Ask HN: What's the worst you've ever screwed up at work?

https://news.ycombinator.com/item?id=7144934

by kadabra9 • 12 years ago

217 307 12 years ago

news.ycombinator.com

We've all been there (most of us, at least). What did you do (or not do), how did you first react, and how did you handle it?

Bonus points for sharing what you learned/key takeaways from the experience.

zimpenfish• 12 years ago

Many years ago, when I was but a fresh faced idiot, the partition that contained the mSQL database which had All The Data filled up. I moved it into /tmp because there was plenty of space.

On a Solaris box.

Hilarity ensued when we next rebooted it.

ambiate2• 12 years ago

For those who don't know, solaris uses tmpfs for /tmp. It is a virtual memory/swap based file system. Anything in /tmp is actually temporary if the machine reboots/powers off.

scott_karana2• 12 years ago

I like setting this up on Linux machines too. There are tons of ephemeral files that get written there, depending on your usage case, and I'd rather not waste the IO for writing pids to lockfiles. Disk is cheap, but RAM is fast and cheap. :)

rfreytag• 12 years ago

About 30 years ago I deleted the JOBCONTROL process on an old VAX 11/780 thinking it might be the reason why someone's process was stuck.

It wasn't a but an hour before I lost sysadmin privileges.

Never "experiment" with a production system - ever.

krak3n_• 12 years ago

The worst thing I have done is terminate a running production instance with no database backups.

Client, not happy.

gmays• 12 years ago

In late 2008 when I was in the Marines and deployed to Iraq I was following too closely behind the vehicle in front while crossing a wadi and we hit an IED (the first of 3 that day).

Nobody was killed, but we had a few injured. Thankfully the brunt of it hit the MRAP in front of us. If it hit my vehicle (HMMWV, flat bottom) instead I probably wouldn't be here.

That was the first major operation on my first deployment, too. Hello, world!

My takeaway? Shit just got real.

We ended up stranded that night after the 3rd IED strike (our "rescuers" said it was too dangerous to get us). It was the scariest day of my life, but in similar future situations it was different. I still felt fear and the reality of the existential threat, but I accepted it. It was almost liberating. Strange.

I deployed for another year after that (to Afghanistan that time). After Afghanistan I left the Corps and started my company. Because if it fails, what's the worst that can happen? Lulz.

kadabra9• 12 years ago

This really puts some of the boneheaded moves I've made in my career in perspective. One thing that's always kept me pretty even keeled after a blowup is to take a breath and tell myself that no matter how bad I've screwed up, I'm still here, still breathing, and there (most likely) is some way out of the hole I've dug, no matter how painful.

Depending on the industry, that might not be the case though. Thanks for your service.

contingencies• 12 years ago

First job, circa 2000, at an ISP that was run very clearly as a business and cutting corners. Not only was it critically understaffed, but management was more interested in laughing their way to the bank than management. They had me - with literally no routing protocol experience - manage a live route advertisement transition between two peering providers. Result: all customers offline, ~24 hours.

Reaction was standard: mostly to point out I did my best in unfamiliar territory and things should be sorted soon.

Take aways were: (1) less support calls than expected - users put up with things. (2) you learn when you fail (3) always have a backup

They kept me on at that job but I left pretty soon anyway as I got a 'real' (as in creative) job hacking perl-powered VPN modules for those Cobalt Raq/Qube devices, and building a Linux-related online retail venture for the same employer ... that worked great, but failed commercially.

penguinlinux• 12 years ago

I worked at an ISP in NY exactly around 1997-2004 we had also the Rac/Qube devices and I had to manage stuff I was not familiair with :) I learned so much by trial by fire.

joshbaptiste• 12 years ago

In 2001 my first IT tech job as help desk analyst I heard beeping in the server room on one of the Solaris/Oracle machines and pressed the power off/power on button on the chassis. DBA came running in and I promptly left saying "oh I think it rebooted itself". The company went bankrupt shortly after so no huge lashing came my way but all my more experienced friends where like "wtf never do that again!"

ambiate2• 12 years ago

Was probably a program running beep codes, haha. I have also had servers executing beep codes that brought great anxiety to me for endless hours. Turns out it was just some debug alert calling the motherboard speaker beep.

kisamoto• 12 years ago

Introducing a master/minion update system to work I ran a batch update to take a certain percentage out of the cluster.

Unfortunately I got my selection criteria wrong and pulled out all of one cluster and half of a second, halting a few thousand operations.

