Interface and Cable Troubleshooting
Most of the network is fine most of the time — and when it is not, the fault is disproportionately often at Layer 1. A single bad patch cable, a mismatched duplex setting, or a run that is just a few metres too long can produce symptoms that look like anything but a cabling problem. This lesson teaches you to read the interface counters that name the fault, and to know the tool for each layer of the problem.
Reading interface state first
Before you chase counters, read the first line of show interfaces. It reports two independent things: the interface status (Layer 1, is there a signal?) and the line protocol status (Layer 2, does the framing and keepalive work?). The pairing tells you instantly which layer to attack.
SW1# show interfaces status Port Name Status Vlan Duplex Speed Type Gi0/1 connected 10 full 1000 1000BaseTX Gi0/2 notconnect 1 auto auto 10/100/1000BaseTX Gi0/3 err-disabled 1 auto auto 10/100/1000BaseTX
| Interface | Line protocol | Meaning and first move |
|---|---|---|
| up | up | Healthy. If users still complain, the fault is higher up the stack. |
| down | down | Layer 1. No signal: cable unplugged, bad cable, dead NIC, wrong port. |
| up | down | Layer 2. Signal is present but framing/keepalive fails: encapsulation, VLAN, or protocol mismatch. |
| administratively down | down | The port was shut down in configuration; run no shutdown. |
shutdown. An exam question that hands you line protocol is down alongside interface is up is testing whether you look one layer above the cable.Collisions and late collisions
A collision happens when two devices on a shared, half-duplex segment transmit at the same time and their signals overlap. A small number of collisions is normal and expected on a healthy half-duplex segment — it is how CSMA/CD shares the medium. On a full-duplex link, however, collisions should be exactly zero, because each direction has its own path. Collisions on a full-duplex port are therefore a symptom, not normal behaviour.
A late collision is more serious. CSMA/CD can detect a collision only during the first 64 bytes (the slot time) of a frame; once a station has transmitted past that point it has effectively won the medium. A collision that arrives after the 64th byte is a late collision, and the frame is already lost. Late collisions almost always point to one of two things:
- A duplex mismatch — one side is full duplex and transmits whenever it likes, while the half-duplex side is still trying to detect collisions. This is the number-one cause.
- A cable or segment that is too long, so the far end’s signal arrives after the sender’s slot-time window has closed.
Duplex mismatch: the classic fault
A duplex mismatch occurs when the two ends of a link disagree: one is configured (or negotiated) full duplex, the other half duplex. The behaviour is asymmetric, and that asymmetry is the clue.
- The half-duplex side still listens and detects collisions, so it reports collisions and, worse, late collisions it can never clear.
- The full-duplex side transmits without listening, so when the half side talks at the same time the full side receives a mangled frame. It reports CRC/FCS errors, runts, and input errors — even though its own configuration looks correct.
The user-visible effect is severe slowness, not a dead link. The link comes up, ping may work, but throughput collapses under load because frames keep being destroyed. The fix is to make both ends agree: either set both to auto autonegotiation (preferred) or set both to the same explicit speed and duplex. Setting only one end manually is what creates the mismatch.
A duplex mismatch is asymmetric. SW2, running half duplex, sees collisions and late collisions; SW1, running full duplex, sees CRC and runt errors because it receives corrupted frames while transmitting. The cure is to make both ends autonegotiate, or to fix both ends to the same speed and duplex.
CRC/FCS errors, runts, and giants
Ethernet appends a Frame Check Sequence (FCS) — a CRC value — to every frame. The receiver recomputes it, and a mismatch means the frame was corrupted in transit. The switch drops the frame and increments a CRC (sometimes called FCS) error counter.
- CRC / FCS errors — the frame arrived damaged. Causes: a bad or damaged cable, a faulty connector, electrical interference, or a duplex mismatch. A steadily climbing CRC count is the single most useful hint that something physical is wrong.
- Runts — frames smaller than the 64-byte minimum. Causes: collision fragments on a half-duplex link, a duplex mismatch, or a failing NIC. A runt is a truncated frame, not a valid short one.
- Giants — frames larger than the maximum (1518 bytes untagged, or 1522 bytes tagged). Causes: a misbehaving NIC that streams too much, or a mismatch when one side expects jumbo frames and the other does not.
Input and output errors
Cisco interface counters group the physical errors into two buckets. Input errors are problems with frames the interface received; output errors are problems with frames it tried to transmit.
| Bucket | Counters inside | Typical cause |
|---|---|---|
| Input errors | runts, giants, CRC, frame, overrun, ignored, no buffer | Damaged cable, bad connector, EMI, duplex mismatch, NIC fault, overrun |
| Output errors | collisions, late collisions, deferred, output buffer failures | Shared/half-duplex congestion, duplex mismatch, over-driven port |
SW1# show interfaces gigabitethernet 0/1
GigabitEthernet0/1 is up, line protocol is up (connected)
Hardware is Gigabit Ethernet, address is 001b.0c00.0001
MTU 1500 bytes, BW 1000000 Kbit/sec, DLY 10 usec,
reliability 255/255, txload 1/255, rxload 1/255
Full-duplex, 1000Mb/s, media type is 10/100/1000BaseTX
47 input errors, 41 CRC, 0 frame, 0 overrun, 0 ignored
0 output errors, 0 collisions, 2 interface resets
0 late collisions, 0 deferred
SW1# show interfaces counters errors Port Align-Err FCS-Err Xmit-Err Rcv-Err UnderSize OutDiscards Gi0/1 0 41 0 47 0 0 Gi0/2 0 0 0 0 0 0
show interfaces counters errors compresses every port into one line, which makes it the fastest way to sweep a whole switch for the one misbehaving port. show controllers goes deeper into the physical chip for serial and some optical interfaces, while show interfaces transceiver reports optical transmit/receive power and temperature for SFP modules — invaluable when a fibre link’s light level is marginal.
