Identify and Monitor Internal Drops, Errors, and Exceptions

Interface Errors

The first step is usually to check interface statistics. Use the extensive option to have the full view.

To identify relevant counters, review the Preclassifier statistics section at the end of the output. This stage occurs early in the ingress pipeline and classifies packets as either network‑control traffic (destined for the Routing Engine or transiting) or best‑effort traffic.

Note:

The real-time field is not used and should remain at 0. This classification is performed per packet based on selected Layer 2 through Layer 4 fields. The final determination of whether traffic is destined for the control plane occurs later during the lookup stage. This early classification helps protect control plane traffic during periods of ingress oversubscription.

Check Layer 1 and Layer 2 statistics next. Review the following counters:

• PCS drops—Indicate possible optical signal degradation
• FEC uncorrected errors—Identify errors that FEC cannot correct when FEC is enabled
• Ethernet CRC errors—Indicate frame corruption
• MAC pause frames—Indicate congestion
• Oversized frames—Identify frames that exceed the maximum size
• Input DA rejects—Indicate frames with an unexpected destination MAC address or untagged frames received on a tagged interface

Use these counters to identify physical or data link layer issues.

Review overall input and output error counters. Focus on the following:

• Input resource errors—Indicate possible resource exhaustion in the ingress lookup engine (high packets per second)
• Output errors—Typically indicate congestion conditions

Depending on the class‑of‑service configuration and traffic profile, some queues might experience tail drop or RED drop. These drops are reflected in the Dropped packets counter.

Use a command to view per‑queue drop statistics. For each queue, monitor tail drop and RED drop counters in real time, along with queue depth (see Queue-depth bytes).

Under normal conditions, even near line rate, the current queue depth should remain close to 0.To determine if a sustained increase in queue depth is causing congestion, packet buffering, or latency, compare the current queue depth against the maximum configured buffer size, see the following snippet:

PFE Drops and Exceptions

Drops can also occur during the forwarding process, not only at the interface level. First, check PFE exception counters. The lookup engine marks packets as exceptions, whether they are control plane or transit traffic.

Run the following command multiple times to identify counters that increment. Based on the ingress interface, review the output for the corresponding FPC (see the Slot field) and focus on counters labeled DISC.

Use the show pfe statistics traffic fpc <slot> command to view traffic statistics for a specific FPC. Apply filters to display drop counters only.

You can typically ignore the Normal discard counter, because it represents expected internal drops that do not affect transit traffic.

Review the Packet Forwarding Engine local traffic statistics section for punted traffic errors. These errors occur when packets are sent to the line card CPU or Routing Engine for additional processing. Drops in this section can indicate host path congestion or overload on the internal path to the CPU.

Use the show pfe statistics traffic command to view per‑PFE details. Review the fpc and pfe fields to identify the FPC slot and PFE ID.

The number of output sections varies by line card model. Each section corresponds to an FPC and PFE pair.

Use additional CLI commands to analyze drop conditions as needed. The commands described in this section address most common drop scenarios. For advanced troubleshooting, contact JTAC support.