CCNA 200-301

CCNP Enterprise

CCNP Security

CCIE Enterprise Lab

CCIE Security Lab

CCNP Service Provider

CCNP Data Center

CCNP Collaboration

CCIE DC Lab

Diagnosing degraded network performance, sluggish application response times, or choppy VoIP calls often starts at the switching layer. Unlike full link outages that immediately trigger down alerts, silent packet drops on a switch port can throttle TCP throughput while every interface still displays an up/up status. Because a Cisco switch can drop frames at the physical transceiver, ingress ring buffer, switching ASIC, or egress QoS queue, isolating the exact point of failure requires a structured diagnostic workflow.

Checking packet loss on a Cisco switch involves inspecting port-level hardware counters via the CLI, interpreting specific error and discard metrics, running active probe tests for intermittent transit drops, and remediating underlying physical or buffer constraints.

Step-by-Step Guide to Checking Interface Packet Loss via the CLI

Inspecting port-level counters in Cisco IOS, IOS-XE, or NX-OS is the fastest way to verify whether a switch is actively dropping frames. Follow this sequential workflow to isolate problematic ports without being misled by historical data.

  1. Verify and Reset Stale Interface Counters
    By default, Cisco switch interface counters accumulate continuously since the last system reboot or manual reset. Before diagnosing an active complaint, verify when the counters were last cleared and reset them to establish a clean baseline:
    ios
    Switch# show interfaces gigabitethernet1/0/1 | include Last clearing
    Last clearing of "show interface" counters never
    Switch# clear counters gigabitethernet1/0/1
    Clear "show interface" counters on this interface [confirm] y

    Wait 5 to 15 minutes (or reproduce the user’s traffic flow) before re-inspecting the port so that any incrementing numbers reflect current network conditions.

  2. Scan All Switch Ports for Errors and Discards
    If you do not know which specific port is dropping packets, run the switch-wide error matrix command to view physical errors and output discards across all interfaces simultaneously:
    “`ios
    Switch# show interfaces counters errors

Port Align-Err FCS-Err Xmit-Err Rcv-Err UnderSize OutDiscards
Gi1/0/1 0 0 0 0 0 0
Gi1/0/2 0 1420 0 1420 84 0
Gi1/0/24 0 0 0 0 0 48912
``
In this view, non-zero columns immediately point you to the faulty ports—such as
Gi1/0/2experiencing physical Receive/FCS errors orGi1/0/24experiencing heavyOutDiscards`.

  1. Inspect Detailed Drop and Error Counters on a Target Port
    Once you identify a suspect interface, inspect its granular packet loss statistics using show interfaces <interface-id>:
    ios
    Switch# show interfaces gigabitethernet1/0/24
    GigabitEthernet1/0/24 is up, line protocol is up (connected)
    Hardware is Gigabit Ethernet, address is 0050.56bf.1a24 (bia 0050.56bf.1a24)
    MTU 1500 bytes, BW 1000000 Kbit/sec, DLY 10 usec,
    reliability 255/255, txload 185/255, rxload 42/255
    Full-duplex, 1000Mb/s, media type is 10/100/1000BaseTX
    Input queue: 0/2000/0/0 (size/max/drops/flushes); Total output drops: 48912
    Queueing strategy: fifo
    Output queue: 0/40 (size/max)
    5 minute input rate 165420000 bits/sec, 24110 packets/sec
    5 minute output rate 728910000 bits/sec, 89400 packets/sec
    14589210 packets input, 9841204112 bytes, 0 no buffer
    Received 41210 broadcasts (38102 multicasts)
    0 runts, 0 giants, 0 throttles
    0 input errors, 0 CRC, 0 frame, 0 overrun, 0 ignored
    89412500 packets output, 114201948200 bytes, 0 underruns
    0 output errors, 0 collisions, 2 interface resets

    To filter this output across multiple interfaces and display only lines relevant to packet loss, use a regular expression pipe:
    ios
    Switch# show interfaces | include line protocol|drops|errors

  2. Check Real-Time Rate and Queue Counters on NX-OS
    On Cisco Nexus switches running NX-OS, standard show interface output follows a slightly different layout. Use the detailed counter view or queuing view to isolate per-class drops:
    nxos
    Nexus# show interface ethernet1/1 counters detailed
    Nexus# show queuing interface ethernet1/1

    Key Takeaway: Always clear historical counters first, use show interfaces counters errors to triage the whole switch at a glance, and drill down into show interfaces <port> to separate ingress errors from egress drops.


