A false smoke alarm almost drowned Jane Street's only NYC datacenter
michael.bacarella.com
michael.bacarella.com
“The superhuman folks at the data center are hauling 55 gallon drums of diesel fuel up 17 flights of stairs. Assuming they're able to sustain a rate of transfer that exceeds the rate of consumption, the data center should run until electricity is restored. The latest estimate from Con Edison for lower Manhattan is 4 days. (The New York Stock Exchange will be open tomorrow, also running on generator power.)”
https://news.ycombinator.com/item?id=4720388 (https://web.archive.org/web/20121101194314/http://status.fog...)
https://arstechnica.com/information-technology/2012/11/how-o...
Though the "55 gallon" drum lift is greatly exaggerated, that was a crew of guys hired off the street who were walking around doing odd jobs in the disaster zone. Kinda wild. For us, we just carried the 5 gallon buckets.
You get out of it what you put into it. ;)
With fuel storage on the same floor for a reasonable runtime, larger fuel storage in the basement and a pump (or two) to move fuel, you're well prepared for lots of things. As long as the fuel throughput on the stairs is better than the fuel use rate, then you're ok for as long as you've got people to move it. Which is ok for exceptional incidents.
It makes me wonder if anyone is paranoid enough to do the same for highrise fuel pumps: sure, the basement may flood, but when the pump in there is designed to flood it's a lot less of a problem! Place the actual power source several floors up, where flooding is impossible outside a The Day After Tomorrow-esque civilization-ending catastrophe and you should be fine.
shocking how much engineering you can get done when the normal process becomes impossible and waiting isn't an option.
Apparently it was a 30-person 5 gallon bucket brigade of customers, contractors, and volunteers all lining up to help fill tanks because of interconnection issues between the supplier they got and where they needed to fuel the tanks up with.
"... neither we nor the fuel company had the proper coupling to tie into that existing infrastructure."
Sometimes "DR" in your plans stands for "didn't realize".
As a bonus though? They were able to in fact get ahead of the fuel burn, and the crew managed an eight hours' head back up on the tanks.
I had made sure to observe the filter replacement process in case we needed to do it again ourselves. We did, and the next time someone recorded me doing it so everyone could learn and we handled most I’d the rest of the swaps from then on.
There were a lot of memorable experiences that week or so.
That's also why it's not a great idea to use diesel for your backup home generator. It will go bad sooner than you expect, and then you're left with drums of gunk that you can't dispose and can't use.
And gasoline won't help you either. It goes bad even faster.
Propane tanks are the best option they can sit in storage indefinitely.
So you get clumps of gunk in your fuel that rapidly clogs the filters.
Definitely a memorable week. I wonder just how many folks in Fidi were learning the finer points of emergency backup generators.
> “They had their brigade of people carrying five-gallon buckets up 18 flights of stairs or whatever it was to get to their auxiliary generator. Quite a task by those guys,” Orchard said.
5 gallons of fuel at a time (hopefully not in a literal bucket) is a lot easier than managing a big drum on stairs. Unless you have a powered stair dolly, but powering your stair dolly up 17 flights of stairs seems like a challenge too. Team lifting a drum up one flight of stairs feels managable, doing it up 17 flights multiple times is a hard no.
Bucket lids are a PITA, but so is spilling diesel on the stairs.
Almost anything that’s not a single use plastic and has a shape to hold will be made out of hdpe or polypropylene, just like a gas can, or a bucket, or those black and yellow boxes that are everywhere, or believe it or not, gas tanks on most newer ice vehicles.
So, yeah, don’t use a trash bag, but that shouldn’t need to be said.
The main reason you shouldn’t use things for normal storage or transportation is because they don’t handle vapors, or pouring without spilling very well. But you can certainly carry gas up the stairs in a bucket, and you can store gas in a bucket as well, but temperature changes can be problematic over time.
I have a stack of clean buckets that I routinely use when I need to remove the fuel from a vehicle to work in the fuel system. Gas or diesel cans are simply to expensive to keep a dozen on hand to use for an afternoon or a week while replacing a fuel pump or the fuel lines.
And in the case of a 17 floor bucket brigade, also likely too expensive and probably hard to find in the quantities they’d need to make that work, if you assume one bucket each way for 17 floors.
You’re probably not going to walk into an Ace and find 34 5 gallon cans especially when anyone with power is running off a generator. But, you will find that many buckets because it’s also a paint store and they store easily and cost nothing.
