROP_CNTEgress IPV4 multicast frame copy is reflection filtered – copies to the original source port are dropped.
RESOLVE_IPM4_TTL_DROP_CNTEgress IPV4 multicast frame copy fails TTL check.   The frame TTL is beyond the programmed threshold.
MODIFY_RPF_FAIL_DROP_CNTIngress IPV4 multicast frames dropped due to reverse-path-forwarding check failed (to prevent looping or duplicated packets); these multicast frames came in from wrong interface.
RESOLVE_VIDMON_FLOWS_DELETED_BY_AGINGInactive Video Monitoring flows deleted by hardware aging mechanism.
RESOLVE_VIDMON_CLASS_FLOW_LIMIT_REACHEDIndicates  that the Video Monitoring Maximum flows per class limit reached.
RESOLVE_VIDMON_FLOW_START_INTRVL_RESETVideo Monitoring flow start interval reset to current Real Time Clock Tick (instead of updating with actual next packet arrival time) because the flow has been inactive for more than 3 interval timeouts.
VIDMON_PUNT_FAILED_NO_TXBUF_CNTPunt frames dropped by Video Monitoring due to no buffer
PARSE_MOFRR_WATCHDOG_INTR_RCVDL3 Multicast Fast Reroute Watchdog interrupts received by nP. The count is summation of both Activity and Loss interrupt counts.
PARSE_MOFRR_SWITCH_MSG_RCVD_FROM_FABNumber of L3 Multicast Fast Reroute Switch (Flood) messages received by nP from Fabric.
RESOLVE_MOFRR_HRT_LKUP_FAIL_DROPNumber of watchdog interrupts are dropped because of an error in MoFRR Hash Table  lookup. This counter indicates that the watchdog interrupt (Activity or Loss) is received but micro code failed to process further because of MoFRR Hash look up failed. This counter indicates an error.
RESOLVE_MOFRR_WDT_INVALID_RESULTS_DROP_CNTNumber of watchdog interrupts are dropped because of an error in Watch Dog counter hash table lookup. This counter indicates that the watchdog interrupt (Activity or Loss) is received but micro code failed to process further because of  lookup failed. This counter indicates an error.
RESOLVE_MOFRR_SWITCH_MSG_HRT_LKUP_FAIL_DROPNumber of  MoFRR switch messages are dropped because of an error with MoFRR Hash Table lookup failure. This counter indicates that the switch message is received from fabric but micro code failed to process further because of MoFRR Hash  look up failed. This counter indicates an error and could be traffic disruptive.
RESOLVE_MOFRR_SWITCH_MSG_SEQNUM_MISMATCH_DROPNumber of MoFRR switch messages are dropped because of sequence number mismatch. This counter indicates that the switch message is received from fabric but micro code failed to process further the sequence number received from ingress LC does not match with sequence number for same route on egress LC. This counter indicates a programming error, race condition or hash result corruption in micro code etc. This is not a usual thing and could be traffic disruptive.
RESOLVE_MOFRR_HASH_UPDATE_CNTNumber of times MoFRR active flag is updated in MoFRR hash table.  This counter indicates the switch message is processed successfully, i.e. RDT/Hash/Direct table lookups are successful and MoFRR Active flag is updated in MoFRR hash table.
RESOLVE_MOFRR_SWITCH_MSG_INGNOREDTotal numbers of MoFRR switch (flood) messages are filtered without being punted to LC CPU. This counter indicates switch message is processed successfully (i.e. lookups successful and MoFRR flag is updated in hardware hash table), but the message is not punted to LC CPU either because the RPF Id is not primary or RPF Id  is not local to the nP.  Note that this counter is informative and does not necessarily mean an error.
RESOLVE_MOFRR_SWITCH_MSG_TO_FABNumber of MoFRR switch (flood) messages generated by nP as a result of Loss Activity interrupts.  This counter indicates that the Loss Activity interrupt is processed successfully by ingress LC and MoFRR switch message is  flooded to all LCs towards fabric.
RESOLVE_MOFRR_EGR_RPF_FAIL_DROPIP Multicast MoFRR enabled packets dropped at egress LC because of RPF check failure.

 

 

 

Layer 2 protocol Debug Counters

  These include: ·         VPLS counts ·         VPWS counts ·         DHCP/IGMP snooping counts ·         CFM protocol counts ·         VPLS Learning/Aging related counts  