Interpreting Cisco Switch Interface Counters: Input Errors vs. Output Drops

Finding non-zero counters in the CLI output is only half the diagnostic process; understanding what each metric represents tells you whether the packet loss stems from damaged cabling, a configuration mismatch, or traffic congestion.

What is the difference between input errors and output drops on a Cisco switch?

Input errors occur on ingress before the switch forwards the frame. They indicate that the switch received a physically corrupted, malformed, or truncated Ethernet frame from the connected device or cable and had to discard it. Output drops (Total output drops or OutDiscards) occur on egress after the switching ASIC has already made a valid forwarding decision. Output drops mean the frame itself was healthy, but the outbound port’s transmit queue buffer was full due to bandwidth congestion or bursty traffic.

What do runts, giants, and CRC errors indicate?

  • Runts: Ethernet frames smaller than the minimum valid 64-byte IEEE 802.3 frame size with a bad FCS. Runts are most commonly caused by half-duplex collisions, duplex mismatches, or severe electrical interference on copper cabling.
  • Giants: Ethernet frames that exceed the interface’s configured maximum transmission unit (standard 1518 bytes untagged or 1522 bytes with an 802.1Q VLAN tag) and also contain a bad FCS. They typically occur when a connected host or upstream trunk sends jumbo frames (e.g., 9000 MTU) into an access or trunk port still set to a 1500-byte MTU.
  • CRC (Cyclic Redundancy Check) / FCS-Err: Frames that arrived with a calculated checksum that does not match the sender’s FCS trailer. Rising CRC errors point directly to Layer 1 faults: damaged patch cables, dirty fiber optic end-faces, failing SFP/SFP+ transceivers, electromagnetic interference (EMI), or duplex mismatches.

How can you tell if packet loss is caused by a duplex mismatch?

A duplex mismatch happens when one link partner is manually hardcoded (for example, speed 100 and duplex full) while the other is set to auto. Because the hardcoded end disables auto-negotiation pulses, the auto-negotiating end falls back to 100 Mbps Half-Duplex per the IEEE standard. This creates a unmistakable counter signature:
* The Full-Duplex side transmits whenever it wants, causing frames to collide in transit, and logs climbing input errors, CRC errors, and runts.
* The Half-Duplex side waits for carrier sense and detects frames arriving while it is transmitting, logging climbing late collisions, deferred transmissions, and output errors.

Why does an interface show high output drops when 5-minute bandwidth utilization is low?

Cisco switches calculate interface bandwidth rates (5 minute output rate) as a rolling average over a default 300-second load-interval. However, switch egress buffers are measured in milliseconds of memory. Microbursts occur when traffic spikes to line rate for just a few milliseconds—most commonly when a 10 Gbps or 25 Gbps uplink forwards a burst of storage or backup traffic out of a 1 Gbps access port. The 1 Gbps port’s egress queue overflows instantly and drops packets, while the 5-minute average utilization still reports under 15%.

What causes input queue drops and “no buffer” counters to increment?

Unlike output drops, input queue drops and no buffer increments mean the switch could not process incoming frames fast enough. On modern hardware-switched platforms, this rarely happens for transit data-plane traffic unless ingress hardware rings are saturated; instead, it indicates excessive control-plane “punt” traffic—such as ARP storms, STP topology floods, TTL=1 packets, or multicast join floods—overwhelming the internal CPU-bound ingress queues.

Key Takeaway: Treat rising input errors, CRC, and runts as Layer 1 or duplex problems, and treat rising Total output drops as egress bandwidth contention or microburst buffer exhaustion.


CLI Templates for Testing Intermittent Packet Loss (Extended Ping and IP SLA)

When local interface counters look clean—or when packet loss only occurs intermittently across a multi-hop Layer 3 switched path—passive counter checks are not enough. Use the following CLI templates directly on your Cisco switch to actively generate test probes and measure live packet loss percentages.