So they were probably using buckets, even if it’s true that you shouldn’t put gasoline in a garbage bag.
Some of us did the full round trip a handful of times, but that was mostly early morning of when the power first when out and we just had to keep the generator going. Once we got into the late morning, enough of us were there that we could have a person at each floor and they'd receive a bucket and relay it forward.
I've done this job. I only had to do it a few times per day, but it sucked each and every time.
(When I showed up due to a report of an outage for our client's gear at the top of the building, the street was flooded. Water was flowing into a giant metal grating in the sidewalk like something from a cartoon. This grating lead to the basement.
And that basement was where the electrical switchgear was.
It was bad. It stayed bad for a long time.)
Context: in 2012, I literally quit Fog Creek, largely over the fact that we had no backup strategy and I was told that a backup strategy was not feasible. I was "convinced" (paid a lot) to stay on for another two years to see if Kiln would succeed. On my way out of that meeting, I commented to a coworker, "I mean, what's the chance a hurricane hits Manhattan in the next 24 months?"
100%. The chance was 100%. And the only reason people had to haul open containers of diesel up 17 flights of stairs was because no one wanted to admit that the chance wasn't 0%.
Anyway, not sure what was going on with Jane Street here, but holding up that Fog Creek/Trello/SquareSpace story as a positive always drives me somewhat nuts. It lionizes a failure.
So many times it's taken herculean effort to make things happen, usually because of poor planning and poor leadership.
Why these are viewed as "successes" or "heroic" baffles me as well.
It's not a mystery, it's exploitation.
With all the business efficiencies we've made with computers as devs and IT, we should all be working 3 days weeks, instead a few people are so rich they can buy elections.
It's never going to work like that, and you don't need any dastardly, sinister, or conspiratorial explanations for why. It's a natural side effect of the market being competitive.
When tools and technology make people more efficient, some people take that to mean they can work fewer hours and get the same job done. Which is true. But other people take that to mean they can work the same amount of hours, and get way more done. This is also true.
In a market in which two companies are competitors, and one takes the former approach and one takes the latter, which company do you think will still be around in ten or twenty years? Obviously the latter. And that pressures people in the former group to demand more output and stable working hours so they can remain competitive.
I think it also remains to be proven that the company that works their employees the hardest will necessarily outcompete the others.
It’s not necessary that the longer hours actually translate into productivity, just that the boss feels like it should.
Well, good organizing, too.
That is an argument of faith (that companies where employees work longer survive more than companies that optimise differently), and I am not convinced it is supported by evidence.
One could argue that all other things being equal, companies with employees doing more hours are more productive (even though I seriously doubt that, there are diminishing returns very quickly). But then I would argue that if a company drives their employees harder, there are leadership and management issues, so other things are not equal. And this is associated with younger workers (adaptable: good, inexperienced: less good), high turn-over (so, long-term information retention issues) and lower worker satisfaction (so the whole thing is self-defeating in the long term).
Besides, there are many, many factors involved that have nothing to with with working time. Again, leadership, strategy, operations, marketing, personal connections, dumb luck…
Personally I think there is an optimum somewhere, and at some point you get more out of your workers by making them more invested and motivated than making them work longer.
But everywhere I have seen people overworked has terrible morale and low productivity in spite of the hours.
Yet after 100 years of explosive technological growth we still have the labouring hours of the 1930s...
The result was that business owners invested in their workers and long term health of their business so it would continue to give them 10 mm per year instead of trying to cash everything out every 1/4.
Your source of confusion seems to be not recognizing that sometimes in life multiple things can be true at once. "Heroism" is generally about the actions of the individuals in response to a dangerous scenario, not how the scenario came about in the first place. "Success" is more overloaded, but can refer to both the overall system in place and the outcome of a given dangerous scenario, again separate from what caused that scenario to come about. This is near perfectly true in particular if, as is usually the case, the ones dealing with merde-hitting-spinning-blades on the ground are not the leadership. It may be idiotic, short sighted, lazy, inept leadership that lead to the problem (or it may simply be an unforeseen alignment of holes in the cheese), but if people on the ground go above and beyond to prevent the worst that's still heroic of them, personally. And if their efforts work that's a success. It'd be pretty strange to insist that the efforts of firefighters shouldn't be recognized just because somebody else did something stupid that resulted in a fire.
