Link flapping, CRC & dirty connectors

A link that bounces up and down, or a port whose CRC and receive-error counters keep climbing, is most often not a failing module and not a broken fibre. It is a marginal optical budget — and the cheapest place to find lost decibels is the connector face.

The economics of contamination

• A single dust particle of 2–5 µm on a connector end-face adds 0.5–2.0 dB of loss — enough to push a marginal link over its budget.
• A clean LC pair contributes a stable 0.1–0.2 dB.

Cleaning is therefore "the cheapest decibel you will ever recover".

Symptoms

• Link flapping — repeated up/down transitions in the log.
• Rising CRC, input errors or receive errors on the port counters.
• Rx power lower than the installation record, or lower on one end than the other.
• A link that fails only under load or at certain temperatures.

The diagnostic ladder

Operators converge on the same sequence:

1. Read the levels on both ends. Compare Tx/Rx with the link passport taken at installation. Look for asymmetry and for DDM thresholds being crossed — see DDM levels.
2. Inspect → clean → re-inspect. Use a fibre inspection scope, then a one-click cleaner or a reel-type cassette cleaner. After any wet cleaning, finish with a dry pass. Never plug in a connector you have not inspected.
3. Reseat, then swap the module. Re-insert the transceiver. If the fault persists, replace it with a known-good module. Field pragmatism: don't spend an hour on a patch cord — clean it once, and if it still fails, replace it.
4. OTDR from both directions. Only now suspect the fibre plant. Ask the provider for a trace. After a fibre repair (a new splice), total loss is frequently higher than before — a link that used to run on 10 km-rated optics may no longer reach.

Why order matters

Each step is cheaper and faster than the next. Swapping a module before cleaning hides the real cause (a dirty patch will dirty the new module too); calling for OTDR before cleaning wastes a technician's visit.

In CodingBox

Use the DDM screen to read a suspect module on the bench and compare its Rx against the same module in the field; log values to CSV to build the passport. If the module itself is fine on the bench, the problem is in the plant or the connectors, not in memory.

On 25G+ links the early warning is the pre-FEC BER, not the Rx power — how to read VDM and host FEC counters: VDM & FEC metrics.

What a flapping optic does to spanning tree, LAG hashing, routing adjacencies and overlays — and how to read those symptoms back to the port: Forwarding basics.

The plant side of intermittent links — contaminated and damaged endfaces, bends, reflectance, water in closures — with OTDR and DDM signatures: Fibre plant faults.

Reflections as a cause of instability — Fresnel physics, typical reflectance per component, laser sensitivity, MPI, what the standards require and how to measure it: Reflections & return loss.