Luckily the monitoring system was very quick to alert me of this and using the same (wrong) selection criteria it was a fairly simple process to stop the update and put them all back in the cluster.

Takeaways? The age old cliche of "With great power comes great responsibility". Oh and have good monitoring!

kentwistle• 12 years ago

git push -f

hluska• 12 years ago

A local Subway franchise was the very first company that hired me. I was extremely young, shy, and intensely socially awkward, yet excited to join the workforce (as I had my eyes set on a Pentium processor).

When I worked at Subway, the bread dough came frozen, but you would put loaves in a proofer, proof it for a certain amount of time, and then bake it. My first shift, however, got busy and I left several trays in the proofer for a very, very long time. Consequently, they rose to roughly the size of loaves of bread, as opposed to the usual buns.

It was my very first shift alone at any job in my life, so I did the most logical thing I could think of and put the massive buns in the oven. They cooked up nicely enough and I thought I was saved. Until I tried to cut into one.

Back in that day, Subway used to cut those silly u-shaped gouges out of their buns. In retrospect, I think this was most likely a bizarre HR technique designed to weed out the real dummies, but at the time I was oblivious (likely because I was one of the dummies they should have weeded out). When I ran out of the normal bread, I grabbed one of my monstrosities, tried to cut into it, and discovered that it was not only rock hard, but the loaf broke apart as I tried to cut it.

That night, my severe shyness and social awkwardness had their first run-in with beasts known as angry customers. I was scared I would get fired, so I promptly made new buns, but spent the rest of my shift trying to get rid of my blunder. I discovered some really interesting things about people that night. First, you'd be surprised how incredibly nice customers are if you are straight up with them. Some customers I never met before met the big, crumbly buns as an adventure and, in doing so, helped me sell all the ruined buns.

In the end, I came clean (and didn't get fired). That horrible night was a huge event in the dismantling of my shell. It taught me an awful lot about ethics. And frankly, that brief experience in food service forever changed how I deal with staff in similar types of jobs.

canadev2• 12 years ago

I gotta say, that's a pretty awesome story. Didn't expect that to be the seeds of transformation.

benaiah• 12 years ago

Dear god the buns. Sympathize greatly with you there.

ericcope• 12 years ago

I think every new hire at Subway does this. They did in my tenure there. I did it more than once...

lostlogin2• 12 years ago

This reminds me of reject analysis week as a radiography student. People would be hiding their crap films (film and chemistry people!) up their tops, behind shelves, basically anywhere. Now days the clever ones know how to dick with the server. I have never deleted films for this reason, but have deleted films to keep incidents quiet.. (Boss must not know I got a chunk of steel in my hand prior to a shift in MRI etc)

michh• 12 years ago

Classic forgetting the full WHERE-part of a manual UPDATE-query on a production system. The worst part is you know you fucked up the nanosecond you hit enter, but it's already too late. Lesson learned? Avoid doing things manually even if a non-technical co-worker insists something needs to be changed right away. And if you do: wrap it in a transaction so you can rollback, leave in a syntax error that you'll only remove when you're done typing the query.

sugerman• 12 years ago

I did this once and have since always typed the WHERE first in an UPDATE.

jhonnycano2• 12 years ago

UPDATE ... UPDATE WHERE ... [go back, go back] UPDATE SET ... WHERE ...

Volundr2• 12 years ago

Been there done that. Usually I always work inside a transaction, and carefully examine the results before typing that all important 'commit'. But a "simple" change at 4:55 and me in a hurry to get home....

dboyd• 12 years ago

Exactly the same...

I was hired by my college to build a grade management system in my second-to-last year there. I was in a hurry due to a lunch meeting with other IT staff at the University, forgot to add the where clause, and suddenly every single student was a Computational-Science major (mine).

Funny part of the story was that the moment it happened I uttered "oh shit." My boss, who sat beside me, said "what'd you do?", and about 15 IT staff from other departments walked into the office to go out for lunch. I'm sure I was an interesting shade of red.

I had to explain what I did in front of all these people. My boss laughed out loud, brought the system offline, and simply said: "well, after lunch we get to test our backup process." We went for lunch.

Two valuable lessons I learned...

People make mistakes, that isn't a problem, it's how they respond that's important.