Counter NameCounter Description
IN_L2_BLOCKEDIngress L2 frames dropped and not forwarded due to ingress interface in spanning tree blocked state.
OUT_L2_BLOCKEDEgress L2 frames dropped and not forwarded due to egress interface in spanning tree blocked state.
IN_DOT1Q_VTP_FILTEREDIngress VTP frames dropped due to dot1q filtering enabled.
IN_DOT1Q_DROPIngress 802.1q frames dropped due to dot1q filtering enabled.
IN_DOT1AD_DROPIngress 802.1ad frames dropped due to dot1ad filtering enabled.
ING_SLOW_PROTO_UNKNOWN_SUBTYPEIngress slow protocol frame on an L3 interface that is not recognized (not LACP/EFM/SYNCE).
IN_ROUTER_GUARD_DROPIngress Layer2  IGMP or PIM dropped due to Router Guard with IGMP snooping enabled.
IN_DHCP_UNTRUSTED_DROPIngress DHCP frame dropped due to the untrusted DHCP bit set in the ingress L2 intf struct.
RESOLVE_L2_DHCP_SNOOP_UNTRUSTED_DROP_CNTEgress DHCP frame dropped due to the untrusted DHCP bit set in the egress L2 intf struct.
ING_DHCP_PW_DROPIngress DHCP frame from AC that is destined for a PW and is eligible for snooping due to the DHCP snoop enabled bit set in the L2 ingress INTF STRUCT.    This is dropped on ingress due to an L2 unicast to PW.   DHCP snooping not supported for AC to PW frames.
EGR_DHCP_PW_DROPIngress DHCP frame from AC that is destined for a PW and is eligible for snooping due to the DHCP snoop enabled bit set in the L2 ingress INTF STRUCT.     This is dropped on egress due to an L2 flood with a PW member. DHCP snooping not supported for AC to PW frames.
L2VPN_DEAGG_NON_VPWS_VPLS_DROPInvalid setting for L2VPN deaggregation set in Label UFIB lookup result for ingress MPLS frame.   This is most likely due to a programming error in the Label UFIB entry.
XID_ZERO_DISCARDIngress VPWS XC to AC frame has a zero destination XID stored in the L2 ingress INTF STRUCT – zero XID index is invalid.
XC2AC_NOMATCH_DROPEgress VPWS XC to AC frame XID lookup fails.   This is caused by an invalid XID index or missing XID entry.
L2UNIC_NOMATCH_DROPEgress VPLS AC to AC frame XID lookup fails.   This is caused by an invalid XID index or missing XID entry.    This can also occur on an IGMP snoop originated inject to an AC/PW/Bundle with invalid XID in the inject feature header.
L2FLOOD_NOMATCH_DROPEgress VPLS flood copy lookup in the flood member table fails.  This could be caused by an invalid copy number or missing entry in the flood member table for that copy.
L2FLOOD_SRCDST_DROPEgress VPLS flood copy is reflection filtered – copies to the original source port are dropped.
L2FLOOD_INACTIVE_MEMBER_DROPEgress VPLS flood copy lookup in the flood member table returns an entry that’s not active (active bit in result is cleared).
L2FLOOD_EGR_XID_NO_MATCHEgress VPLS flood copy XID lookup fails – XID stored in the  OLIST result is invalid or entry is missing.   This can occur if the copy is intended to go over a PW and the PW is down.
EGR_AC_PW_MISMATCHEgress VPWS XC to AC  XID lookup returns a PW type.   For VPWS, PW entry is always on ingress.   Note that for VPLS AC to AC, a PW type XID is allows in order to support IGMP injects to a PW.
RESOLVE_VPWS_LAG_BIT_DROP_CNTAn ingress VPWS XC to Bundle frame with a LAG table lookup fail.  This is due to programming error where the LAG table pointer is invalid or no LAG table entry.
RESOLVE_L2_SXID_MISS_DROP_CNTAn ingress VPLS frame with a source XID lookup fail.   This can be due to an invalid source XID programmed in the L2 ingress intf struct, or a missing source XID entry.
RESOLVE_VPLS_SPLIT_HORIZON_DROP_CNTAn ingress or egress VPLS that is dropped due to a split horizon group mismatch.   Note that the split horizon group check is skipped for zero split horizon group indexes.
RESOLVE_VPLS_REFLECTION_FILTER_DROP_CNTAn ingress unicast VPLS frame that has a source XID that matches its destination XID.   These are dropped since it cannot be sent back to the source port.
RESOLVE_VPLS_MAC_FILTER_DROP_CNTAn ingress VPLS frame has a source or destination MAC with the filter bit set in the L2FIB result – these are dropped.
RESOLVE_VPLS_ING_BD_STR_MISS_DROP_CNTAn ingress VPLS frame that is dropped due to a bridge domain lookup fail.
RESOLVE_VPLS_EGR_BD_STR_MISS_DROP_CNTAn egress VPLS flood frame or mac notify frame that is dropped due to a BD structure lookup miss.
RESOLVE_VPLS_FLOOD_BLOCK_DROP_CNTAn Ingress VPLS frame that is dropped due to its associated flood block bit being set:  Unknown Unicast or Multicast.
RESOLVE_VPLS_STORM_DROP_CNTAn ingress VPLS frame that is dropped due to rate exceed for its associated flood storm control – Unknown Unicast or Multicast or Broadcast.
RESOLVE_VPLS_EGR_FLD_NO_MEMBERS_DROP_CNTEgress L2 flood frame has zero members – no copies are sent to output ports.
