multicast counters to subinterface, remove physical counters from subinterface

[dplore]

Change Scope

• Fixes Add multicast counters to ip interface, remove physical counters from subinterface #905 which unintentionally removed interface ip counters
• Refactor counters for interfaces and subinterfaces to use a common grouping for counters used for both interfaces and subinterfaces
• Add multicast-pkt counters to common sub interface counters
• Add resets counter for interface grouping
• Deprecate in-fcs-error, in-unknown-protosand carrier transitions from subinterfaces
• Removed TODO item regarding additional error counters
• This change is backwards compatible

module: openconfig-interfaces
  +--rw interfaces
     +--rw interface* [name]
        +--ro state
        |  +--ro counters
        |  |  +--ro in-octets?             oc-yang:counter64
        |  |  +--ro in-pkts?               oc-yang:counter64
        |  |  +--ro in-unicast-pkts?       oc-yang:counter64
        |  |  +--ro in-broadcast-pkts?     oc-yang:counter64
        |  |  +--ro in-multicast-pkts?     oc-yang:counter64
        |  |  +--ro in-errors?             oc-yang:counter64
        |  |  +--ro in-discards?           oc-yang:counter64
        |  |  +--ro out-octets?            oc-yang:counter64
        |  |  +--ro out-pkts?              oc-yang:counter64
        |  |  +--ro out-unicast-pkts?      oc-yang:counter64
        |  |  +--ro out-broadcast-pkts?    oc-yang:counter64
        |  |  +--ro out-multicast-pkts?    oc-yang:counter64
        |  |  +--ro out-discards?          oc-yang:counter64
        |  |  +--ro out-errors?            oc-yang:counter64
        |  |  +--ro last-clear?            oc-types:timeticks64
        |  |  +--ro in-unknown-protos?     oc-yang:counter64
        |  |  +--ro in-fcs-errors?         oc-yang:counter64
        |  |  +--ro carrier-transitions?   oc-yang:counter64
        |  |  +--ro resets?                oc-yang:counter64.       <-- added
        +--rw subinterfaces
        |  +--rw subinterface* [index]
        |     +--ro state
        |     |  +--ro counters
        |     |     +--ro in-octets?             oc-yang:counter64
        |     |     +--ro in-pkts?               oc-yang:counter64
        |     |     +--ro in-unicast-pkts?       oc-yang:counter64
        |     |     +--ro in-broadcast-pkts?     oc-yang:counter64
        |     |     +--ro in-multicast-pkts?     oc-yang:counter64        <-- added
        |     |     +--ro in-errors?             oc-yang:counter64
        |     |     +--ro in-discards?           oc-yang:counter64
        |     |     +--ro out-octets?            oc-yang:counter64
        |     |     +--ro out-pkts?              oc-yang:counter64
        |     |     +--ro out-unicast-pkts?      oc-yang:counter64
        |     |     +--ro out-broadcast-pkts?    oc-yang:counter64
        |     |     +--ro out-multicast-pkts?    oc-yang:counter64      <-- added
        |     |     +--ro out-discards?          oc-yang:counter64
        |     |     +--ro out-errors?            oc-yang:counter64
        |     |     +--ro last-clear?            oc-types:timeticks64
        |     |     x--ro in-unknown-protos?     oc-yang:counter64    <-- deprecated
        |     |     x--ro in-fcs-errors?         oc-yang:counter64.          <-- deprecated
        |     |     x--ro carrier-transitions?   oc-yang:counter64.       <-- deprecated

Platform Implementations

• Cisco IOS XR - show interface counters includes multicast and interface resets
• JunOS - show interface statistics. Note this does not mention support for 'resets'
• Arista EOS - show interface counters includes multicast and interface hardware resets.
• Nokia SR Linux - show interface detail includes multicast counters but not interface reset counters.

Error counters analysis

The TODO item to investigate if additional interface error counters are needed was removed based on the following analysis:

I'd like to complain a little bit that this is expanding the scope of the PR which is intended to be small and only seeks to

• add multicast counters
• remove physical interface errors counters from subinterfaces which are invalid

Thank you for reading my complaint. :)

That said, I investigated and I think this TODO can be removed. Here are the error counters which seem to match commonly supported output from 'show interfaces' commands on various implementations. There is variation by This varies by hardware and NOS. Some platforms report only input and output errors without support for additional, more granular errors. Some platforms aggregate errors as input and output errors and also support more granular errors.

• Survey of JunOS platforms and different outputs
• Cisco 8000 series interface errors

Common error counters	OC Path	Description
runts	...ethernet/state/counters/in-undersize-frames	Number of received packets that were too small to be handled. Aggregated to input errors count.
giants	.../ethernet/state/counters/in-oversize-frames	Number of received packets that were too large. Aggregated to input errors count.
ignored	.../state/counters/in-discards	Total number of ignored packet errors. Ignored packets are those that are discarded because the interface hardware does not have enough internal buffers. Broadcast storms and bursts of noise can result in an increased number of ignored packets.
throttles	.../state/counters/in-discards	Number of packets dropped due to throttling (because the input queue was full).
input errors	.../state/counters/in-errors	Total number of received packets that contain errors and hence cannot be delivered. Compare this to total input drops, which counts packets that were not delivered despite containing no errors.
CRC	/interfaces/interface/ethernet/state/counters/in-crc-errors	Number of packets that failed the Ethernet CRC check.
frame	/interfaces/interface/state/counters/in-fcs-errors	Number of packets with bad framing bytes. Perhaps further investigation needed to de-duplicate the in-crc-errors and in-fcs-errors leaves.
overrun	/interfaces/interface/state/counters/out-discards	Number of overrun errors experienced by the interface. Overruns represent the number of times that the receiver hardware is unable to send received data to a hardware buffer because the input rate exceeds the receiver's ability to handle the data.
resets	/interfaces/interface/state/counters/resets	Number of times that the interface hardware has been reset.
output errors	/interfaces/interface/state/counters/out-discards	Number of times that the receiver hardware was unable to handle received data to a hardware buffer because the input rate exceeded the receiver's ability to handle the data.
underruns	/interfaces/interface/state/counters/out-discards	Underruns will occur when the low level transceiver tries to send data, but the buffer is still empty because the device hasn't serviced the packets yet.
output buffer failures	/interfaces/interface/state/counters/out-discards	Number of times that a packet was not output from the output hold queue because of a shortage of MEMD shared memory.
carrier transitions	interfaces/interface/state/counters/carrier-transitions	Number of times the interface state has transitioned between up and down.

[OpenConfigBot]

Major YANG version changes in commit d147c28:

[dplore]

Ready for review @robshakir

One open question for us is: how should multicast IP packets be counted? At the subinterface level or at the ip level or both?

It seems counters are duplicated (where applicable) on the subinterface container and IP container. I am not quite sure what the history is here and if I should duplicate the multicast counter at the ip container level.

[robshakir]

One minor fix. Did we check that this doesn't remove anything unexpectedly?

[OpenConfigBot]

Compatibility Report for commit d147c28:
✅ pyangbind@d122b7a

[dplore]

One minor fix. Did we check that this doesn't remove anything unexpectedly?

Fixed index.

Merged in latest breaking change detection (which now detects changed and deleted leafs and compares to the new semantic version) and it had findings in wifi ap interfaces model which was impacted by the refactoring of the groupings in openconfig-interfaces. (@wenovus +1) This is now fixed.

[dplore]

@mike-albano I made a small refactor to the wifi / ap-interfaces model which doesn't change the tree in any way. It was needed to be compatible with the refactoring done on the openconfig-interfaces model in this PR. Would like your approval.

[mike-albano]

LGTM.