Don't try and solve hard problems when emotions are running high. If shit is going down in production, the most important thing to do is to breathe, and get a glass of water. That little bit of time helps a lot.

marvvelous• 12 years ago

I did something similar once. Now my query writing goes:

UPDATE
UPDATE SET url='new_url' WHERE source_id IN (etc)
UPDATE sources SET url='new_url' WHERE source_id IN (etc)

gargarplex2• 12 years ago

I always add a LIMIT even when not necessary.

Why doesn't MySQL have a version control baked in? Even if it preserves just the last n hours of state..

michaelmior2• 12 years ago

That's what backups are for. With appropriate checkpoints, you can roll back to any point in time with a reasonable level of precision.

icedchai• 12 years ago

It kind of does. It's called the binary log.

DCoder• 12 years ago

Since you're using MySQL, you can start it with --i-am-a-dummy [1]. Not a backup, but still handy.

[1] http://dev.mysql.com/doc/refman/5.6/en/mysql-tips.html#safe-...

Davertron2• 12 years ago

I've done similar and now I almost always write a select first and then only after I've verified I'm getting the rows that I expect do I update my query to an update/delete.

Volundr2• 12 years ago

Reading all of these makes me think, the admin tool for your database of choice should probably put you inside a transaction by default, and require you to explicitly commit changes. For the madmen, it could still have an auto-commit mode, but should be opt-in rather than the default.

codegeek• 12 years ago

In this case though, wouldn't you have to COMMIT before the actual update happens ? Usually in production, it is not a good idea to have auto COMMIT on.

simcop2387• 12 years ago

Yep, I typically do SUPDATE just for fun that way. And only do that AFTER building the where clause with a SELECT * FROM foo WHERE ... so that I always start with the clause when making the update. Might be paranoid but it always seems to work out for me that way.

glogla• 12 years ago

This is why, while I hate Oracle and everything they represent as a company, I kind of like their database because of the flashback feature. You can do

  SELECT * FROM table AS OF TIMESTAMP some_timestamp;

and that is pretty practical. It works online, no restore, no nothing, and while it only works as long as the old data are in logs, on a production system, you should have the spare space to have some history. Theres also FLASHBACK TABLE tab to BEFORE DROP but that shouldn't happen, right?

Of course, you should probably do every update of production data in a transaction, check the result and then commit, and if you want to be sure, you can do UPDATE ... RETURNING to check what's changing. Autocommit on manual access to production is pretty crazy. But still, flashback is useful.

miahi• 12 years ago

I usually allocate a large part of the free space to FRA. In the production system I use right now (about 2TB of data, 50GB changes/day), I can go back a couple of weeks if needed. Fortunately everything is stable now, but that flashback was quite useful a few times.

jraines• 12 years ago

I wish I didn't know the stomach-plunge feeling of realizing you've done this

jpatokal• 12 years ago

This is why you have SET SQL_SAFE_UPDATES=1; (or equivalent) in your DB shell startup. It only takes one UPDATE users SET password='foo'; to learn why...

kchoudhu• 12 years ago

Ah, memories.

I did this in a production database (thought it was a QA environment) and brought trading on the mortgage desk of an investment bank to a grinding halt on September 14th, 2008.

The DBAs saved my 23 year old ass that day. I make it a point to send them beer on 9/14 every year.

JasonFruit• 12 years ago

I sent an email to three thousand insurance agents informing them of the cancellation of policy number 123456789, made out to Someone Funky. I learned to appreciate Microsoft Outlook's message-recall function, which got most of them. I also learned that just because you're using the test database instance doesn't mean nothing can go wrong.

earino3• 12 years ago

easy:

me: "unix definitely won't just let me cat /dev/urandom > /dev/sda"

other: "sure it will"

me: <presses enter>

what I learned? unix will absolutely let you hang yourself. 1998, production server for a fortune 5 company.

krisdol• 12 years ago

Why in heaven's name did you try that in production?

earino3• 12 years ago

it was 1998. i was young and foolish. no, seriously, i was 19. i was also SUPER convinced that it wouldn't work :) I have since learned to be waaaay less convinced since then.

hcarvalhoalves2• 12 years ago

Professional suicide.

earino3• 12 years ago

I respectfully disagree. I was a junior person who was given root level privileges on a production server. There were many layers of process and management that weren't in place. Today, I am a senior person at a multi-billion dollar company. I would never call "professional suicide" at a junior person who made a genuine mistake. I would scold them, and then I would try to figure out how we got into the situation where someone unqualified could cause us such a bad thing to happen.

lauradhamilton• 12 years ago

That's quite funny.