RESOLVE_L2_EGR_INTF STRUCT_MISS_DROP_CNTAn egress VPLS unicast or flood copy frame to an AC or Bundle that is dropped to an egress L2 INTF STRUCT lookup miss.  This is due to an invalid INTF STRUCT index in the XID or a missing egress L2 INTF STRUCT entry.    This can also occur with L2 multicast forwarded frames or L2 multicast inject frames.
RESOLVE_L2_EGR_PW_INTF STRUCT_MISS_DROP_CNTAn egress VPLS unicast or flood copy frame to a PW that is dropped to an egress L2 INTF STRUCT lookup miss.  This is due to an invalid INTF STRUCT index in the XID or a missing egress L2 INTF STRUCT entry.  This can also occur if the PW is down.   .    This can also occur with L2 multicast forwarded frames or L2 multicast inject frames to a PW.
RESOLVE_VPWS_L2VPN_LDI_MISS_DROP_CNTAn ingress VPWS or VPLS unicast frame to a PW with an L2VPN LDI table lookup fail.  This can be due to an invalid L2VPN LDI index in the XID,  or a missing entry in the L2VPN LDI table.
RESOLVE_VPWS_RP_DROP_CNTAn ingress VPWS or VPLS unicast frame to a PW with the RP drop bit or RP dest. bit set in the NON-RECURSIVE ADJACENCY or RX-ADJ lookup results.    Punts are not supported for an L2 PW adjacency.
RESOLVE_VPWS_NULL_ROUTE_DROP_CNTAn ingress VPWS or VPLS unicast frame to a PW with the Null Route bit or IFIB punt bit set in the Leaf lookup or RECURSIVE ADJACENCY lookup results.   Punts are not supported for an L2 PW adjacency.
RESOLVE_VPWS_LEAF_DROP_CNTAn ingress VPWS or VPLS unicast frame to a PW with the Drop bit or Punt bit set in the Leaf lookup or RECURSIVE ADJACENCY lookup results.   Punts are not supported for an L2 PW adjacency.
RESOLVE_VPLS_EGR_NULL_RTE_DROP_CNTAn egress VPLS flood frame copy to a PW with the Null Route bit set in the Leaf lookup or RECURSIVE ADJACENCY lookup results.    Also includes IGMP snoop injects to a PW.   Due to a software problem, this may also get incremented due to a BD structure lookup miss on egress flood.
RESOLVE_VPLS_EGR_LEAF_IFIB_DROP_CNTAn egress VPLS flood frame copy to a PW with the IFIB punt bit set in the Leaf lookup or RECURSIVE ADJACENCY lookup results.   Punts are not supported for an L2 PW adjacency.  Also includes IGMP snoop injects to a PW.
RESOLVE_VPLS_EGR_LEAF_PUNT_DROP_CNTAn egress VPLS flood frame copy to a PW with the Drop bit or Punt bit or Next Hop Down bit set in the Leaf lookup or RECURSIVE ADJACENCY or NON-RECURSIVE ADJACENCY lookup results.     Punts are not supported for an L2 PW adjacency.  Also includes IGMP snoop injects to a PW.
RESOLVE_VPLS_EGR_TE_LABEL_DROP_CNTAn egress VPLS flood frame copy to a PW with the Drop bit or Punt bit or Incomplete Adjacency bit set in the Tx adjacency or Backup Adjacency or NH Adjacency lookup results.    Punts are not supported for an L2 PW adjacency.  Also includes IGMP snoop injects to a PW.
RESOLVE_VPLS_EGR_PW_FLOOD_MTU_DROP_CNTAn egress VPLS flood frame copy to a PW that fails L3 MTU check.   Punts are not supported for an L2 PW adjacency with MTU fail.  Also includes IGMP snoop injects to a PW.
RESOLVE_VPLS_EGR_PW_FLOOD_INTF STRUCT_DOWN_DROP_CNTAn egress VPLS flood frame copy to a PW with egress INTF STRUCT in down state.   Punts are not supported for an L2 PW adjacency with line status down.  Also includes IGMP snoop injects to a PW.
RESOLVE_VPLS_EGR_PW_FLOOD_NO_NRLDI_DROP_CNTAn egress VPLS flood frame copy to a PW with and NON-RECURSIVE ADJACENCY lookup miss.   Also includes IGMP snoop injects to a PW.
RESOLVE_INGRESS_LEARN_CNTTotal number of VPLS Source MAC learns on an ingress frame – learned locals.
PARSE_FAB_MACN_W_FLD_RECEIVE_CNTCount of all frames received from fabric on egress LC with NPH L2 flood type and mac notify included (mac notify for ingress SA learn included with L2 flood)
PARSE_FAB_MACN_RECEIVE_CNTCount of all frames received from fabric on egress LC with NPH mac notify flood type (mac notify for ingress SA learn).      This includes: 1)     forward mac notifies (from ingress LC) 2)     reverse mac notifies (from egress LC in response to flood 3)     mac deletes (due to a mac move or port flush)
PARSE_FAB_DEST_MACN_RECEIVE_CNTCount  of all mac notify frames received from fabric on egress LC that are reverse mac notify type
PARSE_FAB_MAC_DELETE_RECEIVE_CNTCount  of all mac notify frames received from fabric on egress LC that are mac delete type
RESOLVE_LEARN_FROM_NOTIFY_CNTTotal number of VPLS Source MAC learns from mac notify received.
RESOLVE_VPLS_MAC_MOVE_CNTTotal number of source MAC moves detected as a result of an ingress learn local or mac notify receive learn.
RESOLVE_BD_FLUSH_DELETE_CNTTotal number of learned entries deleted due to a Bridge Domain Flush
RESOLVE_INGRESS_SMAC_MISS_CNTTotal number of ingress frames with source MACs that did have a learned entry.  These source MACs will be relearned if learned limits have not been exceeded, learning is enabled, and the address is unicast.
