I haven’t read the article because documentation is overhead but I’m guessing the real reason is because the guy who kept saying they needed to add more storage was repeatedly told to calm down and stop overreacting.
I used to do some freelance work years ago and I had a number of customers who operated assembly lines. I specialized in emergency database restoration, and the assembly line folks were my favorite customers. They know how much it costs them for every hour of downtime, and never balked at my rates and minimums.
The majority of the time the outages were due to failure to follow basic maintenance, and log files eating up storage space was a common culprit.
So yes, I wouldn’t be surprised at all if the problem was something called out by the local IT, but were overruled for one reason or another.
and log files eating up storage space was a common culprit.
Another classic symptom of poorly maintained software.
Constant announcements of trivial nonsense, like [INFO]: Sum(1, 1) - got result 2! filling up disks.
I don’t know if the systems you’re talking about are like this, but it wouldn’t surprise me!
You gotta forward that to Spunk so your logs ain’t filling up the server generating them. Plus you can set up automated alerts for when the result stops being 2.
This message brought to you by Big Splunk.
And yet that’s probably there because sometime, somewhere, it returned 1.9 or 2.00001 or some such nonsense.
Yeah a few levels.
Level 1: complex stand alone devices, mostly firmware.
Level 1a. Stuff slightly more complicated than a list of settings, usually for something like a VFD or a stepper motor controllers. Not as common.
Level 2 PLCs, HMIs, and the black magic robotic stuff. Stand alone equipment. Like imagine a machine that can take something, heat it up, and give it to the next machine.
Level 3: DCS and SCADA. Data control center and whatever SCADA stands for, I always forget. This is typically for integrating or at least data collection of multiple stand alone equipment for level 2.
Level 4: the integration layer between Level 3 and whatever means the company has for entering in sales.
Like everything in software this is all general. Some places will mix layers, subtract layers, add them. I would complain about the inconsistent nature of it all but without it I would be unemployed.
I’m this person in my organization. I sent an email up the chain warning folks we were going to eventually run out of space about 2 years ago.
Guess what just recently happened?
ShockedPikachuFace.gif
Sometimes that person is very silly though. We had a vendor call us saying we needed to clear our logs ASAP!!! due to their size. The log file was no joke, 20 years old. At the current rate, our disk would be full in another 20 years. We cleared it but like, calm down dude.
Can’t you just add a few external USB drives? (heard this more than once at an NGO think tank.)
I mean I’ve worked at a hosting company that had a bunch of static sites running off an SSD connected by usb to the server so this did happen back in the day. I try not to think about those days.
“What’s that? Your accounting front end that’s built in obsolete front page code on an Access database isn’t working again? It’s probably a file lock, I’ll restart IIS.”
Ballast!
Just plonk a large file in the storage, make it relative to however much is normally used in the span of a work week or so. Then when shit hits the fan, delete the ballast and you’ll suddenly have bought a week to “find” and implement a solution. You’ll be hailed as a hero, rather than be the annoying doomer that just bothers people about technical stuff that’s irrelevant to the here and now.
Or you could be fired because technically you’re the one that caused the outage.
Sysadmin pro tip: Keep a 1-10GB file of random data named DELETEME on your data drives. Then if this happens you can get some quick breathing room to fix things.
Also, set up alerts for disk space.
Why not both? Alerting to find issues quickly, a bit of extra storage so you have more options available in case of an outage, and maybe some redundancy for good measure.
A system this critical is on a SAN, if you’re properly alerting adding a bit more storage space is a 5 minute task.
It should also have a DR solution, yes.
Yes, alert me when disk space is about to run out so I can ask for a massive raise and quit my job when they dont give it to me.
Then when TSHTF they pay me to come back.
A lot of companies have minimal alerting or no alerting at all. It’s kind of wild. I literally have better alerting in my home setup than many companies do lol
It’s certainly cheaper to not have any but it will limit growth substantially
There’s cases where disk fills up quicker than one can reasonably react, even if alerts are in place. And sometimes culprit is something you can’t just go and kill.
Had an issue like that a few years back. A stand alone device that was filling up quickly. The poorly designed device could only be flushed via USB sticks. I told them that they had to do it weekly. Guess what they didn’t do. Looking back I should have made it alarm and flash once a week on a timer.
