How I Almost Burned Down Cowboys Stadium — and Proved the Demo Worked

In October 2008, Hewlett-Packard acquired LeftHand Networks, a pioneer in iSCSI SAN and clustered storage. Less than a year later, I was asked to help turn that acquisition into a live, high-stakes demo.

We built a Site A / Site B clustered storage environment in HP’s Houston lab, crated it up, and shipped it to VMworld 2009 in San Francisco (August 31 – September 3, 2009). During the Day 1 keynote, a LeftHand engineer delivered the demo: a streaming video that continued playing while workloads failed over between sites. It worked flawlessly in front of thousands.

Mark Potter, HP’s SVP & GM of Industry Standard Servers, liked the demo enough that he wanted it added to an already-scheduled internal HP sales event at the newly opened Cowboys Stadium in Arlington, Texas.

There was a catch. The VMworld hardware was immediately shipped to a demo center after the show. So I had to build a second Site A / Site B environment back in the Houston lab. It was short notice. I rented an SUV, loaded the equipment myself, and drove it from Houston to Dallas.

Setup Day — October 13, 2009

We set up the environment just outside a medium-sized presentation room inside the stadium. No racks — just portable gear staged for the demo. The environment ran fine all day.

The next morning, Mark Potter and Dallas Cowboys CTO Bill Haggard were scheduled to co-present the demo to about 30–40 HP salespeople. Before that, Mark did a rehearsal. I walked him through the steps, explained what the failover would look like, and trained him on how to give the demo cleanly and confidently.

The Part Where It Almost Went Very Wrong

That evening, Mark and others went to dinner. The LeftHand engineer and I stayed behind to run a few more failover tests.

I had requested two 30-amp power drops. What we actually got were several 15-amp extension cords plugged into 20-amp outlets. It had been working all day — but after multiple failover cycles, with servers rebooting and disks spinning up at the same time, the startup surge caught up with us.

I heard a crackling sound.

Then I saw flames coming from an extension-cord reel hidden under a draped table.

I yanked the plug.

The fire went out immediately. Smoke lingered. The only damage was the cord reel itself.

Right then, Bill Haggard walked in carrying a fire extinguisher, ready to act. Stadium staff brought additional extension cables, and we redistributed the load safely.

Dinner at Pappadeaux

Later that night, we joined the group at Pappadeaux Seafood Kitchen. They were just ordering dessert when we arrived. I had coffee.

We told the story — the crackling, the flame, the smoke, and Bill showing up with a fire extinguisher. What could have been a disaster turned into a mix of relief, laughter, and realization.

Because here’s the irony:

The demo worked.

During the overload, the Site A / Site B environment failed over exactly as designed. The video stream never stopped. The system did precisely what it was supposed to do under real-world stress.

Demo Day — October 14, 2009

The next morning, Mark Potter and Bill Haggard co-presented the demo to the sales team without incident. Calm. Clean. Convincing.

Afterward, Mark bought me a Dallas Cowboys hoodie from the stadium gift shop.

Later, my management chain told me I’d “pulled a rabbit out of a hat.”

Takeaways

• Never assume a venue has the power you requested.
• Extension cords are not infrastructure.
• Real failures are the best demos.
• If your system survives smoke and fire, your sales pitch just got easier.

I went home with a melted cord reel, a hoodie, and one of the best stories of my career.