New and Updated Telemetry Metrics
The following tables list the new, updated, and discontinued gNMI and OTEL metrics in Cumulus Linux 5.19.
New gNMI Metrics
The switch exports these xPaths on Spectrum-6 switches only. For information about the counters behind these metrics, refer to Show and Clear Back-to-sender Notification on Link Down Counters.
| Name | Description |
|---|---|
/components/component[name]/integrated-circuit/pipeline-counters/packet/trim/state/spxm-notify-sender-link-down-pkts |
Number of packets that matched the back-to-sender classifier on link down, both while the route is still in hardware and after FRR withdraws it. Counts notifications the switch attempted, not notifications the switch confirmed the sender received. |
/components/component[name]/integrated-circuit/pipeline-counters/packet/trim/state/notify-sender-trimmed-link-down-pkts |
Number of those packets the switch trimmed and sent as a back-to-sender notification. |
| Name | Description |
|---|---|
/components/component[name]/integrated-circuit/pipeline-counters/control-plane-traffic/vendor/nvidia/spectrum/punt-classifiers/punt-classifier[rule-id]/state/rule-id |
The ID of the punt classifier rule. This is also the list key. |
/components/component[name]/integrated-circuit/pipeline-counters/control-plane-traffic/vendor/nvidia/spectrum/punt-classifiers/punt-classifier[rule-id]/state/matched-packets |
The number of packets that the rule drops. |
/components/component[name]/integrated-circuit/pipeline-counters/control-plane-traffic/vendor/nvidia/spectrum/punt-classifiers/punt-classifier[rule-id]/state/matched-octets |
The number of octets that the rule drops. |
/components/component[name]/integrated-circuit/pipeline-counters/control-plane-traffic/vendor/nvidia/spectrum/punt-classifiers/punt-classifier[rule-id]/state/packet-type |
The packet type that the rule matches: PACKET_TYPE_SPXM or PACKET_TYPE_IPV6. |
/components/component[name]/integrated-circuit/pipeline-counters/control-plane-traffic/vendor/nvidia/spectrum/punt-classifiers/punt-classifier[rule-id]/state/destination-port |
The IPv6 UDP destination port that the rule matches. Present only for a rule with packet type PACKET_TYPE_IPV6. |
/components/component[name]/integrated-circuit/pipeline-counters/control-plane-traffic/vendor/nvidia/spectrum/punt-classifiers/punt-classifier[rule-id]/state/action |
The action that the switch programs for the rule, which is always ACTION_DROP. |
- Every rule with packet type
PACKET_TYPE_SPXMshares one counter. Each of those rule IDs reports the same value, so do not add the values together. - Each IPv6 UDP destination port has its own counter.
- The counters cover the CPU egress of one ASIC, so they are totals for each ASIC instead of for each front panel interface.
- These paths are read-only. Configure the punt classifier with the NVUE commands described in Control Plane Punt Classifier.
- The DSCP values that a rule matches are not available over gNMI.
| Name | Description |
|---|---|
/interfaces/interface[name=<interface-id>]/step-time-estimation/state/step-time-estimate |
Estimated step time, in seconds, of the AI training workload running on the interface. Requires step time estimation. |
| Name | Description |
|---|---|
/performance/interfaces/interface[name]/histograms/microburst/direction[dir][quantity] |
Per-interface microburst telemetry container. dir is rx or tx; quantity is packets or bytes. |
/performance/interfaces/interface[name]/histograms/microburst/direction[dir][quantity]/bin[upper-boundary]/count |
Histogram bin count for the most recent collection window. The value covers a single collection window and resets on read. |
/performance/interfaces/interface[name]/histograms/microburst/direction[dir][quantity]/score |
Burstiness score computed as σ² × μ in bin-index space. |
/performance/interfaces/interface[name]/histograms/microburst/direction[dir][quantity]/mean-bin-index |
Mean bin index μ over the current window. |
/performance/interfaces/interface[name]/histograms/microburst/direction[dir][quantity]/distribution-variance |
Bin-index variance σ² over the current window. |
/performance/interfaces/interface[name]/histograms/microburst/direction[dir][quantity]/peak-value |
Raw max_watermark value for the current window. Interpretation depends on quantity. |
/performance/interfaces/interface[name]/histograms/microburst/direction[dir][quantity]/threshold-state |
Operational threshold state for the configured score threshold. |
/performance/interfaces/interface[name]/histograms/microburst/direction[dir][quantity]/last-triggered |
Last time the threshold entered triggered state. Absent or unset if never triggered. |
The telemetry model names the packets or bytes selector quantity, where the NVUE command that configures it is unit. The two names refer to the same setting; quantity avoids a collision with the standard OTEL instrument descriptor field.