The real pro tip is to segregate the core system and anything on your system that eats up disk space into separate partitions, along with alerting, log rotation, etc. And also to not have a single point of failure in general. Hard to say exact what went wrong w/ Toyota but they probably could have planned better for it in a general way.
Even better, cron job every 5 mins and if total remaining space falls to 5% auto delete the file and send a message to sys admin
10GB is nothing in an enterprise datastore housing PBs of data. 10GB is nothing for my 80TB homelab!
It not going to bring the service online, but it will prevent a full disk from letting you do other things. In some cases SSH won’t work with a full disk.
It’s all fun and games until tab autocomplete stops working because of disk space
This happens. Recently we had a problem in production where our database grew by a factor of 10 in just a few minutes due to a replication glitch. Of course it took down the whole application as we ran out of space.
Some things just happen and all head room and monitoring cannot save you if things go seriously wrong. You cannot prepare for everything in life and IT I guess. It is part of the job.
Bad things can happen but that’s why you build disaster recovery into the infrastructure. Especially with a compqny as big as Toyota, you can’t have a single point of failure like this. They produce over 13,000 cars per day. This failure cost them close to 300,000,000 dollars just in cars.
In my experience, the C-Suite dicks will put the hammer down on someone and maybe fire a couple of folks. They’ll demand a summary of what happened and what will be done to stop it from happening again. IT will provide legit options to resolve this long term, but because that comes with a price tag they’ll be told to fix it with “process changes” and the cycle continues.
If they give IT money that’s less for themselves at EOY for bonuses so it’s a big concern /s
Yea, fair point regarding the single point of failure. I guess it was one of those scenarios that should just never happen.
I am sure it won’t happen again though.
As I said it can just happen even though you have redundant systems and everything. Sometimes you don’t think about that one unlikely scenario and boom.
I blame lean philosophy. Keeping spare parts and redundancy is expensive so definitely don’t do it…which is just rolling the dice until it comes up snake eyes and your plant shuts down.
It’s the “save 5% yearly and stop trying to avoid a daily 5% chance of disaster”
Over prepared is silly, but so is under prepared.
They were under prepared.
Lean philosophy is supposed to account for those dice-rolling moments. It’s not just “keep nothing in inventory”, there is supposed to be risk assessment involved.
The problem is that leadership doesn’t interpret it that way and just sees “minimizing inventory increases profit!”
The problem is that leadership doesn’t interpret it that way and just sees “minimizing inventory increases profit!”
Yep. Managers prioritize short-term gains (often personal gains, too) over the overall health of a business.
There’s also industries where the “lean” strategy is inappropriate because the given industry is one that booms in times of crisis when logistics to get “just in time” supplies go kaput due to the same catastrophe that’s causing the industry to boom. Hospitals and clinics can end up in trouble like this.
But there’s other industries too–I haven’t looked for it, but I’m sure there’s a plethora of analysis already on what Covid did to companies and their supply chains.
In my own impression from the side of software engineering (i.e. the whole discipline rather than just “coding”) this kind of thing is pretty common:
- Start with ad-hoc software development with lots of confusion, redundancy, inneficient “we’ll figure it out as when we get there” and so on.
- To improve on this somebody really thinks things through and eventually a software development process emerges, something like Agile.
- There are lots of good reasons for every part of this processes but naturally sometimes the conditions are not met and certain parts are not suitable for use: the whole process is not and can never be a one size fits all silver bullet because it’s way to complex and vast a discipline for that (if it wasn’t you wouldn’t need a process to do it with even the minimum of efficency).
- However most people using it aren’t the “grand thinkers” of software engineering - software architect level types with tons of experience and who thus have seen quite a lot and know why certain elements of a process are as they are, and hence when to use them and when not to use them - and instead they’re run-of-the-mill, far more junior software designers and developers, as well as people from the management side of things trying to organise a tech-heavy process.
So you end up with what is an excellent process when used by people who know that each part tries to achieve, what’s the point of that and when is it actually applicable, being used by people who have no such experience and understanding of software development processes and just use it as one big recipe, blindly following it with no real understanding and hence often using it incorrectly.
For example, you see tons of situations where the short development cycles of Agile (aka sprints) and use cases are used without the crucial element which is actually envolving the end-users or stakeholders in the definition of the use cases, evaluation of results and even prioritization of what to do in the next sprint, so one of the crucial objectives of use cases - the discovery of the requirement details by interactive cycles with end-users where they quickly see some results and you use their feedback to fine-tune what gets done to match what they actually need (rather than the vague very high level idea they themselves have at the start of the project) is not at all achieve and instead they’re little more than small project milestones that in the old days would just be entries in Microsoft Manager or some tool like that.