RESOLVE_VPLS_MCAST_SRC_MAC_DROP_CNTTotal number of ingress frames with source MACs that were not unicast.  These frames are invalid and are are dropped.   Non-unicast source macs cannot be learned.
RESOLVE_HALF_AGE_RELEARN_CNTTotal mac address from ingress frames that were relearned due to exceeding the half age period.  These entries are refreshed to update the age timeout.
STATS_STATIC_PARSE_PORTFLUSH_CNTEgress VPLS frame that has been detected as port flushed.  Not valid if port based mac accounting is disabled.  The frame will be dropped and mac delete indication sent to the bridge domain.
RESOLVE_PORT_FLUSH_DROP_CNTEgress unicast frames dropped because the destination port has been flushed.  For each a Mac Delete frame is sent to the bridge domain to remove the destination address.
RESOLVE_PORT_FLUSH_DELETE_CNTCount of learned MACs deleted due to their source port being flushed.  Does not include MACs deleted due to MAC Delete message.
RESOLVE_PORT_FLUSH_AGE_OUT_CNTVPLS MAC entries that have been port flushed as detected by the MAC aging scan.  These frames are dropped.  Also can occur if the frame XID lookup fails and the XID is global or learned local – XID has been removed.
RESOLVE_VPLS_LEARN_LIMIT_ACTION_DROP_CNTIngress VPLS Frames dropped to learned limit exceeded – port based or bridge domain based learn limits – these frames are dropped.
RESOLVE_VPLS_STATIC_MAC_MOVE_DROP_CNTA static MAC entry that are deleted due to a MAC move detected – these should never move.
RESOLVE_MAC_NOTIFY_CTRL_DROP_CNTTotal number of processed mac notify frames – forward or reverse mac notify – dropped after processing.
RESOLVE_MAC_DELETE_CTRL_DROP_CNTTotal number of processed mac delete frames – forward or reverse mac notify – dropped after processing.
RESOLVE_AGE_NOMAC_DROP_CNTMAC scan aging message dropped due to an L2FIB lookup fail for the scanned mac entry.
RESOLVE_AGE_MAC_STATIC_DROP_CNTMAC scan aging message dropped due to static MAC address for the scanned mac entry.
EGR_VLANOPS_DROPEgress L2 frame without the proper number of vlans to do egress VLAN ops – the frame from the fabric needs at least one vlan to do a pop operation or replace operation.
EGR_PREFILTER_VLAN_DROPEgress L2 frame that fails EFP pre-filter check – only valid if EFP filtering enabled – attached VLAN mismatch.
RESOLVE_EFP_FILTER_TCAM_MISS_DROPEgress L2 frame that fails EFP filter TCAM – only valid if EFP filtering enabled – likely due to a TCAM missing entry due to a programming error.
RESOLVE_EFP_FILTER_MISS_MATCH_DROPEgress L2 frame that fails EFP filter check – only valid if EFP filtering enabled – attached VLAN mismatch.
CFM_ING_PUNTIngress CFM frame punt count.
CFM_EGR_PUNTEgress CFM frame punt count.
CFM_EGR_PARSETotal egress Layer2 CFM frames processed by the NP – including egress CFM inject and egress CFM forwarded frames.
INJ_NO_XID_MATCH_DROPIngress CFM frame injects with an XID lookup fail – source XID for bridged injects, and dest. XID for pre-route injects – invalid XID index or missing entry.
EGR_CFM_XID_NOMATCHEgress CFM frame inject with a dest.  XID lookup fail – invalid XID index or missing entry.
EGR_CFM_PW_DROPEgress CFM frame injected to a PW – not supported.
PARSE_INGRESS_L3_CFM_DROP_CNTIngress CFM frame on a layer 3 interface dropped due to CFM not enabled.
RESOLVE_VPLS_MC_ING_RTE_DROP_CNTIngress L2 multicast frame route lookup fails with no match.
RESOLVE_VPLS_MC_EGR_RTE_DROP_CNTEgress L2 multicast frame route lookup fails with no match.
RESOLVE_VPLS_MC_NO_OLIST_DROP_CNTEgress L2 multicast frame has zero olist members – no copies are sent to output ports.
EGR_L2MC_OLIST_NOMATCHEgress L2 multicast frame has olist drop bit set  (no copies are sent to output ports), OR egress L2 multicast frame olist lookup fails.   Olist lookup failure could be due to a bad copy number, or a missing entry for that copy.
EGR_L2MC_SRCDSTEgress L2 multicast frame copy is reflection filtered – copies to the original source port are dropped.
EGR_L2MC_XID_NOMATCHEgress L2 multicast frame copy XID lookup fails – XID stored in the  OLIST result is invalid or entry is missing.   This can occur if the copy is intended to go over a PW and the PW is down.
RESOLVE_L2_PROT_BYTE_ERROR_DROP_CNTInvalid L2 protocol action set – this is an internal software programming error on an ingress or egress L2 frame if incremented.
RESOLVE_VPLS_ING_ERR_DROP_CNTAn ingress VPLS frame has an invalid processing type.  This is likely an internal error in software.
RESOLVE_VPLS_EGR_STRUCT_MISS_DROP_CNTEgress VPLS frame protocol handling falls through to an invalid case.  This is likely a software implementation error.