Linux will indeed let you hang yourself. I maintain that every experienced Linux user has seriously messed up their system at least once. More often personal / dev / test environment than production though...

killertypo• 12 years ago

During a server migration for our web based file sharing system our lead engineer (at the time) forgot to ensure that all cron jobs (for cleaning up files and sending out automated emails) had been turned back on.

Queue me 7mos later reviewing the system. Realizing that critical jobs were no longer running and that our users were all essentially receiving 100% free hosting for however much storage they wanted. SOOOO i turned the jobs back on.

The lead engineer before me left no documentation of what the jobs did other than that they should be run. In my stupor i did not review the code. The jobs sent out a blast of emails warning that files would be deleted if not cleaned up or maintained. Then seconds later deleted said files...

We nuked around 70GB worth of files before we realized what happened. WELL GET THE TAPES! Turns out our lead engineer ALSO forgot to follow up w/ system engineers and the backups were pointed at the wrong storage.

No jobs lost, thankfully the manager at the time was a word smith of the highest degree and can play political baseball like a GOD.

michaelochurch• 12 years ago

Tried to prevent a massive product failure.

It failed anyway, but I wasn't around when it did and there would have been no "I told you so" credit even if I were.

One of those "big company" lessons, but probably applicable to startups (which have an even higher ego density).

kenesom1• 12 years ago

Bummer, they should've listened to you...

Some companies are using internal prediction markets, where employees can speculate on various initiatives:

http://en.wikipedia.org/wiki/Prediction_market#Use_by_corpor...

Open allocation seems like the ultimate prediction market - people deciding on what initiatives to work on and invest their time in is a stronger signal than dysfunctional internal politics. And people are likely to be more motivated working on projects they think have value.

reppic• 12 years ago

One time I tried to change a column name in a production database. I learned that when you change a column name, mysql doesn't just change a string somewhere, it creates a new table and copies all the values from the old table into the new one and when that table has millions of rows in it, it really slows down your production server.

xutopia2• 12 years ago

I keep thinking that's the most ridiculous thing ever.

cgh• 12 years ago

I was in a remote meeting and failed to realise my laptop's camera was broadcasting. A roomful of people saw me, clad in horrid workout clothes, jam my finger up my itchy nose and scratch my balls.

Key takeaway: always check the cam.

xauronx• 12 years ago

And make sure your phone is muted. The first conference call is easy. When you have one every day for a year and it becomes so common place... sometimes you forget. I've heard some people on my team coughing obnoxiously, yell at people driving, doing the dishes, etc. Mute your shit, and tell your team mates immediately when they aren't muted.

webstonne• 12 years ago

I asked them for a job in the first place.

yan• 12 years ago

Not the worst at all, but probably one I found most amusing. One of my jobs included some sys admin tasks (this wasn't the position, but we all did dev ops), among my other responsibilities. I spent half a day going through everything with the person responsible for most of the admin tasks at the time. She was an extremely dilligent and competent admin, did absolutely everything through configuration management and kept very thorough personal logs and documentation on the entire network. One of my first tasks was to change backup frequency (or other singular change) and going by how I usually did things at the time, just sudid a vi session, changed the frequency and restarted the service.

She found out about it pretty quickly due to having syslog be a constant presence in one of her gnu screen windows and gave me a look. She quickly reverted what I did, updated our config management tool, tested it, then deployed it, while explaining why this was the right way to do things. I slowly came around to doing things the right way and haven't thought much about the initial incident until we found her personal logs that she archived and left on our public network share for future reference.

In the entries for the day that I started, we saw the following two lines:

    [*] 2007/09/09 09:58 - yan started. gave sudo privs and initial hire forms.
    [*] 2007/09/09 10:45 - revoked yan's sudo privs.

cunac• 12 years ago

not so bad 47 minutes in first day :-)

sillysaurus22• 12 years ago

She found out about it pretty quickly due to having syslog be a constant presence in one of her gnu screen windows

I'm amazed that this is possible. How would I set something like that up? A realtime log of only the most significant events of a remote system?

In fact, I'd like to take this opportunity of ignorance-admitting to ask the community for general linux/bsd sysadmin resources. What books should I read, or what topics should I study? I want to become an expert at modern sysops. Modern deployment, hardening, backup, managing dozens of boxen, etc.

