[drob518]

Frequently, when you see these massive failures, the root cause is an alignment of small weaknesses that all come together on a specific day. See, for instance, the space shuttle O-ring incident, Three-Mile Island, Fukushima, etc. These are complex systems with lots of moving parts and lots of (sometimes independent) people managing them. In a sense, the complexity it the common root cause.

[burningChrome]

This is the same thing that happened with the 35W bridge collapse in Minneapolis. The gusset plates after the disaster were examined and found to be only 1/2" thick when the original design called for them to actually be 1" thick. The bridge was a ticking time bomb since the day it was built in 1967.

As the years went on, the bridge's weight capacity was slowly eroded by subsequent construction projects like adding thicker concrete deck overlays, concrete median barriers and additional guard rail and other safety improvements. This was the second issue, lining up with the first issue of thinner gusset plates.

The third issue that lined up with the other two was the day of the bridges failure. There were approximately 300 tons of construction materials and heavy machinery parked on two adjacent closed lanes. Add in the additional weight of cars during rush hour when traffic moved the slowest and the bridge was a part of a bottleneck coming out of the city. That was the last straw and when the gusset plates finally gave way, creating a near instantaneous collapse.

[linuxguy2]

It's like the Swiss Cheese model where every system has "holes" or vulnerabilities, several layers, and a major incident only occurs when a hole aligns through all the layers.

https://en.wikipedia.org/wiki/Swiss_cheese_model

[Ringz]

I use this model all the time. It's very helpful for explaining the multifactorial genesis of catastrophes to ordinary people.

[anonymars]

Also perhaps worth a read:

https://devblogs.microsoft.com/oldnewthing/20080416-00/?p=22...

"You’ve all experienced the Fundamental Failure-Mode Theorem: You’re investigating a problem and along the way you find some function that never worked. A cache has a bug that results in cache misses when there should be hits. A request for an object that should be there somehow always fails. And yet the system still worked in spite of these errors. Eventually you trace the problem to a recent change that exposed all of the other bugs. Those bugs were always there, but the system kept on working because there was enough redundancy that one component was able to compensate for the failure of another component. Sometimes this chain of errors and compensation continues for several cycles, until finally the last protective layer fails and the underlying errors are exposed."

[jacquesm]

I've had that multiple times. As well as the closely related 'that can't possibly have ever worked' and sure enough it never did. Forensics in old codebases with modern tools is always fun.

[magicalhippo]

> As well as the closely related 'that can't possibly have ever worked' and sure enough it never did.

I had one of those, customer is adamant latest version broke some function, I check related code and it hasn't been touched for 7 years, and as written couldn't possibly work. I try and indeed, doesn't work. Yet customer persisted.

Long story short, an unrelated bug in a different module caused the old, non-functioning code to do something entirely different if you had that other module open as well, and the user had disciverdd this and started relying on this emergent functionality.

I had made a change to that other module in the new release and in the process returned the first module to its non-functioning state.

The reason they interacted was of course some global variables. Good times...

[anonymars]

By the way, a corollary I encountered, I think with one of the recent AWS meltdowns, is that a paradoxical consequence of designing for "reliability" is that it guarantees that when something does happen, it's going to be bad, because the reliability engineering has done a good job of masking all the smaller faults.

Which means 1. anything that gets through, almost by definition, is going to be bad enough to escape the safeguards, and 2. when things do get bad enough to escape the safeguards, it will likely expose the avalanche of things that were already in a failure state but were being mitigated

The takeaway, which I'm not really sure how to practically make use of, was that if a system isn't observably failing occasionally in small ways, one day it's going to instead fail in a big way

I don't think that's necessarily something rigorously proven but I do think of it sometimes in the face of some mess

[jacquesm]

Global variables... the original sin if you ask me. Forget that apple.

[roenxi]

> See, for instance, the space shuttle O-ring incident

That wasn't really a result of an alignment of small weaknesses though. One of the reasons that whole thing was of particular interest was Feynman's withering appendix to the report where he pointed out that the management team wasn't listening to the engineering assessments of the safety of the venture and were making judgement calls like claiming that a component that had failed in testing was safe.

If a situation is being managed by people who can't assess technical risk, the failures aren't the result of many small weaknesses aligning. It wasn't an alignment of small failures as much as that a component that was well understood to be a likely point of failure had probably failed. Driven by poor management.