 

3.  Punt Reason Counters

 

For each punt reason, there are 2 counters.  The total number of frames attempted to be punted by microcode is the sum of these 2 counters.

 

  • Actual      punted
  • Punt      drops due to policing

 

If the punt reason lookup fails, then software will drop the packet with MODIFY_PUNT_DROP_CNT increment.    A layer3 interface is either a router gateway or a default interface for untagged frames (or both).

 

 

Counter NameCounter Description
PUNT_INVALID PUNT_INVALID_DROPInvalid Punt Reason – this should not occur.
PUNT_ALL PUNT_ALL_EXCDPunt All Set – not implemented by software.
CDP CDP_EXCDPunt Ingress CDP protocol frames – layer3 interface only.
ARP ARP_EXCDPunt Ingress ARP protocol frames – layer3 interface only.
RARP RARP_EXCDPunt Ingress Reverse ARP protocol frames – layer3 interface only.
CGMP CGMP_EXCDPunt Ingress CGMP protocol frames – layer3 interface only.
LOOP LOOP_EXCDPunt Ingress ELOOP protocol frames – layer3 interface only.
CFM_OTHER CFM_OTHER_EXCDIngress NP Punt or Egress NP Punt of CFM other packets (link trace or perf. mon.).   Ingress punts can occur on a Layer2 or Layer3 interface Egress punts occur on a Layer2 interface only.
CFM_CC CFM_CC_EXCDIngress NP Punt or Egress NP Punt of CFM CC packets (continuity check).   Ingress punts can occur on a Layer2 or Layer3 interface Egress punts occur on a Layer2 interface only.
DHCP_SNOOP_REQ DHCP_SNOOP_REQ_EXCDIngress Punt of DHCP snoop request frames – layer2 interface only and DHCP snooping enabled.
DHCP_SNOOP_REPLY DHCP_SNOOP_REPLY_EXCDIngress Punt of DHCP snoop reply frames – layer2 interface only and DHCP snooping enabled.
MSTP MSTP_EXCDIngress NP Punt or Egress NP Punt of MSTP frames.  For Layer2 VPLS or Layer3, punt is on ingress frames only.  For Layer3 VPWS, punt is on egress frames only.
MSTP_PB MSTP_PB_EXCDIngress NP Punt or Egress NP Punt of MSTP Provider Bridge frames.  For Layer2 VPLS or Layer3, punt is on ingress frames only.  For Layer3 VPWS, punt is on egress frames only.
IGMP_SNOOP IGMP_SNOOP_EXCDIngress punt of snooped IGMP or PIM frames – layer2 interface only and IGMP snooping enabled.
EFM EFM_EXCDIngress Punt of EFM Protocol frames – layer3 interface only.
IPv4_OPTIONS IPv4_OPTIONS_EXCDIngress Punt of IP frames with options – layer3 interface only – IPV4, IPV6.
IPV4_PLU_PUNT   (IPV4_FIB_PUNT) IPV4_PLU_PUNT_EXCDIngress or Egress punt of IPV4 frames with punt flag set in Leaf results or RECURSIVE ADJACENCY results – layer3 interface only.
IPV4MC_DO_ALL  IPV4MC_DO_ALL_EXCDIngress punt of IPV4 Mcast frames due to punt only set for route, OR Egress punt of IPV4 Mcast frames due to punt and forward bit set for mcast group member – layer3 interface only.   IPV4 Mcast punts not support for egress route.
IPV4MC_DO_ALL_BUT_FWD IPV4MC_DO_ALL_BUT_FWD_EXCDIngress punt and forward of IPV4 Mcast frames due to punt only set for route, OR Egress punt and forward of IPV4 Mcast frames due to punt and forward bit set for mcast group member – layer3 interface only.   IPV4 Mcast punts not support for egress route.  This can also occur as a result of  the RPF source interface check.
PUNT_NO_MATCH   (PUNT_FOR_ICMP) PUNT_NO_MATCH_EXCDAn Ingress or Egress IPV4/IPV6 unicast frame punt due to ICMP Unreachable (i.e., next hop down).
IPV4_TTL_ERROR    (TTL_EXCEEDED) IPV4_TTL_ERROR_EXCDAn Ingress IPV4 frame with a TTL of 1 is punted (TTL of 0 is dropped).
