The story starts in a somewhat amusing fashion, an argument over metric conversion. Normally a 767 is fueled almost completely automatically using a device known as the Fuel Quantity Information System Processor, which runs all of the internal pumps and reports to the pilots on the status of the fuel load. However Flight 143's FQIS was not working properly, a problem later traced to a bad solder joint in the capacitance gauges in the fuel tanks. The fuel load was instead measured with a dripstick, a sort of dipstick for planes, giving the total volume of fuel in the tanks.
The problem occurred when it became time to calculate how much fuel was needed for the trip from Montreal to Edmonton. The calculations were based not on volume, but weight, so the measurements had to be converted. The 767 measured fuel in kilograms, whereas all of the other manuals and planes in the Air Canada fleet used pounds. Looking in their notes for the conversion they used the factor of 1.77 pounds/liter, but a plane measured in kg should have used .8 kg/liter instead. After using the 1.77 figure they punched in 20,400 to the computer, indicating kg, and the computer said there was enough fuel. In fact they had only 9144 kg onboard, not nearly enough for their flight to Edmonton.
Both the pilots and the fueling crew had misgivings about the math, but after trying three times and coming up with the same number, the pilot, Pearson, finally stated that's it, we're going. Flight 143 then flew the short distance from Montreal to Ottawa, on landing they had the tanks re-dripped just to make sure, and once again the numbers added up – incorrectly. They then took off on the second leg of the trip, from Ottawa to Edmonton.
At 41,000 feet over Red Lake Ontario, the cockpit warning system chimed four times, and indicated a fuel pressure problem on the left side. The pilots thought a fuel pump had failed, and turned it off. The computer said there was still lots of fuel, but of course it didn't really know. A few moments later a second alarm sounded, and the pilots decided to divert to Winnipeg. Within seconds the left engine failed and they prepared for a one-engine landing.
While they attempted to re-start the engine and communicate with controllers in Winnipeg for an emergency landing, the warning system sounded again, this time with a long "bong" that no one had ever heard. The sound was in fact the "all engines out" sound, something none of the training simulated. Seconds later the right side engine stopped, and the 767 lost all power and went quiet, allowing the cockpit voice recorder to easily pick out "oh fuck".
The 767 is based on a "glass cockpit", in which mechanical instruments, which are hard to read and fail regularly, are replaced with several monitors in the cockpit. This works very well with the exception that the jet engines also deliver electical power to the aircraft, so most of the instrumentation suddenly went dead. One of the lost instruments was the vertical-rate indicator, which would let them know how fast they were sinking and how far they could go.
The engines also supply power to the hydraulic systems, without which a plane the size of the 767 could not be controlled. However Boeing actually planned for this and included a device known as a ram air turbine that automatically popped open on the side of the plane, spinning a propeller to provide enough power to the hydraulics to make it controllable.
The pilots immediately opened the emergency guide to the section on flying the aircraft with both engines out, only to find there was no such section. Pearson glided the plane at 220 knots, his best guess as to the optimum airspeed. Copilot Quintal began making calculations to see if they'd reach Winnipeg, using the altitude from one of the mechanical backup instruments and distance travelled supplied by the air traffic controllers in Winnipeg by measuring the distance the spot moved on their radar screens. The plane had lost 5,000 ft in ten nautical miles (18.5 km), giving a glide ratio of approximately 11:1 (which is actually quite good). The controllers and Quintal both calculated that Flight 143 would not make Winnipeg.
At this point Quintal selected his former RCAF base at Gimli as the landing spot. Unbeknownst to Quintal, since being in the service Gimli had "gone public" and was now being used for drag racing, and to make matters worse, on this particular day the area was covered with cars and campers for "Family Day".
As they approached Quintal did a power-off "gravity drop" of the main gear, but the nose wheel, which opens to the front, wouldn't lock. The ever reducing speed of the plane also reduced the effectiveness of the ram air turbine, and the plane became increasingly hard to control. As they grew nearer it became apparent they were too high, and Pearson executed a manuver known as a "sideslip" to increase their drag and reduce their altitude. As soon as the wheels touched the runway they stood on the brakes, blowing several tires. The plane came to rest, nose down due to the unlocked nose gear, only a few hundred feet from Family Day at the end of the runway.
No one was hurt in the landing, although there were minor injuries when exiting via the rear slide, which, at the tail, was almost vertical. A minor fire in the nose area was soon put out by the racers who rushed over with fire extinguishers. The injuries were soon addressed by a doctor who just happened to be about to take off on Gimli's other runway, still being used for a flying club and skydivers.
Mechanics soon repaired the minor damage and flew the plane out two days later. Within weeks it was fully repaired and back in service. As an amusing side note, the mechanics sent from Winnipeg Airport ran out of fuel on their way to Gimli, and found themselves stranded in the backwoods of Manitoba. Another van was sent to pick them up.