Yes, it's absolutely just as important what comes after. Ideally it's like the commercial aviation industry, where heroism on the parts of flight crews can be recognized and said heroism results in a top-to-bottom full systemic investigation and efforts to ensure that ideally nobody will ever be heroes in quite that way ever again. And have constant efforts to make sure heroism isn't ever required in the first place.
The opposite extreme of some organization using it as an excuse for inaction and cheapness (at least until the day arrives when heroic efforts are not enough and a true disaster really does occur) is abominable. Even worse, because in that sort of utterly failed leadership environment the people on the ground will probably get scapegoated too. And it's 100% worth acknowledging that there can be a "success trap" and Normalization Of Deviance situation here, where heroics keep averting the worst which lets bad leadership/management escape correction. But it's a trap precisely because each individual failure can still be bad, up to and including literally people dead, so the pressure to succeed is still there each time.
But if employees, be they civilian or military for that matter, perform far beyond the call of duty why would that ever not be heroic?
If I flip a fair coin and don't look at it as it sits in my hand, the probability of heads is 50%. Your estimate should be the same. Before or after the flip doesn't matter because our state of knowledge is the same.
If I look (and you don't), the probability that it is heads from my state of knowledge is 100% or 0%. Your estimate would be unchanged because you didn't get any more information.
If somebody looks at the coin and tells you what they say, you estimate should be somewhat less than 100% or more than 0% depending on how much you trust them.
All of this is the long way around the barn to say that after the hurricane, the probability that Sandy hit Manhattan is definitely 100%. That's past tense probability, however. It is still also true that the best estimate a year before it hit would be much, much lower. Using words like "probably that a hurricane would hit Manhattan" makes interpretation much harder, of course, because of the uncertain time frame.
My company during Sandy was operated in both NY and SF, and our experience was a bunch of us on the west coast filling in for those who were not only unable to get to the office, but without power themselves in NYC. It was all unfamiliar stuff for us (different business functions were done there, but the NYC office’s work was time critical) but did build camaraderie.
"Everyone coming together to solve a problem by any means necessary" still loses something if the problem is sufficiently imaginary.
Like, carrying fuel up 17 flights to keep a children's hospital powered on, or a MUD that's been running continuously since '93? Sure, inspiring as hell.
"Our company's product will be offline for a few days and we'll theoretically lose money/marketshare, but all our customers will be super understanding because they were all watching the news and, holy shit, did you see that tunnel get flooded, and then everything will turn back on and be fine with no extraordinary effort on our part...?"
If MUD is the type of game you're falling into the same trap as the one you're signaling but worse. You're putting in the same sentence "saving kids' lives" and "the bragging rights for a game's uninterrupted session" on equal footing.
When we went offline customers were super understanding!
Except one customer who was outraged and we felt totally comfortable banning their account.
This was a startup. The company getting fucked could mean our jobs going away. It was our equity too.
Not everybody is just punching a clock, while resenting the owners of the company. In our case, we wanted our company to do well because we liked working together.
In the case of Fog Creek during Sandy, because the software we hosted provided planning and project management for a lot of people, going offline would have a pretty large ripple effect. Sure maybe 95% would've been fine, but I knew of customers who e.g. managed emergency services and crisis response, and so keeping them online was meaningful to me.
Quitting a job because of this is so funny to me. 2012 was truly a different time.
If there was electricity to run the window washing crane, they wouldn't need to haul diesel.
You can do a barrel knot, though with a modern metal barrel, you can also generally just do a hook on top an bottom and lift horizontally.
My comment was more about the goofy shit people do to make the failure mode simpler next time instead of eliminating the failure mode all together. Some of the places I've worked would probably have made one of us get trained on the lift or something instead of make sure we have the fittings on had or a fuel pump and hoses with 17 floors of head pressure. That would be the temporary fix and then when they got the quote for the pump it would become the fix.
The chance of a hurricane within 24 months was never 100%. Looking back, it did happen, but you can’t judge future-facing decisions on post-hoc information that you didn’t have at the time.
There is a difference between “no one wanted to admit it wasn’t 0%” and “no one else thought the risk-adjusted return was positive.” When people decline to take some preventative measure, that’s usually what they’re saying implicitly.