I've been thinking of going through any MIT OCW on the subject, but it seems like hard-earned experience might not necessarily translate well to an academic setting. What would you recommend I do?

ixmatus2• 12 years ago

Papertrail is great for this...you can of course setup syslog to route to a central server and just be logged in tmux / screen on that particular machine to read off the stream of logs (I prefer papertrail though + saved searches and hipchat "pings" when saved searches are matched on incoming events).

General devops / sysops/ sysadmin knowledge can be had through a variety of means - I got most of my knowledge from simply reading the FreeBSD manual and making a lot of mistakes with my own servers.

sentenza2• 12 years ago

What position are you starting from? My old workplace was a university group where we (admins) were recruted from the available pool of PhD students. So I'm used to guiding people from "no knowledge" to "enough knowledge to be dangerous". The first step was to force the prospective admins to run a specific system on their "productive machine" and keep it in such conditions that _everything_ works.

This way, a complete admin newbie would learn about digging through the systems by working out the kinks of practical everyday problems. Remember, this is only the most basic instruction, nowhere near enterprise-grade.

If there was a "prospective admin" who had never before run Linux, I'd tell them to install and use Ubuntu/Mint. (Those guys whould usually only be trained to be a helping hand for a "senior" admin.)

If he'd already used Ubuntu at home, I'd tell them to start using Debian and work out how to set up an SSH server and set up their home machine so they could access it remotely.

If they had dabbled with Debian, Fedora, SuSe or something similar, I'd tell them to install Arch and set up some "interesting things", like a mail server or a nis server.

If they were using Arch or Gentoo at home, I'd just personally show them the important things about our system and have them wingman with me for a few days.

If you are already an advanced Linux or BSD user, my approach is of course not appropriate. Instead I'd recommend to pick skills that you want to learn (iptables? Exim?) and set that up. Read Manuals! Read RFCs!

Best of luck.

kordless• 12 years ago

Splunk: http://splunk.com/. Free download, free to use up to X many GB.

dredmorbius• 12 years ago

Remote syslogging, generally.

There are any number of log tailers. One I've recently discovered and love is multitail by Folkert vanHeusden:

http://www.vanheusden.com/multitail/

Incidentally, a related utility (and one which works really well with multitail) is rsstail:

http://www.vanheusden.com/rsstail/

One option is to get an RSS/Atom feed of critical system events.

azth• 12 years ago

I've been thinking of going through any MIT OCW on the subject

Are there any MIT OCW courses on this? I haven't come across any.

atsaloli• 12 years ago

Check out www.sabok.org

Sys Admin Body of Knowledge

ggreer3• 12 years ago

One summer in college, I got an internship at a company that made health information systems. After fixing bugs in PHP scripts for a couple weeks, I was granted access to their production DB. (Hey, they were short on talent.) This database stored all kinds of stuff, including the operating room schedules for various hospitals. It included who was being operated on, when, what operation they were scheduled for, and important information such as patient allergies, malignant hyperthermia, etc.

I was a little sleepy one morning and accidentally connected to prod instead of testing. I thought, "That's weird, this UPDATE shouldn't have taken so long-oh shit." I'd managed to clear all allergy and malignant hyperthermia fields. For all I knew, some anesthesiologist would kill a patient because of my mistake. I was shaking. I immediately found the technical lead, pulled him from a meeting, and told him what happened. He'd been smart enough to set up hourly DB snapshots and query logs. It only took five minutes to restore from a snapshot and replay all the logs, not including my UPDATE.

Afterwards, my access to prod was not revoked. We both agreed I'd learned a valuable lesson, and that I was unlikely to repeat that mistake. The tech lead explained the incident to the higher-ups, who decided to avoid mentioning anything to the affected hospitals.

If it's any consolation, the company is no longer in business.

Just remember when you screw things up: Your mistake probably won't get anyone killed, so don't panic too much.

munificent2• 12 years ago

You didn't screw up here. The entire infrastructure, org chart, and policies that allowed you to accidentally modify a production database containing critical medical information screwed up.

Blaming yourself here is like blaming yourself for being hurt after being told to drive a car with no seatbelt or brakes.

ggreer3• 12 years ago

Sure there's plenty of blame to spread around, but I still would have felt terrible if someone had been hurt or killed.

What system would you put in place to prevent this? The issue was that I connected to prod when I thought I was connecting to a test DB. We each had different credentials for prod vs everything else, but the SQL client remembered my username and password. Anyone with prod access could have made the same mistake.

pc86• 12 years ago

He's not entirely innocent. It's more like blaming himself for being hurt after being told to drive a car with no seatbelt or brakes, while knowing that the car has neither, fully understanding what could happen, agreeing to it anyway, then driving 80 mph down a residential street while still groggy from waking up.