This is IMHO the “problem” with any advanced systematic process in a complex domain: it’s excellent in the hands of those who have enough experience and understanding of concerns at all levels to use it but they’re generally either used by people without that experience (often because managers don’t even recognize the value of that experience until things unexpectedly blow up) or by actual managers whose experience might be vast but is actuallly in a parallel track that’s not really about dealing with the kinds of technical concerns that the process is designed to account for.
it’s excellent in the hands of those who have enough experience and understanding of concerns at all levels to use it but they’re generally either used by people without that experience
You’re just saying that skilled people can do stuff better than unskilled people. This is hardly a software engineering issue. It is common in all aspects of life
The difference with software engineering is that the field is still relatively young enough to not have figured out a working model or sets of working models (unlike farming, for example, or finance).This field is realistically 30 years old during which it continuously evolves and redefines itself so it’s still not going to produce good working models.
Agile, since you picked on it, is very difficult to implement because it specifically relies on engineering figuring out how to work and how to deliver. It’s really not a model at all. It’s just a set of guidelines meant to create the environment in which the figuring out happens. It’s no wonder that it only works when people have the ability to figure it out.
I work in a manufacturing company that was owned by the founder for 50 years until about 4 years ago when he retired. He disagreed with a lot of the ideas behind lean manufacturing so we had like 5 years worth of inventory sitting in our warehouse.
When the new management came in, there was a lot of squawking about inefficiency, how wasteful it was to keep so much raw material on the shelf, and how we absolutely needed to sell it off or get rid of it.
Then a funny little thing happened in 2020.
Suddenly, we were the only company in our industry still churning out product. Other companies were calling us, desperate to buy our products or even just our raw material. We saw MASSIVE growth the next two years and came out of the pandemic better than ever. And it was mostly thanks to the old owners view that “Just In Time” manufacturing was BS.
Cool story, but a once every 150 years pandemic is hardly a good reason to keep wasting money on storing stuff. A fire or a flood was much more likely to wipe it all out in 50 years.
Even in your anecdote the owner never actually benefited from the extra costs.
Depending on what you’re producing costs to maintain extra inventory of raw materials can be massive and for the company the size of Toyota, multiply that by million.
Oh man. I work for an OEM and have almost the same story except the CEO didn’t retire so our shelves were never bare. A large part of our sales last year was because our competition couldn’t source parts and we could. Also since they have a skeleton engineering crew they couldn’t figure out how to improvise (at least this is what our salespeople are claiming).
2022 I spent 3 hours a day, every single working day, just making ECNs for parts out of stock.
Lean JIT is such fucking crap. When a customer line goes down they are bleeding who knows how many millions a day. They want a solution right freaken now. So they call the OEM and we tell them that part that died we have right here on the shelf and we can overnight it.
But you know how you really know that JIT Evengelical types are full of it? Talk to any of them now and they won’t say something like “look, normally it works but no one could have predicted this” they will double down on it and say it works but was implement incorrectly. It isnt “true” JIT.
Many years ago I read a sentence that I use every single week of my life “all ideology can do is reassert itself endlessly”. When I hear crap like that, how X works if and only if it is true-X I know I am dealing with ideology. When I hear that X works under very specific situations and only gives specific results I know I am dealing with facts. Yes if the customer is never in a rush, there isn’t a world shaking event, and you are running out of capacity for storage JIT might be an answer.
There’s some irony to every tech company modeling their pipeline off Toyota’s Kanban system…
Only for Toyota to completely fuck up their tech by running out of disk space for their system to exist on. Looks like someone should have put “Buy more hard drives” to the board.
not to mention the lean process effed them during fukashima and covid, with a breakdown in logistics and a shortage of chips, meant that their entire mode of operating shut down, as they had no capacity to deal with any outages in any of their systems. Maybe that has happened again, just in server land.
Toyota was the carmaker best positioned for the COVID chip shortage because they recognized it as a bottleneck. They were pumping out cars a few months longer than the others (even if they eventually hit the same wall everyone else did).
It wasn’t just Fukushima. There was a massive flood in Thailand at the same time that shut down a load of suppliers. It was a really bad bit of luck but they did learn from that.