IPV4_DF_SET_FRAG_NEEDED_PUNT  IPV4_DF_SET_FRAG_NEEDED_PUNT_EXCDAn egress IPV4 or MPLS frame with MTU exceeded is punted.
RP_PUNT RP_PUNT_EXCDAn ingress IPV4 or MPLS has the RP punt bit set in the NON-RECURSIVE ADJACENCY or RX adjacency.
PUNT_IFIB PUNT_IFIB_EXCDAn ingress IPV4 ucast or IPV6 frame has an IFIB lookup match resulting in an IFIB punt.
PUNT_ADJ PUNT_ADJ_EXCDAn ingress IPV4 ucast or MPLS frame  has the adjacency punt bit set in the NON-RECURSIVE ADJACENCY or RX adjacency.  An egress IPV4 ucast or MPLS frame  has the adjacency punt bit set in the NON-RECURSIVE ADJACENCY or TX adjacency.
PUNT_UNKNOWN_IFIB PUNT_UNKNOWN_IFIB_EXCDAn ingress IPV4 ucast frame has the IFIB bit set in the LEAF or RECURSIVE ADJACENCY and the IFIB lookup does not get a match, OR an ingress MPLS frame has the IFIB bit set in the LEAF or RECURSIVE ADJACENCY.
PUNT_ACL_DENY PUNT_ACL_DENY_EXCDIngress or Egress  frame has a deny action set in the ACL policy and ACL punt (instead of drop) is enabled.
PUNT_ACL_LOG PUNT_ACL_LOG_EXCDIngress or Egress  frame has an ACL policy match and ACL logging is enabled.    Layer3 frames only.
PUNT_ACL_LOG_L2 PUNT_ACL_LOG_L2_EXCDIngress or Egress  frame has an ACL policy match and ACL logging is enabled.    Layer2 frames only.
IPV6_LINK_LOCAL IPV6_LINK_LOCAL_EXCDIngress IPV6 frames received that are link local packets – these are punted – layer3 interface only.
IPV6_HOP_BY_HOP IPV6_HOP_BY_HOP_EXCDIngress IPV6 frames received that are hop by hop packets  - these are punted – layer3 interface only.
IPV6_TTL_ERROR IPV6_TTL_ERROR_EXCDIngress IPV6 frames that have a TTL error – these are punted – layer3 interface only.
IPV6_PLU_PUNT IPV6_PLU_PUNT_EXCDIngress or Egress punt of IPV6 frames with punt flag set in Leaf results or RECURSIVE ADJACENCY results – layer3 interface only.
IPV6_PLU_RCV IPV6_PLU_RCV_EXCDNot implemented by software.
IPV6_ROUT_HEAD IPV6_ROUT_HEAD_EXCDIngress IPV6 frames received that are router header packets – these are punted – layer3 interface only.
IPV6_TOO_BIG IPV6_TOO_BIG_EXCDEgress IPV6 frames that are too bit – exceed MTU – these are punted – layer3 interface only.
IPV6_MISS_SRC IPV6_MISS_SRC_EXCDNot implemented by software.
IPV6MC_DC_CHECK IPV6MC_DC_CHECK_EXCDNot implemented by software.
IPV6MC_DO_ALL IPV6MC_DO_ALL_EXCDNot implemented by software.
IPV6MC_DO_ALL_BUT_FWD IPV6MC_DO_ALL_BUT_FWD_EXCDNot implemented by software.
MPLS_PLU_PUNT  (MPLS_FIB_PUNT) MPLS_PLU_PUNT_EXCDIngress MPLS frame punted due to Punt set in the Leaf or RECURSIVE ADJACENCY, or an Ingress MPLS router alert frame – these are punted.
MPLS_FOR_US MPLS_FOR_US_EXCDIngress MPLS frame punted due to IFIB (for-us) set in the Leaf or RECURSIVE ADJACENCY.
PUNT_VCCV_PKT PUNT_VCCV_PKT_EXCDIngress MPLS frame that is a PW VCCV frame – must be a PWtoAC de-aggregation flow.
PUNT_STATISTICS PUNT_STATISTICS_EXCDStatistics gathering frame punted to LC host – generated by statistics scanning machine.
NETIO_RP_TO_LC_CPU_PUNT NETIO_RP_TO_LC_CPU_PUNT_EXCDPreroute egress inject frame received from the RP that is sent to the LC host.
MOFRR_PUNT MOFRR_PUNT_EXCDIPV4 multicast over fast reroute punt frames.
VIDMON_PUNT VIDMON_PUNT_EXCDIPV4 multicast video monitoring punt frames.
VIDMON_PUNT_FLOW_ADD VIDMON_PUNT_FLOW_ADD_EXCDIPV4 multicast video monitoring new flow added punt frames.
PUNT_DBGTRACEDebug Trace frames that are punted.
PUNT_SYNCEPunt Ingress SYNCE protocol frames – layer3 interface only.