The switch exports these xPaths on Spectrum-6 switches only. For information about the counters behind these metrics, refer to Packet Trimming Counters.
| Name | Description |
|---|---|
/components/component[name]/integrated-circuit/pipeline-counters/packet/trim/state/out-trim-pkts |
Number of packets that the switch trimmed and sent successfully. |
/components/component[name]/integrated-circuit/pipeline-counters/drop/state/trim-pkts |
Number of packets that the switch trimmed and dropped. |
/interfaces/interface[name]/state/counters/dropped-trim-pkts |
Number of packets that the switch trimmed and dropped on the port. |
/qos/interfaces/interface[interface-id]/output/queues/queue[name]/state/out-trim-pkts |
Number of packets that the switch trimmed and sent successfully on the port and queue. |
/qos/interfaces/interface[interface-id]/output/queues/queue[name]/state/dropped-trim-pkts |
Number of packets that the switch trimmed and dropped on the port and queue. |
| Name | Description |
|---|---|
/components/component[name]/state/name |
Component instance name. Cumulus Linux 5.19 adds the name for the ASIC and for transceivers. |
/components/component[name]/state/type |
OpenConfig component type. Cumulus Linux 5.19 adds the type for the ASIC, transceivers, fans, storage disks, and temperature and leakage sensors. |
Cumulus Linux adds the following metrics for unreachability AFI SAFI.
| Name | Description |
|---|---|
/network-instances/network-instance[name]/lldp-exception/address-families/address-family[address-family=IPV4]/state/exception-count |
Number of unreachable IPv4 prefixes the switch holds to export to LLDP, excluding the default route exception. |
/network-instances/network-instance[name]/lldp-exception/address-families/address-family[address-family=IPV4]/state/default-exception-active |
Whether the IPv4 default route exception is active. |
/network-instances/network-instance[name]/lldp-exception/address-families/address-family[address-family=IPV6]/state/exception-count |
Number of unreachable IPv6 prefixes the switch holds to export to LLDP, excluding the default route exception. |
/network-instances/network-instance[name]/lldp-exception/address-families/address-family[address-family=IPV6]/state/default-exception-active |
Whether the IPv6 default route exception is active. |
/lldp/state/counters/exception-add-total |
Number of unreachable prefixes sent to LLDP since the routing service started, including resynchronization replays. |
/lldp/state/counters/exception-remove-total |
Number of unreachable prefix withdrawals sent to LLDP since the routing service started. |
/lldp/state/counters/default-exception-add-total |
Number of default route exceptions sent to LLDP since the routing service started, counted separately for each VRF and address family. |
/lldp/state/counters/default-exception-remove-total |
Number of default route exception withdrawals sent to LLDP since the routing service started. |
/network-instances/network-instance[name]/protocols/protocol[identifier=BGP][name=BGP]/bgp/global/afi-safis/afi-safi[afi-safi-name=IPV4_UNREACH]/state/total-prefixes |
Total number of unreachable IPv4 prefixes. |
/network-instances/network-instance[name]/protocols/protocol[identifier=BGP][name=BGP]/bgp/global/afi-safis/afi-safi[afi-safi-name=IPV6_UNREACH]/state/total-prefixes |
Total number of unreachable IPv6 prefixes. |
/network-instances/network-instance[name]/protocols/protocol[identifier=BGP][name=BGP]/bgp/global/afi-safis/afi-safi[afi-safi-name=IPV4_UNREACH]/state/total-prefixes-all |
Total number of all unreachable IPv4 prefixes. |
/network-instances/network-instance[name]/protocols/protocol[identifier=BGP][name=BGP]/bgp/global/afi-safis/afi-safi[afi-safi-name=IPV6_UNREACH]/state/total-prefixes-all |
Total number of all unreachable IPv6 prefixes. |
/network-instances/network-instance[name]/protocols/protocol[identifier=BGP][name=BGP]/bgp/global/afi-safis/afi-safi[afi-safi-name=IPV4_UNREACH]/state/total-prefixes-local |
Total number of unreachable local IPv4 prefixes. |
/network-instances/network-instance[name]/protocols/protocol[identifier=BGP][name=BGP]/bgp/global/afi-safis/afi-safi[afi-safi-name=IPV6_UNREACH]/state/total-prefixes-local |
Total number of unreachable local IPv6 prefixes. |
For information about gNMI, refer to gNMI Streaming.