I’ve been where you are—frustrated that you predicted and told people exactly what would happen, they didn’t listen, and you were right—many times in my career. It’s infuriating. Part of learning from it is figuring out how to get people to change, not merely figuring out how to be right.
My guess is that this kind of black/white Strong Opinions approach has bitten you throughout your career.
This anecdote hits differently for me. I grew up reading about Fog Creek and consuming all of their writings (and that one film) as an example of what a software development company should look like. It feels strange to have the curtain peeled back to reveal that something like backups, which would be considered basic requirements today, were not part of their early operations.
It really was a shocking wake up period for me in the realities of the real world vs. how we often perceive things.
IIRC it was the whole "Wasabi" thing that kind of opened up my eyes that there was a lot more smoke and mirrors involved than I had realized before.
Or you could label it Hyrox/Crossfit/$FITNESSTREND and charge people money to do it. #barrelpump #verticalmarathon #fitnessgoals #blessed
Im all for doing some dirty work when needed, but if they expect 100% uptime they should be ready for something as common as a storm flood without risking injuries hauling fuel drums by hand.
If you drop back to 30m of lift, you get still have 36PSI which is enough to get some flow. But that means that you need a holding tank (or lots of buckets) and another pump for another 30m of lift. That gets you to the 16th floor so you need another holding tank and a another pump.
That's a lot of complication. And getting those pumps to sit still can be a challenge.
Firefighters are heroes who bravely enter burning buildings, beloved by children and adults alike, muscular figures who get pictured on sexy calendars and thanked by tearful families.
Officious paper-pushers who make a fuss about a few cardboard boxes stored under the fire escape stairs (like me!) seldom get pictured on sexy calendars.
A firefighter entering a burning building is literally risking his or her life at that moment. A paper pusher who works on preventing the fire in the first place is normally very far from danger.
That's a common bias in management. Heroic fire fighting gets rewarded while silently preventing a crisis with proper planning isn't. Nobody ever gets credit for fixing problems that never happened.
Exactly why the general public thinks Y2K was a "hoax" or "overblown".
Someone hit on the notion that it’s Stockholm Syndrome. You bond with people you go through shit with. You don’t bond with people who prevent the shit. That was a super awkward conversation. The asshole in charge quit shortly after it, but 2 years later we still occasionally called each other out for making plans that accounted for a toxic chain of command that was no longer toxic.
Unfortunately, that typically meant when you were paged, the problem was probably significant, hard, or it was another team and you were just being triaged in because you were affected or might be able to help root cause.
Anyway, I was the first Fog Creek employee on site at 55 broadway (one of our founders got their first, and two peer1 employees were also on site) and also part of the first wave of the "bucket brigade". Another dozen of us showed up over the coming hours as we assessed and fixed things.
The media coverage on this was never super accurate-- we were carrying (and later, relaying) 5 gallon covered plastic buckets (like the home depot buckets with sealed lids) up to the landing where the generator was to top up its reserve tank. At one point SquareSpace (who was also running exclusively out of this DC) hired a group of muscle to bring a 55 gallon drum up, but that wasn't us.
I don't view this as lionizing failure, though I can see your perspective. Across our products (then FogBugz, Kiln, Trello) we had different backup strategies in play, but we were definitely putting a lot of eggs in the local-datacenter-that-claimed-to-have-resilient-power basket. Kiln in particular had decided to use SANs for data resilience, and that proved particularly troublesome for trying to find a workaround once the storm hit-- FogBugz on the other hand was able to spin up an EC2 version and start a slow-but-would-have-worked-eventually migration (and we even had an employee hit the road for us-east-1 location with a bunch of encrypted backups so that we could get a fast data restore online), and Trello fast-tracked-and-completed its local-to-cloud migration. While it was true that the DC had a backup generator, the fuel tanks for the generator being in the basement and becoming unavailable was not part of their resiliency plan. I don't fault us for not catching that.
...although I did have to later write the dreaded "I'm sorry, we've been hacked" blog post as a direct consequence of us not properly deprovisioning a "temporary" AWS credential that we created during that sprint. I will fault us for not catching that.
Shortly after Sandy, I switched my role from core dev on FogBugz over to DevOps, and I built out the backup and restore resiliency we had been missing. It wouldn't have kept us online during Sandy, though.
It was dangerous, and a couple folks slipped and fell. That's not a good thing and I won't celebrate that. Many employees opted out of the bucket brigade for family, safety, or personal reasons. I threw out all the clothes I wore that week as they reeked of diesel... not great.