 

 

Related Information

n/a

 

Xander Thuijs - CCIE #6775

Sr Tech Lead ASR9000

Comments
Carlos T
Level 1
Level 1

Hi Xander,

Interesting article, thinking about npus and bridge in the linecards, I just try to understand, what is the criteria or best practice, when choosing npus and bridges in the linecard for special links, like the core links which aggregate all traffic, or multicast sender/receiver links (multicast replication)

For better understanding I thought in 2 scenarios. Kindly appreciate if you can clarify each one:

1. Scenario A: an asr9k as a PE router with a 40x1G low queue linecard. 3x 1G ports are used for uplink to the core, load balancing in this 3 ports is achieved by ecmp of the igp. Only unicast traffic is sent/received by these 3 core uplinks.

Question: What npu/bridges could be choosen for the 3 uplink ports, could be safe to choose one npu/bridge for the 3 core uplinks? or is it better to chose one different npu for each uplink?

2. Scenario B: an asr9k as a PE router with a 40x1G low queue linecard, 1 port is connected to the source of multicast traffic, and 3 ports are connected to the receivers of the multicast traffic.

Question:What npu/bridges could be choosen for the link to the sender, and to the receivers? Trying to avoid excess multicast replication in the linecard, could be safe to choose one npu/bridge for the sender and receivers? or is it better to chose one different npu for the sender, and other npu for the receivers? Also, if choosing different npus, both belonging to the same bridge? or or to different bridges?

Thanks,

Carlos Trujillo.

xthuijs
Cisco Employee
Cisco Employee

Hi Carlos,

If you have multiple uplinks I would try to split them across different linecards, so that when one LC would fail, the other paths/links are not affected.

Next I would spread them over different NPU's. If the NPU would reset for whatever reason, not all your uplinks are going to be affected.

Bridge can be omitted in this regard because it is non blocking (just doing header translation) and the only reason why the brdige might drop traffic is when the egress NPU is applying backpressure, then the bridge drops temporarily traffic (separated by Priority, default and mcast in separate classes) until the backpressure has reached the source.

For your scenario 2, if the mcast is overwhelming the NPU, then mcast will be flowed off, not affecting priority or default class unicast traffic.

This is the concept of virtual output Queues (see article on qos architecture on the supportforums I wrote) whereby on the FIA's we have sets of VOQ's (here P1, P2, default ucast and mcast) per egress NP.

This means that one interface getting overloaded or flowing off on the NP may flow off other interfaces connecting to the same NP.

thanks!

xander

acowdell99
Level 1
Level 1

Hi Alexander,

Very informative post, thank you for this.   Could you let me know what the equivalent command to "show controllers np counters all loc x/x" would be for the A9K-SIP-700/ SPA cards?

Thanks

Anthony

xthuijs
Cisco Employee
Cisco Employee

Thanks anthony, that's nice to hear!

the command for CPP (/SIP700) is show controller pse qfp [0-1] [datapath/flag-utilization | statistics/drop] location X

it is not the same type of output as the trident/typhoon but it gives you an idea what the CPP is working on or doing.

The designator 0-1 is the ingress or egress CPP.

regards

xander

acowdell99
Level 1
Level 1

Great, the suggested command works fine but this now leads on to a further question; we have a large number of drops in the "Ipv4Unclassified" counter, and i can;t find any reference to this mnemonic.

I'm aware that the SIP700 will drop ingress traffic destined to a L2 address, would this scenario be be included in this counter?

xthuijs
Cisco Employee
Cisco Employee

I thought it had something to do with a drop adjacency. I am checking the cpp drivers to find out when this gets set.

more to follow (if there is news on that )

xander

xthuijs
Cisco Employee
Cisco Employee

confirmed, this is a drop adjacency

AARON WEINTRAUB
Level 1
Level 1

I'm seeing quite a few counters that are not seen in this doc:

204  INTR_FRAME_TYPE_7                                        589          11

405  RSV_DROP_ING_BFD                                          62           1

439  RSV_DROP_IPV4_URPF_CHK                                   848          18

592  RSV_SNF_SAMPLE_PKT                                     11952         218

593  RSV_SNF_IPV4                                           11952         218

594  RSV_SNF_ING_PUNT                                        7085         129

595  RSV_SNF_EGR_PUNT                                        4867          89

604  RESOLVE_REMOTE_RACK_PREP_CNT                             593          11

This is on a

A9K-24x10GE-TR on 4.2.3

xthuijs
Cisco Employee
Cisco Employee

hey aaron!

whoopsy! here you go:

204  INTR_FRAME_TYPE_7     NP interrupts frames for NPU messaging inside the chip. It is not a worrisome or drop counter.