Regardless, I am very glad they are now out of business. :)

dbtc• 12 years ago

To be fair, it's still a little bit your fault if you end up crashing the car.

smartician• 12 years ago

Did a similar thing, but in a less critical domain (warehouse management). Updated the status of all packages to "NEW", which would have meant that everyone who ever ordered something from that company would have gotten another delivery for free, provided the articles were in stock.

We were able to restore the data pretty quickly, but we had to interrupt warehouse workflow for a few minutes. They were surprisingly accommodating, almost amused by my mistake.

matellis• 12 years ago

Uh, has anyone on this thread heard of HIPAA? I'm pretty sure having a summer intern get full access to actual patient data shouldn't be possible under a properly implemented set of HIPAA processes, and the same goes for the accidental UPDATE.

The story reminds me of the day I was introduced to "BEGIN TRANS", "COMMIT" and "ROLLBACK" when someone upgraded the Sybase console and helpfully changed the default setting so we didn't need those pesky semi-colons to finish a query any more. The result was:

  DELETE * FROM TABLE x
  131054 rows deleted
  WHERE a = "foo"
  >> Malformed query <<

Phone starts to ring a few seconds later as all the users saw their morning's work disappear.

This stuff is way too easy for us noobs. Thank goodness that with modern technology we've found ways to make sure it doesn't happen any more... :-)

HeyLaughingBoy2• 12 years ago

Honestly, as long as he's had HIPAA training, there's no difference between an intern making the changes and a 10-year employee doing so.

bradbatt2• 12 years ago

"That's weird, this UPDATE shouldn't have taken so long-oh shit."

That right there is one of the worst feelings in the world. I imagine that everyone on HN just felt it with you as they were reading it!

Totally agree that the higher-ups were responsible for not putting better roadblocks in place to prevent this type of thing from happening.

Davertron2• 12 years ago

I wrote an update script for a database table not realizing I had the key wrong (I'm kind of fuzzy on the details, but essentially I think it was a composite key but I was only using one of the columns in my WHERE clause...) and accidentally updated all customers addresses in our database to the addresses of one account.

Luckily we had backups from that morning so we only lost any address updates people would have done that day, but it made for some interesting customer service calls for awhile...

TheCapn• 12 years ago

I may, and/or may not have caused a production site's PLC to go into STOP mode during daily operations while making network updates remotely.

Possible outcomes of unplanned system haults include plugged machinery that would need to be manually cleared, mixed products which would become immediate net losses for the company and damaged motors.

Thankfully no product was being run at the time. I have also implemented changes across the board to our client sites that prevent this type of shit from ever happening again. You know when you look at a system and go "this is going to bite us in the ass eventually?" This was one of those systems, they just needed a new hire to give them the push.

yen223• 12 years ago

I wrote a piece of code controlling an assembly line machine. These machines require manual operation, and would come with a light curtain, which detects when someone places their hand near the moving parts, and should temporarily stop the machine.

A relatively minor bug in the software that I wrote caused the safety curtain to stop triggering when a certain condition was met. We discovered this bug after an operator was injured by one of these machines. Her hand needed something like 14 stitches.

Lessons learnt:

1. Event-driven code is hard.

2. There's no difference between a 'relatively minor' bug and a major one. The damage is still the same.

BuildTheRobots3• 12 years ago

Couldn't read your comment without a shudder and my brain going straight to the Therac 25 incident(s).

dkokelley• 12 years ago

I hadn't heard about this so I looked it up. Chilling.

http://en.wikipedia.org/wiki/Therac-25

HeyLaughingBoy2• 12 years ago

Another lesson your company should have learned is that a safety-critical system like this should not be left to software. Sure, monitor the curtain by software and send errors, but hardware should immediately stop the machine when the light curtain is broken.

frogpelt• 12 years ago

Non-tech-related:

I was doing HVAC work while I was in college and we were removing an old air handler from underneath a house. Just inside the crawl space, under the access door was a water pipe. My boss told me to make sure I held it down while we slid the air handler out through the hole. I lost my grip on the pipe and the air handle snapped it in two, at which point gallons of water began to gush into the crawl space.