Speed mismatch, EMI, and attenuation
Speed mismatch. If one end is hard-set to 100 Mbps and the other to 1000 Mbps, autonegotiation cannot agree. The result is usually a link that fails to come up at all (down/down) or one that comes up at the lower speed with errors. The rule is the same as for duplex: let both ends autonegotiate, or set both ends identically.
Electromagnetic interference (EMI). Motors, fluorescent ballasts, power cables, and radio transmitters can induce noise onto a copper cable. Twisted pairs twist precisely to cancel interference, and shielded twisted pair (STP) adds a foil or braid shield, but a cable routed alongside high-voltage wiring will still suffer. EMI shows up as random CRC/FCS errors that grow with the amount of nearby electrical activity.
Attenuation. Every cable weakens a signal with distance. Copper Ethernet is specified to 100 metres precisely because beyond that the signal has attenuated too far to be reliably decoded. Symptoms of over-long runs are CRC errors and intermittent link drops that worsen over time as connectors age. Fibre attenuates far more slowly, which is why it is used for long runs, but it too has distance limits and can be damaged by tight bends.
clear counters) and watch them over a few minutes before drawing a conclusion — a count that has sat unchanged for months may be ancient history.Cable problems, length limits and the tools to diagnose them
Beyond the signal itself, the physical cable is the usual suspect. The faults you must recognise:
- Incorrect pinouts. A straight-through cable connects the same pins at both ends and is used device-to-switch. A crossover cable swaps the transmit and receive pairs (wired 568A on one end, 568B on the other) and was required between like devices such as switch-to-switch. Most modern ports support auto-MDIX, which detects and corrects the pairing automatically, but older gear does not — a wrong cable there simply never links.
- Damaged connectors. A cracked RJ45 tab, a bent pin, or a poorly crimped end causes intermittent CRC errors and link flapping. Wiggling the cable to reproduce the fault is a quick test.
- Wrong cable category or rating. Cat5e and Cat6 support gigabit to 100 m; older or damaged cable may not. Mixing categories rarely breaks a link outright but can cause CRC errors at higher speeds.
- Length limits. Copper Ethernet: 100 m. Keep the horizontal run plus patch cords under that. Fibre has far greater ranges that depend on the fibre type (multimode vs single-mode) and the transceiver.
The right tool matches the layer you are testing. Match each symptom to its instrument:
| Tool / command | What it does | Use it for |
|---|---|---|
show interfaces | Status, line protocol, speed/duplex, and all counters | First look at any interface problem |
show interfaces counters errors | One line per port of error totals | Sweeping the switch for the faulty port |
show controllers | Physical-layer chip detail (cable, clocking, signalling) | Serial/optical physical faults |
show interfaces transceiver | Optical power, temperature, and diagnostics | Marginal or failing fibre links |
| Cable tester / wiremap | Checks each pair, continuity, and pinout | Bad crimps, wrong pinouts, opens and shorts |
| Time-domain reflectometer (TDR) | Sends a pulse and times the reflection | Locating a break or the distance to a fault in copper |
| OTDR (optical TDR) | The fibre equivalent of a TDR | Locating breaks, splices, and bends in fibre |
| Loopback plug | Loops transmit back to receive | Testing a single port in isolation |
| Toner and probe | Traces a cable through a bundle by injected tone | Identifying which cable is which in a patch panel |
show interfaces, then show interfaces counters errors, then swap the patch cable. Swapping a $2 cable solves more tickets than any console session.Key terms
- Line protocol — the Layer 2 state of an interface; “up/down” points to a Layer 2 fault.
- Collision — two devices transmitting at once on a shared half-duplex segment; normal in small numbers there, but wrong on full duplex.
- Late collision — a collision after the first 64 bytes; the fingerprint of a duplex mismatch or an over-long cable segment.
- Duplex mismatch — one end full duplex, the other half duplex; causes collisions/late collisions on one side and CRC/runts on the other.
- CRC / FCS error — a frame that failed its checksum and was dropped; indicates corruption in transit.
- Runt — a frame shorter than 64 bytes; a fragment from a collision or a fault.
- Giant — a frame larger than the maximum (1518/1522 bytes).
- Attenuation — the weakening of a signal over distance; the reason copper Ethernet is limited to 100 m.
- Auto-MDIX — a feature that detects and corrects straight-through versus crossover wiring automatically.
- TDR / OTDR — time-domain reflectometer / optical TDR; locates the distance to a break in copper or fibre.
Review
Check what you have learned with these review questions.
Sources & references
- Cisco. Troubleshooting Ethernet and Interface Errors (collisions, late collisions, CRC, runts, giants). cisco.com.
- Cisco. CCNA (200-301) Exam Topics, Network Fundamentals — troubleshoot interfaces and cabling. cisco.com.
- IEEE. IEEE 802.3: Ethernet (frame size, FCS, and CSMA/CD). standards.ieee.org.
- Cisco. CCNA: Introduction to Networks — “Network Media” and “Network Troubleshooting”. netacad.com.
Target exams: CCNA 200-301 v1.1 · CyberOps Associate 200-201. Last content verification: 2026-10-11.