> Fukushima

This one too. Wasn't the reactor hit by a wave that was outside design tolerance? My memory was that they were hit by an earthquake that was outside design spec, then a tsunami that was outside design spec. That isn't a number of small weaknesses coming together. If you hit something with forces outside design spec then it might break. Not much of a mystery there. From a similar perspective if you design something for a 1:500 year storm then 1/500th of them might easily fail every year to storms. No small alignment of circumstances needed.

[cpgxiii]

In reality the "swiss cheese" holes for major accidents often turn out to be large holes that were thought to be small at the time.

> [Fukushima] No small alignment of circumstances needed.

The tsunami is what initiated the accident, but the consequences were so severe precisely because of decades of bad decisions, many of which would have been assumed to be minor decisions at the time they were made. E.g.

- The design earthquake and tsunami threat

- Not reassessing the design earthquake and tsunami threat in light of experience

- At a national level, not identifying that different plants were being built to different design tsunami threats (an otherwise similar plant avoid damage by virtue of its taller seawall)

- At a national level, having too much trust in nuclear power industry companies, and not reconsidering that confidence after a number of serious incidents

- Design locations of emergency equipment in the plant complex (e.g. putting pumps and generators needed for emergency cooling in areas that would flood)

- Not reassessing the locations and types of emergency equipment in the plant (i.e. identifying that a flood of the complex could disable emergency cooling systems)

- At a company and national level, not having emergency plans to provide backup power and cooling flow to a damaged power plant

- At a company and national level, not having a clear hierarchy of control and objective during serious emergencies (e.g. not making/being able to make the prompt decision to start emergency cooling with sea water)

Many or all of these failures were necessary in combination for the accident to become the disaster it was. Remove just a few of those failures and the accident is prevented entirely (e.g. a taller seawall is built or retrofitted) or greatly reduced (e.g. the plant is still rendered inoperable but without multiple meltdowns and with minimal radioactive release).

[roenxi]

To be blunt; that isn't an appropriate application of the swiss cheese model to Fukushima. It isn't a swiss cheese failure if it was hit by an out-of-design-spec event. Risk models won't help there. Every engineered system has design tolerances. And that system will eventually be hit by a situation outside the tolerances and fail. Risk models aren't to overcome that reality - they are one of a number of tools for making sure that systems can tolerate situations that they were designed for.

If Japan gets traumatised and changes their risk tolerance in response then sure, that is something they could do. But from an engineering perspective it isn't a series of small circumstances leading to a failure - it is a single event that the design was never built to tolerate leading to a failure. There is a lot to learn, but there isn't a chain of small defence failures leading to an unexpected outcome. By choice, they never built defences against this so the defences aren't there to fail.

> Many or all of these failures were necessary in combination for the accident to become the disaster it was.

Most of those items on your list aren't even mistakes. Japan could reasonably re-do everything they did all over again in the same way that they could simply rebuild all the other buildings that were destroyed in much the same way they did the first time. They probably won't, but it is a perfectly reasonable option.

Again I'm going from memory with the numbers but doubling the cost of a rare disaster in a way that injures ... pretty much nobody ... is a great trade for cheap secure energy. It isn't a clear case that anything needs to change or even went wrong in the design process. Massive earthquakes and tsunamis aren't easy to deal with.

[cpgxiii]

> It isn't a swiss cheese failure if it was hit by an out-of-design-spec event

First of all, the design basis accident is a design choice by the developers of the plant and regulators. The decision process that produced that DBA was clearly faulty - the economic and social costs of the disaster so clearly have exceeded those of a building to a more serious DBA.

> Again I'm going from memory with the numbers but doubling the cost of a rare disaster in a way that injures ... pretty much nobody ... is a great trade for cheap secure energy. It isn't a clear case that anything needs to change or even went wrong in the design process. Massive earthquakes and tsunamis aren't easy to deal with.

This is absolute nonsense. For the cost of maybe maybe tens of millions at most in additional concrete to build the seawall a few meters higher, the entire disaster would have been avoided entirely (i.e. plant restored to operation). With backup cooling that could have survived the tsunami (a lower expense than building a higher seawall), all that would have happened at Fukushima Daiichi is what happened at its neighbor Fukushima Daini (plant rendered inoperable, no meltdown, no significant radioactive release). Instead, we are talking about a disaster that will cost a (current) estimated $180 billion USD to clean up (and there is no way this estimate is realistic, when the methods required to perform the cleanup barely exist yet).