I ran for all I was worth to the road, which in this case was about 600 feet away, to turn off the water at the water meter. I ran up and down the road in front of the house and never found the water meter. So I ran back to the house and inside and told the homeowner who promptly informed me that they used well water. She called her husband and he told us where to turn off the well pump.

It wasn't really that bad in the grand scheme of things but letting the homeowner's water gush under the house for about 15 minutes does not bode well when you are supposed to be there to fix problems not create them.

jmspring• 12 years ago

Early on in the implementation of one of the PKCS "standards" while at a browser company many years ago, due to an improper interpretation of a spec that was still in flux. There wasn't enough testing and "release bits" went live.

I had to quickly get a patch in for the improper code and had to maintain that buggy implementation. In addition, the "standard" itself got a rather scathing write up from Peter Gutmann, which is completely valid:

https://www.cs.auckland.ac.nz/~pgut001/pubs/pfx.html

This is a critique on the "standard" itself, the process was just as ugly.

byoung2• 12 years ago

When I worked at ClearChannel back in 2010, we rebuilt Rush Limbaugh's site. When migrating over the billing system, I realized a flaw that granted at least 20,000 people free access to the audio archive ($7.95/month). The billing provider processed the subscriptions, but their system would only sync with our authentication database once a week with a diff of accounts added or removed in the past 7 days. You got the first 7 days free for this reason. If this process failed (e.g. due to a connectivity issue, timeout, or SQL error), all accounts after the error would not be updated. Anyone with a free trial or people who cancelled during a week with an error would get a permanent free trial. I rewrote the code to handle errors and retry on failure so that errors wouldn't happen in the future, but my downfall was running a script that updated all accounts to the correct status. Imagine angry Rush Limbaugh fans used to getting something for free now getting cut off (even though it shouldn't have been free). Management quickly made the decision to give them free access anyway, so I rolled back the change.

dougbarrett• 12 years ago

I used to work at Fry's Electronics right before iPhones were released and MP3 players were seeing better days. Creative had come out with a nice, $300 MP3 player and I was in charge of creating the sign tags in my department because I was the only one who could get it done the quickest. I would do hundreds a day, and sometimes there would be slight slip ups, in this case I forgot a 0, so there was a lucky customer that got a $300 MP3 player for $30 that day.

Luckily, there was no slap on the wrist or anything, the store manager knew that after doing thousands of these cards this was only one of a few slip ups I've made so they just brushed it off and moved on.

_dp9d• 12 years ago

I added some products to a system on a Thursday, not remembering we added some new columns to the product definitions, and the columns were nullable.

I was off Friday, so I come in Monday morning to see that ~20k customers have been getting free stuff since Thursday lunchtime.

Lost something like $200k because of two nullable columns :(

munificent2• 12 years ago

Dramatically increased user satisfaction for relatively small marketing budget of $200k!

grecy2• 12 years ago

Actually, the customers get pissed off even when they get something for free, because it's just another sign of how incompetent we are :(

SDGT• 12 years ago

Ticked a debug output flag on prod for a specific IP (Proprietary CMS, couldn't replicate the issue on test even with a full codebase and db sync), brought down the entire server for an hour.

edit: This was after I asked for permission to do this.

Lesson learned: Don't EVER use Coldfusion as a web server.

itwasme2• 12 years ago

I once worked for a company that schedules advertising before films. This wasn't in the US and the company had a monopoly over all of the ads shown across the country. It was my first programming job and done during university holidays, so I was there for a couple of months and then back to university. Toward the end of the following year I get a phone call: something was wrong with the system, it was allowing agents to overbook advertising slots. I diagnosed the problem over the phone and they put a fix in but management decided it was too late for the company to go back and cancel all of the ads that were already booked. This was not surprising as it was the most money they'd ever made. Conveniently, the parent company owned the cinemas so they did a deal where they just showed all of the ads that were booked.

Because of me, one December, everyone in the country who went to the cinema got to watch anywhere between 30 and 45 minutes of ads before the main presentation started.

Lesson learned: write more tests, monitor everything.

mcintyre1994• 12 years ago

That time isn't normal where you are? How much did you make selling your system to the UK cinema industry? :)

disgruntledphd2• 12 years ago

I remember going to the cinema in the UK once in December, and being shocked by the ads. For a brief moment, I thought that I had been a victim of the tale.

itwasme2• 12 years ago

Haha. It was quite a long time ago otherwise I would have remembered the usual maximum booking time. I wouldn't be surprised if they exported the bug, given its success.

failsrails• 12 years ago

This one time we used Ruby on Fails. That was the worst screw up ever!

smoyer• 12 years ago

I didn't down-vote you, but I think the key to your mistake is "one time". You're never going to learn a framework enough to use it the first time (without an experienced mentor on board).