I wouldn't wish that experience on anyone, but having come through it with fortune and no slip-and-falls, I still think about:
- Walking around FiDi post storm and seeing the flooded tunnels, clogged with cars that had been washed into them from the street parking.
- Dropping off big armfuls of coffee and bagels in the lobby for the building work crews. The folks who were essentially dredging out the basement and getting the building back online did much harder work than us.
- Going shopping for aquarium supplies in order to best jerry-rig the pump that our founder has repurposed from the office fish tank in order to transfer diesel from the drums on the curb into our 5g buckets.
Something that kept me going was that I knew that one of our FogBugz customers was the emergency response service in Haiti, and they got absolutely walloped by the storm. They were using FogBugz to coordinate relief, so there wasn't any way I was going to let FogBugz go offline if I could help it. Turns out that meant climbing a lot of stairs.
The issue that I remember was that we had so many SQL Server databases that it was impossible to keep anything vaguely approaching a sane backup strategy; SQL Server just isn't designed to handle as many as we had. (I in fact remember talking to a famous SQL Server consultant [edit: this was likely Brent Ozar] we were using who said "that won't work," and then I or someone else was tasked with talking to a SQL Server dev at Microsoft about it and they ALSO said that it couldn't work, and when I brought this up, I was told that Microsoft was wrong. IIRC, the implication was that it would take an extremely long time (many days) to restore, and that the snapshots would likely be days old. I don't think any of that contradicts your memory about the very a slow restore going on, but might add some color.
(Likewise, Kiln supported multiple backend instances and had admin commands to copy and move repos between them, so it the SAN had specifically been the issue, we had workarounds that would've been fine with whatever EBS was up to those days. But I remember the databases being the blocker.)
Our "One database per tenant" model was also the reason we were so slow to go to AWS-- the MSSQLServer licensing terms were cost-prohibitive.
However recognizing that around 90% of the FogBugz databases were free-tier and didn't get daily changes meant that teaching the backup scripts to skip creating redundant backups solved most of the issue.
Post-Sandy, I was able to solve most of this by:
1) Shipping full, diff, and incremental backups at different schedules
2) Scripting the ability to create warm standby servers powered by the latest fulls, then diffs as they come online, then the logs at the cutover time. (so you can provision a new server and seamlessly cut over to it)
3) Using our own scripts instead of the RedGate tool (which was _fine_ just not equipped to handle as many DBs as we had) to orchestrate the backups and restores. This in particular was needed for the "only back up modified dbs" fix.
This got us to most of what we wanted-- 0-downtown transitions of databases across servers, nightly offsite backups, fast backup runs. The final problem-- if our datacenter evaporated and we had to reprovision everything from scratch -- was left as an acceptable risk of "we'll fetch the nightlies from s3/Glacier while we do the slower part of rebuilding our servers."
But to your point, during Sandy, none of the above had been fixed yet. I think our sysadmin team was very strong, but they weren't programmers (and/or didn't have enough time to step back and program), and I was able to pick a lot of low hanging fruit in DevOps
There is no reason that they couldn’t have had more than 1 DC when Sandy hit.
I guess it helps that I grew up in a rural place and did construction for a while, so lugging gas to a generator wasn't a new thing for me.
Also I was... 25? I'd be more cautious today, I think.
Awesome, story time! I was working at a not brand new but still relatively small tech company (~$100m yearly revenue) and we were in two major datacenters, one in Las Vegas, and another in Manhattan. We were essentially telephony in the cloud, back when that was still fairly novel. At the time, most of our revenue came from handling various 1-800 campaigns, like "Ronco Knives" if you remember that, and also big money raising telethons. Our platform was active/active between the two datacenters. Of course not for a given phone call, but we could pretty easily balance traffic between both datacenters, even fully away from one of them, though as we'll see later, at some peril.
Our main networking guy was also the main datacenter person, and he'd been in NY to help set us up there. I can't remember which datacenter it was, but we had like half of a floor well above the street.
I was watching Sandy carefully, and once the odds of it hitting Manhattan were substantial, I started to press for plans to move all traffic out of the NY datacenter no less than 24 hours before landfall.
The network guy disagreed; he said that he'd been there, on site. He saw the redundant generators. The redundant network ingress and egress. The site people joked that if somehow the first floor ever flooded, the the rest of the datacenter would continue operating normally.
