FR
live

Two HPE Juniper routers were enough to ground all US flights for four hours

On Friday 13 November 2026, two HPE Juniper routers installed by contractor L3Harris began hammering the FAA’s aging copper network at a thousand connections every four to five seconds, and a nationwide ground stop froze American aviation. It was neither a cyberattack nor, strictly speaking, a hardware failure: it is a continent-scale demonstration of what a single point of failure on legacy infrastructure costs.

A stripped copper cable snapped on a telephone-pole crossarm, its two frayed ends almost touching, a single amber spark at the break.

13 November 2026. American aviation stops. On 14 November, Transportation Secretary Sean Duffy points at an upgrade job. On 15 November, the investigation narrows the culprit to two HPE Juniper routers installed at two rural sites by contractor L3Harris. Why it matters: the longest nationwide ground stop since 11 September 2001 was not caused by a cyberattack but by two network devices that, by hammering a copper network several decades old, brought down a country’s entire air traffic. This is not an aviation story; it is a textbook case in network resilience.

The mechanism: a thousand connections every four to five seconds

The first alarm came at Indianapolis Center, which watches the airspace over northern Kentucky, southern Indiana, southern Ohio, and part of West Virginia. Controllers could no longer reach flight crews. The agency then ran a nationwide hunt and found two possible sources: an FAA radio site in El Dorado, Arkansas, and a second site in Texas.

FAA technology chief Rebecca Guy described the symptom with a precision that explains the whole failure: “we saw two places that were actually pinging our network at a very high rate, a thousand connections over every four to five seconds.” In engineering terms, that is a connection flood — a volume any modern network would absorb without flinching, but not a copper network designed decades ago.

The FAA network is mid-migration to fiber, a project costing taxpayers billions of dollars. Until that migration completes, the legacy copper remains in service, and it is the copper that buckled under the load. FAA administrator Bryan Bedford had already raised concerns about the risks of this transition period.

What the failure was not

Two points need to be settled up front, because they change the diagnosis.

It was not a cyberattack. Sean Duffy said so without ambiguity: no sign of foreign intrusion. The scenario is not an attacker flooding the network, but a device doing it all by itself.

It was not, strictly speaking, a hardware failure. HPE pushed back publicly: the routers “did not fail,” and the feature in question “worked as intended.” That is the most instructive part of the whole affair. A system can be built from components that all work “as intended” and still fail, the moment their interaction exceeds the capacity of a link older than they are. The router did its job; the network did not hold.

The TCAS collision-avoidance system, which runs on hardware aboard the aircraft and talks plane-to-plane, was unaffected. But losing communication with high-flying aircraft made it impossible to coordinate the thousands of planes aloft at any moment, forcing the agency to stop traffic.

Four hours to restore, days to absorb

It took roughly four hours for the FAA to bring its communications network back to normal. That number is misleading: it measures only the network recovery, not aviation’s return to normal. Passengers on thousands of canceled and diverted flights had to be rebooked while new passengers kept arriving for their scheduled departures. The closest reference point is the CrowdStrike crash of 2024, after which airlines needed days to clear the backlog.

The political fallout was immediate. New York governor Kathy Hochul called for Sean Duffy’s resignation. Donald Trump broke his silence and demanded funding for the FAA. Sean Duffy himself set the operational goal: “we have to get off old network.” The phrasing is right, but it moves the question: getting off copper will take years, and through the whole transition the two worlds will have to coexist.

The lesson for any critical network

The FAA incident reads like a manual, and its three lessons apply to any infrastructure that cannot afford to stop.

A single point of failure is still a single point of failure, even when it looks redundant on paper. Two sites were implicated, but the underlying copper network was itself a single path. Redundancy has to cover the full path — device, link, subsystem — not just one layer.

Monitoring must catch the flood, not just the outage. A thousand connections every four to five seconds is a trivial signal for any modern monitoring tool to detect. The absence of real-time detection turned a local incident into a national stop. An alert on an abnormal connection rate could have cut the problem at the root in seconds.

A change without a rollback path is not a change; it is a bet. The failure happened during a telecom network upgrade. On a critical system, every change must ship with a tested rollback plan, not a hope that “it will be fine.”

The HPE controversy — “the product worked as intended” — is the most quoted part but the least important. What the incident proves is that legacy networks do not forgive implicit assumptions: a modern device that leans on a capability the old network never had is not an upgrade, it is a time bomb.

The verdict

If you operate a network whose downtime costs more than its modernization, map your single points of failure this week and require a tested rollback plan for every configuration change. The FAA incident shows that a four-hour outage can be paid for in days of chaos and political casualties — the cost of real redundancy is trivial by comparison.

If your monitoring does not alert on an abnormal connection rate within seconds, add that metric before your next maintenance window. The gap between “two sites are pinging too hard” and “the country is grounded” was measured in hours, not days. That is exactly the kind of signal your stack should already be catching.

References

The cyber brief, every Tuesday

The flaws that matter and the patches to apply, in a ten-minute read.

No spam. One-click unsubscribe.
read next

On the same topic

← Back to the feed

Type at least two characters.

↑ ↓ navigate ↵ open esc dismiss