1. Extended Ping Stress Test Template (IOS / IOS-XE)

Use extended ping with a high packet count, a shortened 1-second timeout, a specified source VLAN/interface, and a 1500-byte payload with the Don’t Fragment (DF) bit set to uncover both intermittent drops and hidden MTU bottlenecks:

Switch# ping
Protocol [ip]: ip
Target IP address: 10.20.30.1
Repeat count [5]: 1000
Datagram size [100]: 1500
Timeout in seconds [2]: 1
Extended commands [n]: y
Source address or interface: Vlan10
Type of service [0]: 0
Set DF bit in IP header? [no]: yes
Validate reply data? [no]: yes
Data pattern [0x0000]: 0xABCD
Loose, Strict, Record, Timestamp, Verbose[none]: none
Sweep range of sizes [n]: n

2. Continuous IP SLA ICMP-Echo Probe Template

To track intermittent packet loss over hours or days without keeping an SSH session open, configure a background IP SLA icmp-echo probe:

! Configure the IP SLA ICMP probe
Switch(config)# ip sla 10
Switch(config-ip-sla)# icmp-echo 10.20.30.1 source-interface Vlan10
Switch(config-ip-sla-echo)# request-data-size 1400
Switch(config-ip-sla-echo)# timeout 1000
Switch(config-ip-sla-echo)# frequency 5
Switch(config-ip-sla-echo)# exit

! Schedule the probe to run continuously starting immediately
Switch(config)# ip sla schedule 10 life forever start-time now
Switch(config)# end

! Verify accumulated packet loss and round-trip latency
Switch# show ip sla statistics 10

3. Bidirectional IP SLA UDP-Jitter Template (Separating Upstream vs. Downstream Drops)

Standard ICMP pings only report round-trip failure, leaving you guessing whether packets were dropped on the forward path or the return path. Configure an ip sla responder on the target Cisco switch and run a udp-jitter probe to measure directional loss (Packet Loss SD vs. Packet Loss DS):

! --- ON REMOTE TARGET SWITCH (10.20.30.1) ---
Remote-Switch(config)# ip sla responder
Remote-Switch(config)# exit

! --- ON SOURCE SWITCH ---
Source-Switch(config)# ip sla 20
Source-Switch(config-ip-sla)# udp-jitter 10.20.30.1 16384 source-ip 10.10.10.1 num-packets 100 interval 20
Source-Switch(config-ip-sla-jitter)# frequency 60
Source-Switch(config-ip-sla-jitter)# exit
Source-Switch(config)# ip sla schedule 20 life forever start-time now
Source-Switch(config)# end

! Check directional packet loss (SD = Source to Destination, DS = Destination to Source)
Source-Switch# show ip sla statistics 20 | include Packet Loss|RTT

Key Takeaway: Use extended pings with the DF bit for immediate MTU and link stress testing, and deploy IP SLA udp-jitter probes when you need to isolate which direction of a full-duplex path is dropping packets.


How to Fix Output Drops, Buffer Exhaustion, and Physical Layer Packet Loss

Once you have pinpointed the faulty port and verified whether the drops are occurring at ingress (physical/duplex) or egress (queue buffer starvation), apply the matching remediation steps below.

  1. Resolve Physical Layer and Duplex Mismatches (For Input Errors, CRC, and Runts)
  2. Standardize Auto-Negotiation: Check the port configuration (show running-config interface <id>) and the connected host settings. Set both link partners to auto-negotiate speed and duplex unless hardware restrictions dictate otherwise:
    ios
    Switch(config)# interface gigabitethernet1/0/2
    Switch(config-if)# speed auto
    Switch(config-if)# duplex auto
  3. Verify Optical Transceiver Power Levels (DOM): On fiber links logging CRC or input errors, inspect Digital Optical Monitoring (DOM) metrics to check if Rx/Tx optical power (in dBm) falls outside the module’s warning or alarm thresholds:
    ios
    Switch# show interfaces gigabitethernet1/1/1 transceiver detail

    If receive power is below threshold, clean the fiber patch cord connectors with an optical cleaner, swap the fiber patch lead, or replace the degraded SFP/SFP+ module.

  4. Expand Egress QoS Softmax Buffers on Cisco Catalyst 9000 Switches (IOS-XE)
    On Catalyst 9200, 9300, and 9500 Series switches, microbursts commonly trigger output drops on 1G downlinks even when total switch memory is underutilized. You can permit congested queues to borrow additional dynamic shared buffer memory by increasing the qos queue-softmax-multiplier from its default value (100) up to 1200:
    ios
    Switch(config)# qos queue-softmax-multiplier 400
    Switch(config)# end
    Switch# show platform hardware fed switch active qos queue config interface gigabitethernet1/0/24

    Raising the multiplier to 400 or 800 quadruples or octuples the dynamic buffer ceiling available to absorb transient microbursts before tail-drop occurs.

