CASE 014
CLOSED
Multicast
Streams dropped at the top of every hour. Every switch reported healthy.
ROOT CAUSE — IGMP querier election flapping between a redundant pair.
TECHNICAL INVESTIGATIONS
A technical investigation practice for the failures that survive vendor escalations, firmware swaps, and forum threads. Bring the evidence. Leave with the cause.
fig. 0 — a healthy stream, until it isn't.
1 typically: two vendors, three engineers, and one very long forum thread.
No retainers. No roadmaps. No transformation initiatives.
One unsolved problem at a time, worked until the mechanism of failure is proven — and shown to you in the capture, not asserted in a slide.
representative cases — details altered to protect the guilty hardware
Streams dropped at the top of every hour. Every switch reported healthy.
ROOT CAUSE — IGMP querier election flapping between a redundant pair.
Lip-sync drift measured in frames. Offsets wandered only under load.
ROOT CAUSE — transparent-clock corrections discarded by one intermediate switch.
Dante flows died on one VLAN. Only on Tuesdays.
ROOT CAUSE — a backup job saturating an uplink shared with the clock domain.
One camera chain lost genlock during live segments. Never in rehearsal.
ROOT CAUSE — a reference distribution amp failing under show-day thermal load.
The graphs said fine. The room said otherwise.
ROOT CAUSE — microbursts invisible at 30-second polling, convicted at packet timestamps.
Touch panels dropped during events. Never in testing.
ROOT CAUSE — multicast flooding when spanning tree reconverged on the show-day topology.
Evidence in. Cause out.
01 / SUBMIT EVIDENCE
Captures, configs, topology, timeline — and the list of everything already tried. The failed fixes are evidence too.
02 / ANALYSIS
Reproduce, instrument, correlate. Every claim is tested against the capture, not the data sheet.
03 / ROOT CAUSE
A proven mechanism, not a plausible story. You see the packet, the config line, the timing diagram that convicts it.
04 / RESOLUTION
The fix, applied and verified against the original failure — documented so it stays fixed after the case closes.
The firmware is current. The hardware was swapped. The vendor blamed the network, and the network team blamed the vendor. The problem is still there.
Everything you've ruled out narrows the search. That's not the end of the road — that's where a case begins.
One engineer. One open case at a time.
Twenty years across data centers, broadcast plants, stadium multicast, and casino IPTV — the rooms where failure is public and "intermittent" is not an acceptable answer. Cases are accepted only when they can be solved.
FIELD
Submit a case.
Describe the failure the way you'd describe it to another engineer. If it isn't a fit, you'll hear that quickly — with a pointer toward who might help instead.
CASE RECEIVED
Case is on the docket.
A confirmation is on its way to your inbox. If the case is a fit, the next step is a short reply with a secure destination for evidence.