I noted that while all that might be true, we don't have any such assurances for all of the other equipment that'll be underwater needed to support the network.
I ended up escalating to our boss, the head of operations. He was one cube over. :) We had a vigorous discussion, and the VP agreed to pull our traffic out of NY. So we did.
Probably 75% of our capacity at the time was in LV and 25% in NY. We were running pretty hot utilization wise, but LV could, under normal circumstances, handle the whole platform by itself.
Sandy came and flooded. Generators kicked on. The network stayed up. Then the DC/networking guy got an emergency call from the datacenter.
They had plenty of fuel on site, but it turns out that the fuel inlets for the generators were...under water and could not be reached. So everything was going to go down in a couple of hours.
And it surely did, for reasons I've forgotten, in a very crappy way. Our logs indicated that power hard flapped for all of our equipment quite a few times.
We lost quite a few PSUs due to that. Yes, we definitely should have shut our equipment down cleanly. Oops.
Sandy power disruption always reminds me of this photo - burned into my brain - https://archive.is/DoZsr
sandy was a wreck
the only time i remember executing on DR plans (which went well!)
(i became a consultant by the time covid hit, so i didn't get to do the whole covid scramble every sysadmin did)
The NYSE, NASDAQ, and most other exchanges and brokerages run in datacenters in Cartaret, NJ and Mahwah, NJ. I can't think of a good reason why anyone would put a datacenter in Manhattan.
This is the actual NYSE datacenter https://maps.app.goo.gl/Da1E6Je9eJAGA1RK8
This is the actual NASDAQ datacenter https://maps.app.goo.gl/ZS9VJcZmf8VtvWEd6
One of the best decisions I made around that time was installing two backup Towerstream lines. When everyone else was out for weeks we were up and running, full operation.
There was about $5M of gear in my wife’s grossly overloaded minivan. The bean counters tried to deny my mileage because I was using a personal vehicle without prior auth. lol. My boss was ready to kill.
Not fun but educational.
It took me an embarrassingly long time to realize that (my 30's), and now I keep my conversations with my coworkers very utilitarian and divulge minimal information about myself. I try not to be a dick or anything, I'm polite enough, but pretty much everything I talk about is either directly relevant to work and/or so surface level as to border on "vapid".
ETA:
Just as an example, I used to have really bad sleep issues in my 20's that would make it so I wouldn't be able to get to sleep until around 4am, if at all. I made the mistake of bringing that up at work to my manager. Two weeks later I got a phone call at 3am because there was something he wanted to look at and "he knew I would be up". I was up, but I didn't really have much desire to get my laptop out and do work. I liked that job at the time and I liked my manager, so I did it anyway, and it wasn't until years later did I think about that incident and how utterly weird it was that I was sort of being punished for my insomnia.
My sleep issues have gotten a lot better now, so if someone calls me at 3am I probably won't hear it because I'll be asleep, but I remember that incident now and have it constantly spinning in my brain when I chat with coworkers.
(otherwise where I work, I also avoid talking about politics for instance, but this is a different theme)
A) because picturing at my current work place everybody learning of all my opinions at once and the resulting drama
B) I am in progressive spaces quite often, but they are far from an echo chamber with everyone having the same opinions either ..
* No
* Only if I get on-call pay
Still, I do think there's merit in keeping your personal life to yourself and not telling your coworkers more than they need to know. I'm not saying you should be stoic or a dick or anything; by all means engage in conversations and be nice. I'm just saying that it's important to have boundaries.
I tend to be someone who overshares a lot (hence my enormous word count on Hacker News), so I just try and keep this in mind.
Maybe I'm lucky. I've so far not had to endure a job for which I was underpaid and expected to meet and exceed every contractual obligation placed on me.
I received many strange requests like to come in Friday night while drinking at a concert and ransack the offices looking for a laptop or possible phone left behind that they suspected was triggering alerts by repeatedly trying to connect to a foreign server. "You want me to drunkenly rob the offices at night and walk out the lobby carrying a stack of laptops? Ok"
* https://www.f-t.com/post/emergency-power-off-in-data-centers
EPO disconnects for generators was around earlier than the above:
* https://www.mgiepss.com/blog/emergency-power-off-epo
You'd generally not want to have the sprinklers go off while things were energized, but there could be a mechanism to monitor flow, and only cut power if sprinklers were actually sending out water. But even if this was present, your first priority would be for the sprinklers to not go off as there would be a lot of clean-up / drying needed after the fact.