  5. Rebalance Queue-Set Buffers on Legacy Catalyst Switches (2960-X / 3750-X / 3850)
    On older Catalyst switches using MLS QoS (mls qos), enabling QoS often carves port buffers into restrictive partitions where the default traffic queue (Queue 2) only receives a small fraction of buffer space, while unused voice/priority queues hoard reserved memory. Reallocate buffer ratios and increase the maximum reserve threshold on the active queue-set:
    ios
    ! Allocate 70% of buffer memory to Queue 2 (default best-effort queue)
    Switch(config)# mls qos queue-set output 1 buffers 10 70 10 10
    ! Allow Queue 2 to borrow up to 400% of its reserved threshold from the common pool
    Switch(config)# mls qos queue-set output 1 threshold 2 200 200 100 400

    On Catalyst 3650/3850 platforms using MQC policy-maps, adjust the queue-buffers ratio within your egress queuing policy-map instead.

  6. Mitigate Persistent Speed Mismatches and Oversubscription
    Buffer tuning only resolves transient microbursts lasting milliseconds; it cannot fix sustained bandwidth oversubscription where 10 Gbps of continuous traffic attempts to exit a 1 Gbps port. For sustained congestion:

  7. Bundle multiple physical links into an LACP EtherChannel (channel-group <number> mode active) to increase aggregate egress bandwidth.
  8. Apply traffic shaping (shape average) on bursty upstream senders so traffic is smoothed before hitting lower-speed access ports.

Key Takeaway: Fix physical cabling and auto-negotiation to stop ingress errors, expand dynamic QoS queue buffers (qos queue-softmax-multiplier or queue-set) to absorb microbursts, and upgrade or bundle links when sustained traffic exceeds port capacity.


Official Cisco Platform Architecture and Hardware Diagnostic Reference

When standard interface counters do not provide enough visibility into internal ASIC drops, administrators must query hardware-specific forwarding engines. Cisco switching families use distinct ASIC architectures, operating systems, and official documentation trees for diagnosing packet drops.

Cisco Catalyst 9000 Series (IOS-XE / UADP ASIC Architecture)

  • Architecture Definition: Catalyst 9200, 9300, 9400, and 9500 switches run Cisco IOS-XE over the Unified Access Data Plane (UADP) or Silicon One ASIC. Packet buffering is managed by the Forwarding Engine Driver (FED) across two shared buffer pools (2 queues on strict priority and 6 queues on normal default sharing, depending on the QoS policy).
  • Hardware-Level Diagnostic Commands:
  • show platform hardware fed switch active qos queue stats interface <interface-id> — Displays exact per-queue enqueue and drop counters in both packets and bytes at the UADP ASIC level.
  • show platform hardware fed switch active fwd-asic drops exceptions — Displays internal ASIC exception drops (such as VLAN spanning-tree blocks, ACL drops, or L3 MTU failures) that do not appear as standard interface output drops.
  • Official Cisco Documentation Reference: Troubleshoot Output Drops on Catalyst 9000 Switches and Cisco Catalyst 9000 Series Switching Platform QoS and Buffer Architecture White Paper (located under Cisco Support > Switches > Catalyst 9300/9500 Series Switches > Troubleshoot and Alerts).

Cisco Nexus 9000 Series (NX-OS / Cloud Scale ASIC Architecture)

  • Architecture Definition: Cisco Nexus 9300 and 9500 data center switches run NX-OS on Cisco Cloud Scale ASICs (Tahoe/Sugarbowl/Homewood families). Buffers are allocated per slice across ingress and egress port groups (qos-group 0 through qos-group 7), supporting lossless RoCEv2/No-Drop classes alongside tail-drop and WRED classes.
  • Hardware-Level Diagnostic Commands:
  • show queuing interface ethernet <slot/port> — Reports per-qos-group unicast, multicast, and span egress discard counters and peak buffer occupancy waters-marks.
  • show hardware internal errors — Queries all ASIC slices for internal drop causes, parity errors, and forwarding lookup drops.
  • show system internal qos Buffer-stats detail — Inspects real-time shared pool cell consumption across ASIC slices.
  • Official Cisco Documentation Reference: Cisco Nexus 9000 Series NX-OS Troubleshooting Guide (Chapter: Troubleshooting Packet Drops) and Cisco Nexus 9000 Series NX-OS Quality of Service Configuration Guide.

