This is the fault that eats an hour and produces a perfectly clean meter reading. Light levels are fine at both ends, the optics pass, the patch cords are known-good — and the link stays down. The usual suspect is not a part at all: it is polarity. Somewhere between the two switches, the pair got crossed, so each transmitter is shouting into the other transmitter and neither receiver hears anything. The check takes five minutes once you stop swapping parts and start testing the pair.
A duplex fiber link is two one-way streets. Side A transmits on fiber 1 and receives on fiber 2; side B must receive on fiber 1 and transmit on fiber 2. Every patch panel, cross-connect, and adapter in between has to preserve that flip exactly once across the whole path. Cross it zero times or twice and the link works; cross it once or three times and it does not — even though every individual segment tests fine on its own.
| Path state | What the meter shows | What the link does |
|---|---|---|
| Correct polarity (odd number of flips) | Light both directions | Up |
| Crossed pair (even number of flips) | Light both directions | Down — TX into TX, RX into RX |
| Broken fiber or dirty connector | Light low or missing one direction | Down — a loss fault, not polarity |
That middle row is the trap: a meter cannot tell you polarity. It only measures power, and a crossed pair delivers full power to the wrong receiver. If your readings are healthy and the link is still down, stop measuring light and start checking the pair.
On MPO trunk links the same rule applies at cassette level, and the method matters. Method A keeps the same fiber positions end to end (straight), Method B flips them (key-up to key-down), Method C flips pairs. Mixing methods across a trunk — say a Method A cassette on a Method B trunk — produces exactly the crossed-pair symptom above, at 12 fibers at once. When a whole MPO link is dark with good light, check that every cassette and trunk in the path uses the same method before touching any optic.
A polarity fix that is not written down will be "fixed" again by the next person who re-patches the panel. The minimum useful line:
SW-04 xe-0/0/5 ↔ SW-07 xe-0/0/5 via ODF-A 14 / ODF-B 22. Loopback pass both ends (optics + ports good). Light levels normal both directions, link down → polarity suspected. Swapped duplex pair at ODF-B only → link up, 0 errors. Root cause: pair crossed at ODF-B cassette (TX→TX). Corrected at ODF-B; do NOT re-swap at ODF-A.
The shape to copy: port pair, loopback results, symptom (light good, link down), the one swap and where, link state after, root-cause panel, and the do-not-touch warning. Seven fields, and the next tech knows the pair is correct as it stands.
See RackReady Free sample Fiber no-link checklist
Support: support@hasherventures.com
Free tool: turn your polarity fix into a structured ticket note with the RackReady Ticket Builder.