405  RSV_DROP_ING_BFD     Ingress BFD frames "for me"

439  RSV_DROP_IPV4_URPF_CHK     uRPF check failed and dropped

592  RSV_SNF_SAMPLE_PKT       Frames which are sampled by Netflow processing and punted to local LC CPU

593  RSV_SNF_IPV4            Frames which are sampled by Netflow processing and send to 2nd pass processing for Source IP lookup.

594  RSV_SNF_ING_PUNT   ingress frames samples

595  RSV_SNF_EGR_PUNT egress frame sampled

604  RESOLVE_REMOTE_RACK_PREP_CNT     cluster messaging

regards

xander

msrilists
Community Member

Xander,

I couldn't find an explanation for this counter

 

RSV_DROP_MPLS_LEAF_NO_MATCH    1000
 

This is on IOS-XR 5.1.1

 

Thanks!

Bryan Garland
Cisco Employee
Cisco Employee

 

 RSV_DROP_MPLS_LEAF_NO_MATCH  = Frames dropped due to inconsistent software state which may occur during reprogramming

So these can be seen during transition times.

Hendro Marwanto
Level 1
Level 1

Thank you Sir for sharing that explanation, but im seeing a counters that are not seen in this doc:

RP/0/RSP0/CPU0:Hendro#sh controllers np counters np1 location 0/3/CPU0 | i DROP              
Tue Jul  1 15:46:53.001 UTC
 242  PARSE_FAST_DISCARD_LOW_PRIORITY_DROP_0               1337977        3804
 243  PARSE_FAST_DISCARD_LOW_PRIORITY_DROP_1               1335913        3814
 244  PARSE_FAST_DISCARD_LOW_PRIORITY_DROP_2               1335158        3862
 245  PARSE_FAST_DISCARD_LOW_PRIORITY_DROP_3               1337969        3829
 394  RSV_NO_RFD_DROP                                           10           0
1033  DROP_FRM_FRM_ERR_XAUI4                                    15           0
1037  DROP_FRM_FRM_ERR_XAUI6                                     9           0
1171  MDF_PUNT_POLICE_DROP                                   96437           2

Offset number 242 - 245 comes up after i set port SPAN from port Te0/3/1/0 to port Te0/3/0/3

Anyone can help to explain what the counters mean ?

 

Thanks,

Hendro 

xthuijs
Cisco Employee
Cisco Employee

Hi Hendro,

PARSE_FAST_DISCARD_X is seen when the npu is running out of steam, pps wise, and therefore needs to throttle incoming packets into the pipeline. SPAN is a reasonable cause for this because it replicates the packet and sends them back in the pipeline for processing towards its span destination.

Because the pps rate of the original traffic and the added span is getting too high together, we start throttling packets coming in based on their priority.

SPAN itself is low priority, so it wont push "good" traffic away, and is the first to be subject to throttling.

regards

xander

Hendro Marwanto
Level 1
Level 1

Hi Xander,

 

Excellent, base on your explanation i try to change SPAN direction from both to tx_only, and after 3 hours monitoring i don't seen again PARSE_FAST_DISCARD_X in np counters.

 

Now i run 2 SPAN and each SPAN forward around 6 Gbps with 700k pps

SPAN 1 : Te0/3/1/0 to Te0/3/0/3

SPAN 2 : Te0/3/1/2 to Te0/3/1/1

 

I have read documentation from Cisco Traffic Mirroring on the Cisco ASR 9000 they said :

Performance Impact with Traffic Mirroring
It is recommended that you do not mirror more than 15% of your total transit traffic. On the 
Cisco ASR 9000 Ethernet Line Card, that uses Ten Gigabit Ethernet interfaces or bundle interfaces there is a limit of 1.5G of data on each of the ingress and egress traffic that can be mirrored. This limitation is not applicable on the Cisco ASR 9000 Enhanced Ethernet Line Card.

 

Now our SPAN run with more than 60% from total traffic on each port, is that ok ? because i just curious it will be impact much on our ASR9K Performance.

I also confuse what the different Ethernet Line Card with Enhanced Ethernet Line Card ?

 

Regards,

Hendro

xthuijs
Cisco Employee
Cisco Employee

Hi Hendro;

cool yeah that makes sense that if you lower the pps rate, the fast discards are eliminated.

The "enhanced ethernet services" cards are the typhoon based ones. The ethernet services cards are the Trident based ones.

You have I believe typhoon based ones because of the fast discard drops (trident called them ifdma discards).

The limitation for typhoon cards is truly and only pps as its fabric ports are 60G of bw per npu and the fabric access is that same. For trident however, the BW capability of the NPU is ~15G only and also the fabric bw is limited, this means that trident is subject to both pps and bw exhaustion when it comes to span.

For that reason we advised to keep span on trident at 15%.

regards

xander

Getting Started

Find answers to your questions by entering keywords or phrases in the Search bar above. New here? Use these resources to familiarize yourself with the community:

Quick Links