Add ingress low-TTL discard counter to interface and subinterface counters

[ostromart]

Add ingress low-TTL discard counter to interface and subinterface counters

This change introduces a new operational state counter in-low-ttl-discards under common interface counters (/interfaces/interface/state/counters/in-low-ttl-discards and /interfaces/interface/subinterfaces/subinterface/state/counters/in-low-ttl-discards) to track ingress packets dropped due to Time-To-Live (TTL) or Hop Limit expiration during transit forwarding.
The version has been bumped to 3.9.0 in openconfig-interfaces.yang.

Change-Id: I2eb15fbed7f42d17dbca3e6a789befd41993f6f9

Change Scope

• Description: This PR introduces in-low-ttl-discards to openconfig-interfaces.yang (under interface-common-counters-state) to count ingress packets discarded on a routed interface or subinterface due to TTL or Hop Limit expiration.
• Backwards Compatibility: This change is fully backward compatible as it only adds a new read-only operational state leaf counter to the existing interface-common-counters-state grouping.

Use case

This counter is designed for production network debugging and automated failure pinpointing. Silent packet discards in routed fabrics are difficult to diagnose using generic discard counters alone. Providing an explicit counter for TTL expiration drops (in-low-ttl-discards) allows network operators and SRE tools to quickly identify routing loops and TTL-expired discards at both interface and subinterface granularity.

Platform Implementations

• SAI Specification: SAI specification defines an explicit debug drop reason for TTL expiration: SAI_IN_DROP_REASON_TTL (SAI Debug Counters Proposal).
• Google Switch Telemetry: Google network switches export granular ingress TTL drop telemetry via UMF and SAI debug counters.
• Cisco IOS-XE: Exposes bad hop count traffic statistics (show ip traffic).
• Nvidia / Mellanox (WJH): Tracks router drops due to TTL expiration (Router Drops -> TTL value is too small).
• SONiC / Broadcom Telemetry: Pipeline counters support per-port ingress drop counting for TTL expiration events.

Tree View

 module: openconfig-interfaces
   +--rw interfaces
      +--rw interface* [name]
         ...
         +--ro state
         |  +--ro counters
         |     +--ro out-discards?            oc-yang:counter64
         |     +--ro out-errors?              oc-yang:counter64
         |     +--ro last-clear?              oc-types:timeticks64
+        |     +--ro in-low-ttl-discards?     oc-yang:counter64
         ...
         +--rw subinterfaces
            +--rw subinterface* [index]
               ...
               +--ro state
                  +--ro counters
                     +--ro out-discards?           oc-yang:counter64
                     +--ro out-errors?             oc-yang:counter64
                     +--ro last-clear?             oc-types:timeticks64
+                    +--ro in-low-ttl-discards?    oc-yang:counter64

[gemini-code-assist]

Code Review

This pull request updates the openconfig-interfaces YANG module to version 3.9.0 and introduces a new in-low-ttl-discards counter to track inbound packets dropped due to expired TTL or Hop Limit. The reviewer suggested a wording improvement in the counter's description to make it more implementation-agnostic, changing 'upon routing lookup' to 'during transit forwarding'.

[navaneethyv]

OC operator review 28th. Looks good.

Keeping in ready to discuss for further comments.

[navaneethyv]

OC operator review August 4th : no comments yet, keeping it for a week before we move to the next phase.

[navaneethyv]

August 11th.
OC operator's review , moving it to the last call

[earies]

Similar to prior comments on #1528 and #1529 - A few general comments/doubts:

• I'm not sure about the references being entirely correct here as show ip traffic commands have tended to be system wide, possible punt path, etc.... not at an interface or subinterface granularity - @rgwilton @romeyod @jsterne to confirm from their end - I'm not seeing show ip traffic as an actual command for EOS so also not sure if references are LLM assisted
• From a Juniper/HPE ASIC perspective, this is not tracked per interface but rather collective per ASIC - this could rather be represented under integrated-circuit counters
• RE: SAI references are imo not necessarily references to ASIC side implementation so I'd be curious if this really fits the criteria of ultimately being neutral or the likely case of implementations deviating and conveying as best as possible via other means

[rgwilton]

Also similar to Ebben's comments, Cisco's IOS XR doesn't support per interface counter drops due to IP TTL (I doubt that IOS XE does either). Again, we can gather/report these counters on a per ASIC/LC basis, and probably you can get more detail from NETFLOW/sFLOW.

Without knowing more details about the individual packets or traffic flows it is unclear to me how useful these more fine grained interface drop counters really are.