Incident Library — Graded Troubleshooting

Sixteen optical-transport incidents, each played the way a real call goes: you get a brief and nothing else. You gather directional evidence, you rank hypotheses, you commit to a root cause and a next action — and only then does the engine reveal whether your reasoning held up. It grades not just the answer but the method: did you scope the blast radius, did you get the discriminating reading before committing, did you avoid a destructive move before you localised, and did you do it efficiently. The point is not to memorise sixteen answers; it is to drill the habit that solves the seventeenth.

The method Every case is worked in the same disciplined order. Internalise the loop and any optical fault becomes tractable:

Scope the blast radius. One wavelength, one span, or one site? This single question routes everything after it — service/transponder vs fiber/line-system vs power/facility.
Gather directional evidence. Faults are directional. Compare local Rx against remote Tx, presence against quality, near-end against far-end — until you find the hop where healthy becomes broken.
Rank hypotheses. Hold several candidates at once and let the evidence sort them, rather than falling in love with the first guess.
Choose the next action / escalation. The correct move is often get one more reading or hand back across the demarc — not swap the card.
Write the root-cause report. Name the fault, the evidence that proves it, and the fix.

Scope note Every alarm name, PM counter, CLI fragment, and dBm reading in these cases is representative / illustrative — chosen to teach the reasoning, not to reproduce any specific Ekinops release. Screen labels, exact thresholds, and menu paths change between products and versions; verify exact behaviour against current Ekinops / Celestis documentation. What transfers is the diagnostic loop, not the literal strings.

The incident engine

Pick a case, work it, and get a graded scorecard. Your progress ticks the checklist below as you close cases correctly.

How this is graded

Each case is scored out of 100 across four dimensions, then mapped to a letter grade. The weighting deliberately rewards method over luck — you can name the right root cause and still lose points if you guessed it without the discriminating evidence, or if you reached for a destructive action first.

Dimension	Weight	What earns the points
Root cause	45	You selected the correct root cause. This is necessary but not sufficient — the other 55 points police how you got there.
Discriminating evidence	30	Before committing, you scoped the blast radius and gathered the specific reading that separates the real cause from its look-alikes (e.g. comparing local Rx to remote Tx to fix direction).
Safety / sequencing	15	You did not take a destructive or hardware action (RMA a card, pull a fiber, force a protection switch) before localising the fault. Acting blind is penalised even when the guess happens to be right.
Efficiency	10	You gathered evidence close to the case's `optimal` step count. Thrashing through every action or committing on too little both cost points.

Letter bands: A 90+, B 80–89, C 70–79, D 60–69, F below 60. PASS/AMBER/FAIL colour cues flag each dimension at a glance: green = clean, amber = did it but late or inefficient, red = missed or unsafe.

Field-method summary The muscle memory these cases build, distilled to a card you can carry:

Scope the blast radius first — one wave, one span, or one site decides the whole investigation
Fix the direction — compare local Rx to remote Tx before you dispatch anyone to the glass
Separate presence (is there light?) from quality (is the light good?) at every hop
Localise before you touch — never RMA, re-patch, or force-switch on a hunch
Prove which side of the demarc owns the fault before handing it back
Close with a root-cause packet: fault, the evidence that proves it, and the fix

Case progress

These tick automatically as you close each case correctly in the engine above; you can also mark them by hand.