Unless you have a mist system (NFPA 750), sprinklers put out a lot of water.
The CRAC won't feed a fire oxygen as it's just recirculating the same oxygen depleted air.
https://signalsandthreads.com/the-thermodynamics-of-trading/
He had to deal with so much absolute bullshit to get the diesel generator I mentioned in the story done, and then Hurricane Sandy destroyed it instantly a few years later. Felt bad for him.
I've once done the opposite: the office "server room" (a small room with three racks, two of them for the patch panels, one for the servers and switches) was cooled by a mini-split which was prone to locking up with the compressor enabled (the relay which controls the compressor got stuck) and freezing, which then stopped the air from circulating due to all the ice. I wrote a script which ran on one of the servers which read the ambient temperature through IPMI to detect when the room was too cold, email a warning, and power off (using ssh) the neighboring servers and then itself.
We mostly hear about engineers from Jane Street here, and not much from their traders and finance people. I wish we could have some other perspective sometimes.
They publish a great podcast, Signal and Threads, I really recommend.
The latency impact would be huge
If there is a fire the decision is: all equipment might be lost in a fire OR all equipment will be lost by water damage.
But in addition, this particular story takes place in a skyscraper in Manhattan, with the data center in question occupying a portion of one floor, with office space surrounding it. So letting a fire burn would create significant risk to human life as well.
remember sandy? that was crazy
Many many years ago someone told me that your data is not safe if is not at least in 3 very different physical places. Today, I would probably add the requirement to be on different providers as well.
Why not just turn the CRAC back on?
> you don’t trust the automation unless there’s a big red-painted, mechanically connected lever you can pull to turn it off
Seems like this is one of those cases where you want to override the automation.
Incidentally, we actually had a fire in the office building I worked in long ago. The HVAC didn't shut down, and smoke started to come out of random vents throughout the building. It was chaos. And yet, plenty of workers wanted to finish and send that email they were working on, commit code and push it to the remote server, etc. What sticks with me most is the one guy, a very senior manager, running around yelling at people to get off their computers and get out of the building, because no job is ever worth your life.
Building fires are no joke. Get the F out.
33 Thomas St is basically one huge datacenter, that's why it doesn't have any windows. https://en.wikipedia.org/wiki/33_Thomas_Street
Some of them have half-floors for all the cable routing, where you can't really stand up.
I think this is why I like them. To me they're not just an ugly box but I always wonder what's inside.
Nitpick, but the headline is not confusing at all, even though a couple of people are apparently confused. I suspect that is the fault of a hasty read through, not the headline itself.
I guess I’m the last one to hear it.
But it's only used in military and aviation these days anyway.
So fire suppression and data centers don't mix too well
https://www.hack.org/~mc/texts/vaxen-joke.txt
Realizing that I'm spoiling some of it... but we're here, so some of it is inherently spoiled...
The fire alarm klaxon went off and the siren warning
of imminent halon gas release was screaming. We
started to panic but the data center manager shouted
over the din, "Don't worry, the halon system failed
its acceptance test last week. It's disabled and
nothing will happen."
He was half right, the primary halon system indeed
failed to discharge. But the secondary halon system
observed that the primary had conked and instantly
did its duty, which was to deal with Dire Disasters.
It had twice the capacity and six times the discharge
rate.But yes, I've seen low-O2 installations also. These have other issues though, like no protection when humans need to be in there for maintenance. And the need to be very very careful when reducing it again, making sure everyone is out. It happened to my colleague once that this went wrong and while it's not low enough that you'll die, he wasn't exactly feeling great for the next few hours.
You just have to reinforce the ceiling panels and the windows so they don't get blown out, but it's not really an issue otherwise. It was all certified and regularly checked. Water is indeed a bit counterproductive as it's not really much better than fire for the equipment in there.
PS: Halon itself is of course gone but it's other less harmful gases these days.
What also happens is low-oxygen environments (they only raise the O2 when maintenance is being done).
Many people (like me) use "halon" when they are talking about general inert gas based fire suppression as there's not really a good short word for them otherwise.
Reminds me of DirectNic's datacenter flooding during Katrina (New Orleans), we all kept posting on Somethingawful until they finally shut it down.
I smell AI