Platform Command Matrix: Catalyst IOS-XE vs. Nexus NX-OS

Diagnostic ObjectiveCisco Catalyst (IOS-XE) CommandCisco Nexus (NX-OS) Command
Port Error & Discard Summaryshow interfaces counters errorsshow interface counters errors
Per-Queue Egress Dropsshow platform hardware fed switch active qos queue stats interface <id>show queuing interface <id>
Internal ASIC Drop Reasonsshow platform hardware fed switch active fwd-asic drops exceptionsshow hardware internal errors
Control Plane (CPU Punt) Dropsshow platform software fed switch active punt cpuq allshow policy-map interface control-plane

Key Takeaway: Use standard show interfaces commands for initial triage, then transition to show platform hardware fed on Catalyst IOS-XE or show queuing interface on Nexus NX-OS to inspect per-queue ASIC drops.


Best SNMP Network Monitoring Tools for Tracking Cisco Switch Packet Loss

While CLI commands are essential for live troubleshooting, manually checking hundreds of switch ports cannot catch transient packet drops that happen overnight. Network monitoring platforms automate this process by polling standard SNMP MIBs—specifically IF-MIB (ifInErrors, ifInDiscards, ifOutErrors, ifOutDiscards), CISCO-CLASS-BASED-QOS-MIB for per-queue drops, and CISCO-RTTMON-MIB for IP SLA statistics—to graph historical drop rates and fire alerts automatically.

  1. SolarWinds Network Performance Monitor (NPM)
    SolarWinds NPM provides deep out-of-the-box integration for Cisco Catalyst and Nexus environments. Beyond polling standard IF-MIB error and discard counters, NPM natively discovers duplex mismatches by cross-referencing CDP/LLDP neighbor states with error signatures, polls Cisco Class-Based QoS (CBQoS) to chart per-queue drops, and integrates with SolarWinds VoIP & Network Quality Manager (VNQM) to automate and graph Cisco IP SLA jitter and packet loss probes.

  2. Paessler PRTG Network Monitor
    PRTG uses a sensor-based architecture featuring dedicated SNMP Traffic, SNMP Cisco System Health, and SNMP Cisco IP SLA sensors. Administrators can enable “Discards” and “Errors” channels inside every switch port’s SNMP Traffic sensor and configure granular threshold triggers (for example, alerting via Slack or PagerDuty whenever ifOutDiscards exceeds 50 packets per second for two consecutive polling intervals).

  3. LibreNMS
    For teams seeking an open-source, zero-licensing-cost solution, LibreNMS offers automated network discovery via CDP, LLDP, and ARP alongside comprehensive native support for Cisco IOS, IOS-XE, and NX-OS MIBs. It automatically generates RRDtool time-series graphs for interface errors, discards, and DOM optical transceiver power levels, and includes a flexible rule builder to alert on sudden spikes in CRC errors or output drops.

  4. Zabbix
    Zabbix is an enterprise-grade open-source monitoring platform that ships with official Cisco IOS by SNMP and Cisco Nexus 9000 Series by SNMP templates. These templates calculate rate-of-change triggers on ifInDiscards, ifOutDiscards, ifInErrors, and ifOutErrors, automatically flagging interfaces where the packet drop percentage exceeds a customizable baseline relative to total packet throughput.

  5. LogicMonitor
    LogicMonitor is a cloud-based observability platform with extensive pre-built Cisco LogicModules. In addition to tracking standard interface error rates, LogicMonitor automatically discovers Cisco QoS policy-maps and hardware environmental sensors, correlating switch buffer exhaustion with upstream router or firewall utilization so teams can pinpoint whether packet loss is localized to a single access switch or cascading across a distribution tier.

Key Takeaway: Deploy an SNMP monitoring tool that polls both IF-MIB (ifOutDiscards/ifInErrors) and CISCO-CLASS-BASED-QOS-MIB at 1-to-5-minute intervals so you can correlate exact packet drop timestamps with application traffic spikes.


Conclusion

Checking and resolving packet loss on a Cisco switch requires following the lifecycle of a frame from the physical cable through the internal ASIC buffers. By resetting stale counters and running show interfaces counters errors and show interfaces <id>, you can immediately separate Layer 1 input errors (such as CRC faults, runts, and duplex mismatches) from egress congestion (Total output drops). Pairing passive counter inspection with active Extended Ping or IP SLA probes, tuning dynamic QoS queue buffers (qos queue-softmax-multiplier), and maintaining continuous SNMP visibility ensures silent packet drops are identified and eliminated before they impact end-user applications.

Please follow and like us:
Last modified: November 17, 2026

Author

Comments

Write a Reply or Comment

Your email address will not be published.