Updated gNMI Metrics
The switch replaces these packet trimming xPaths with new ones. If you subscribe to the old xPaths, update your subscriptions.
| Old xPath | New xPath |
|---|---|
/qos/packet-trimming/state/counters/trimmed-unicast-pkts |
/components/component[name]/integrated-circuit/pipeline-counters/packet/trim/state/trim-pkts |
/qos/interfaces/interface[interface-id]/packet-trimming/state/counters/trimmed-unicast-pkts |
/interfaces/interface[name]/state/counters/trim-pkts |
/qos/interfaces/interface[interface-id]/packet-trimming/state/counters/trimmed-tx-unicast-pkts |
/interfaces/interface[name]/state/counters/out-trim-pkts |
/qos/interfaces/interface[interface-id]/packet-trimming/output/queues/queue[name]/state/trimmed-unicast-pkts |
/qos/interfaces/interface[interface-id]/output/queues/queue[name]/state/trim-pkts |
The switch exports these xPaths on Spectrum-4, Spectrum-5, and Spectrum-6 switches, except /interfaces/interface[name]/state/counters/out-trim-pkts, which the switch exports on Spectrum-5 and later switches only.
New OTEL Metrics
The switch exports these metrics on Spectrum-6 switches only. For information about the counters behind these metrics, refer to Show and Clear Back-to-sender Notification on Link Down Counters.
| Name | Description |
|---|---|
nvswitch_qos_spxm_notify_sender_link_down_pkts |
Number of packets that matched the back-to-sender classifier on link down, both while the route is still in hardware and after FRR withdraws it. Counts notifications the switch attempted, not notifications the switch confirmed the sender received. |
nvswitch_qos_notify_sender_trimmed_link_down_pkts |
Number of those packets the switch trimmed and sent as a back-to-sender notification. |
| Name | Description |
|---|---|
nvswitch_control_plane_punt_classifier_matched_packets |
The number of packets that a control plane punt classifier rule drops. |
nvswitch_control_plane_punt_classifier_matched_bytes |
The number of bytes that a control plane punt classifier rule drops. |
Both metrics are 64-bit counters and carry four labels: rule_id, packet_type (spxm or ipv6), udp_dport (the destination port or none), and action, which is always drop. The switch exports them only when you enable the punt-classifier class. For information about the feature behind these metrics, refer to Control Plane Punt Classifier.
| Name | Description |
|---|---|
nvswitch_histogram_interface_microburst |
Per-window microburst histogram distribution. Bucket counts are exported after each poll cycle; temporality is delta. The values are the packet-delta or byte-delta per sampling window, depending on the value of quantity. |
nvswitch_histogram_interface_microburst_score |
Burstiness score σ² × μ for the current window. |
nvswitch_histogram_interface_microburst_mean_bin_index |
Mean bin index μ for the current window. |
nvswitch_histogram_interface_microburst_distribution_variance |
Bin-index variance σ² for the current window. |
nvswitch_histogram_interface_microburst_peak_value |
Peak raw watermark for the current window (packets or bytes, depending on quantity). Omitted when no valid watermark is available for the window. |
nvswitch_histogram_interface_microburst_threshold_state |
Threshold state for the configured score threshold. Omitted when no threshold is configured. |
nvswitch_histogram_interface_microburst_last_triggered |
(Unix epoch) timestamp of the last transition into triggered state. Omitted when the threshold has never triggered. |
Each microburst metric carries an interface label, a direction label (rx or tx), and a quantity label (packets or bytes). The telemetry model names this selector quantity, where the NVUE command that configures it is unit; the two names refer to the same setting. nvswitch_histogram_interface_microburst is a histogram and the rest are gauges, with threshold_state reporting 0 for clear and 1 for triggered. For information about the feature behind these metrics, refer to Microburst Histogram.