[roenxi]

> The decision process that produced that DBA was clearly faulty - the economic and social costs of the disaster so clearly have exceeded those of a building to a more serious DBA.

That isn't clear at all. We're effectively sampling from the entire globe and we've had 2-3x bad nuclear disasters since the 70s. Our safety standards appear to be overcautious given the relatively small amount of damage done vs ... pretty much every alternative. The designs seem to be fine. I'm still waiting to see the justification for the evacuations from Fukushima; they seemed excessive. People died.

> For the cost of maybe maybe tens of millions at most...

You haven't thought for long enough before you typed that. For this particular disaster, sure. But hardening against all the possible disasters is what needs to happen when you become less risk tolerant. It is the millions of dollars to prevent against this disaster multiplied by the number of potential disasters that you have to consider. Safety is expensive.

The numbers aren't small, safety of that magnitude might not even be economically feasible. To say nothing of whether it is actually sensible. And once you get into one in 500 or thousand year events, some really catastrophic stuff starts happening that just can't be reasonably defended against. San Francisco and its fault springs to mind, I forget what sort of even that is but it is probably once a millennium or more often.

[drtgh]

Fukushima was designed to be constructed on a hill 30-35 meters above the ocean, but someones decided would be cheaper to construct it at sea level in order to reduce costs in water pumping, others decided to approve this, and much latter, one decade before the disaster when was requested to reinforce the security measures within all the reactors at Japan, those in charge of Fukushima decided to ignore it, again, pushing for extensions year after year until it all blew up. Decades of bad decisions with a strong smell to corruption.

https://warp.da.ndl.go.jp/info:ndljp/pid/3856371/naiic.go.jp...

https://warp.da.ndl.go.jp/info:ndljp/pid/3856371/naiic.go.jp...

https://web.archive.org/web/20210314022059/https://carnegiee...

[dreamcompiler]

There was a strong corporate cultural component to Fukushima as well. Tepco had spent decades telling the Japanese public that nuclear power was completely safe. A tall order in Japan obviously, but by and large it worked.

During the operation of Fukushima Daiichi, various studies had been done that recommended upgraded safety features like enlarging the seawall, moving the emergency generators above ground so they couldn't be flooded, etc.

In every case, management rejected the recommendations because:

1. They would cost money.

2. Upgrading safety would be tantamount to admitting the reactors were less than safe before, and we can't have that.

3. See 1.

[drob518]

I’m not sure why you think those are not a confluence of smaller events or that something outside the design spec isn’t one of those factors. By “small,” I don’t mean trivial. I mean an event that by itself wouldn’t necessarily result in disaster. Perhaps I should have said “smaller” rather than “small.” With the O-rings, the cold and the pressure to launch on that particular day all created the confluence. With Fukushima, the earthquake knocked out main power for primary cooling. That would have been manageable except then the backup generators got destroyed by the tsunami. It was not a case of just a big earthquake, whether outside or inside the design spec, making the reactor building fall down and then radiation being released.

[roenxi]

If Fukushima get hit by a disaster that is outside the design spec then the engineering root cause of the failure is established. There isn't some detailed process needed to figure out how a design should tolerate out-of-design events. And there isn't a confluence of smaller events, it is a very cut and dry situation (well, unstable and wet situation I suppose). There was one event that caused the failure. An event on a biblical scale that was hard to miss.

If you want Fukushima to tolerate things it wasn't designed to tolerate or fail in ways it wasn't designed to fail in then the swiss cheese model isn't going to be much help. You're going to need to convince politicians and corporate entities that their risk tolerance is too high. Which in a rational world would be a debate because it isn't obvious that the risk tolerances were inappropriate.

[namibj]

The design spect tsunami resistance is for getting away with just a couple days downtime plus what the grid concerns.

A much higher much rare case is what happened and which they didn't have a plan ready on hand.

Even if you treat the box as the special being they wre...

[amelius]

It usually starts with a broken coffee machine.

[drob518]

When that happens, get ready.