The switch exports these metrics on Spectrum-6 switches only. For information about the counters behind these metrics, refer to Packet Trimming Counters.
| Name | Description |
|---|---|
nvswitch_qos_trimmed_tx_unicast_pkts |
Number of packets that were trimmed and sent successfully. |
nvswitch_qos_trimmed_drop_unicast_pkts |
Number of packets that were trimmed and dropped. |
nvswitch_interface_trimmed_drop_unicast_pkts [interface] |
Number of packets that were trimmed and dropped on the interface. |
nvswitch_interface_tc_trimmed_tx_unicast_pkts [interface][tc] |
Number of packets that were trimmed and sent successfully on the interface and traffic class. |
nvswitch_interface_tc_trimmed_drop_unicast_pkts [interface][tc] |
Number of packets that were trimmed and dropped on the interface and traffic class. |
| Name | Description |
|---|---|
node_radius_auth_request_total |
Access-Request packets sent to the RADIUS server to authenticate an interactive user login, including retransmissions. |
node_radius_auth_accept_total |
Access-Accept messages received from the RADIUS server. |
node_radius_auth_reject_total |
Access-Reject messages received from the RADIUS server. |
node_radius_auth_timeout_total |
Access-Request messages that timed out and required retransmission. |
node_radius_auth_retransmit_total |
Retransmitted Access-Request messages. |
node_radius_auth_connection_error_total |
Login authentication attempts for which the switch could not reach the RADIUS server. |
node_radius_auth_bad_response_total |
Malformed or otherwise invalid RADIUS responses received during an interactive login. |
Each metric carries a single server label holding the RADIUS server address. For information about the counters behind these metrics, refer to Show and Clear RADIUS Counters.
Cumulus Linux adds the following metrics for unreachability AFI SAFI.
| Name | Description |
|---|---|
nvrouting_bgp_unreach_prefix_count |
Number of prefixes in the BGP unreachability table. |
nvrouting_lldp_exception_count |
Number of unreachable prefixes the switch holds to export to LLDP, excluding the default route exception. |
nvrouting_lldp_default_exception_active_count |
Whether the default route exception is active. |
nvrouting_lldp_exception_add_total |
Number of unreachable prefixes sent to LLDP since the routing service started, including resynchronization replays. |
nvrouting_lldp_exception_remove_total |
Number of unreachable prefix withdrawals sent to LLDP since the routing service started. |
nvrouting_lldp_default_exception_add_total |
Number of default route exceptions sent to LLDP since the routing service started, counted separately for each VRF and address family. |
nvrouting_lldp_default_exception_remove_total |
Number of default route exception withdrawals sent to LLDP since the routing service started. |
nvrouting_lldp_export_channel_connected |
Whether the LLDP export channel is up. |
nvrouting_lldp_export_channel_errors_total |
Number of unreachable prefix updates that failed to reach LLDP since the routing service started. |
| Name | Description |
|---|---|
nvswitch_interface_step_time_estimate |
Estimated step time, in seconds, of the AI training workload running on an interface. |
This metric is a gauge carrying a single interface label holding the interface name. The switch exports a data point only for an interface for which the algorithm produces an estimate, and exports no metric at all when it produces no estimate for any interface. For information about the feature behind this metric, refer to Step Time Estimation.
In addition to the following metrics, the switch also exports two WJH events, wjh.aggregate_drop and wjh.l1, as OTEL log records, not metrics. For more information, refer to WJH Events and Metrics Format.
| Name | Description |
|---|---|
nvswitch_wjh_total_events |
The total number of WJH drop-trap events received on the channel, before aggregation. |
nvswitch_wjh_channel_trigger |
Which triggers are enabled for each channel. |
For information about OTEL, refer to OTEL Telemetry Export.
Updated Transceiver Metric
In Cumulus Linux 5.19, the OTEL nvswitch_platform_transceiver_ethernet_pmd metric and the gNMI path /components/component[name=*]/transceiver/state/ethernet-pmd report the PMD of the transceiver media instead of the host interface:
- Optical modules report the advertised media type, such as ETH_400GBASE_DR4.
- Active optical and active copper cables report ETH_100G_AOC or ETH_100G_ACC at 100G. Other speeds report OTHER.
- Passive copper cables report the operating PMD, such as ETH_100GBASE_CR4, only while the link is up.
For CMIS transceivers (QSFP-DD and OSFP), the reported PMD comes from the first application the module advertises. If the first application is not the full rate of the module, such as a twin-port 800G module that advertises 400GAUI-4, the metric shows the lower rate. If the first application is InfiniBand, an active cable reports its PMD only while the link is up.