Those that did down-vote you are responding to the "Ruby-on-Fails" part of the sentence ... you were the failure, as you would have been with JavaEE, Django or Drupal. Your comment was needlessly offensive (to some at least) - if you're going to be critical, you should learn enough to contribute back too.

double051• 12 years ago

We shipped an Android app that didn't like the way we had our HTTPS certs configured, so I had logic in there to accept the connection if the cert matched the one we had.

Two months later, the certs were expiring soon and we changed our configuration to something Android liked by default. The bad news was that our production Android app rejected the new configuration and only wanted to accept the current certs.

We ended up quickly shipping a hotfix that accepted the current and upcoming configuration a few days before the certs expired. There technically wasn't any 'downtime' as long as users updated the app, but this all took place right before 'holiday vacations', and the QA team had to test the fix while all the devs were away.

bengarvey• 12 years ago

I poured gasoline into the tractor's radiator instead of the gas tank.

Thankfully, someone stopped me before I turned it on.

benburleson• 12 years ago

Is this really that bad? IANAMechanic, but if the coolant system is sealed, shouldn't the gas just work similar to antifreeze?

seanhandley2• 12 years ago

Not if it's diesel - it'll be heated up (so it'll partially vaporize) and it'll be under pressure. There'll be air in there as well. Could quite happily combust if it got warm enough!

NamTaf• 12 years ago

Diesel doesn't require a spark to combust, just head and pressure (and oxygen).

discardorama• 12 years ago

I bet > 66% of these are something to do with databases. :-)

My story (though I wasn't directly responsible): we were delivering our software to an obscure government agency. Based on our recommendation, they had ordered a couple of SGI boxes. I wrote the installation script, which copied stuff off the CD, etc. Being a tcsh afficianado, I decided to write it in tcsh with the shebang line

   #!/usr/local/bin/tcsh

Anyways: we send them the CD. Some dude on the other side logs in as root, mounts the CD, and tries to run "installme.csh". "command not found" comes the response. So he peeks at the script, and sees that it's a shell script. He knows enough of unix that "shell == bash". So he runs "bash installme.csh" . A few minutes go by, and lots of errors. So he reboots; now the system won't come up. The genius that he is, he decides to try the CD on the second SGI box. Same results.

In the script, the first few lines were something like:

    set HOME = "/some/location"
    /bin/rm -rf $HOME/*

Hint: IRIX didn't ship with /usr/local/bin/tcsh. And guess what's the value of "HOME" in bash?

dllthomas4• 12 years ago

'And guess what's the value of "HOME" in bash?'

In the rm line of the snippet above, "/some/location". Magic variables in bash tend to lose their magic once set.

Robin_Message• 12 years ago

I assume `set HOME = /some/location` is the tcsh syntax to set a variable.

In Bash, it doesn't do anything useful.

hapless• 12 years ago

nope. try it: set HOME; echo $HOME.

on IRIX, $HOME for root is almost certainly root. So the resulting command is rm -rf /

tilt_error• 12 years ago

  # cd /etc
  # emacs inetd.conf
  # ls
  ...
  ... inetd.conf
  ... inetd.conf~
  ...
  # rm * ~
  # ls
  # ls

yankoff• 12 years ago

Man, for some reason that double 'ls' made me laugh for 5 minutes. Just tried to imagine the surprised look on your face.

jonalmeida2• 12 years ago

irl, that would have been:

  $ ls
  $ sleep 5
  $ ls

binarygrizzly• 12 years ago

I did something similar once. Always be careful with rm and the star. Don't let him get away from its character neighbours.

0xbadcafebee• 12 years ago

"Let go" a few hours into the first day on the job.

A friend had referred me for a sysadmin job opening at a web hosting company in Florida. After a brief interview I got the job for a pretty decent salary and was told when I could start. What they hadn't told me was that my schedule would be tuesday to saturday. I had informed the hiring manager of my preferred schedule (monday-friday), but I guess nobody mentioned it to the manager of the group.

When I got there they told me my schedule and I immediately told them that's not what I signed up for. So they asked me to sit for a while so they could figure out what to do next. I took a tour of the NOC, and saw one of their tier 1 technicians was chatting and watching a movie. I walked up and asked him "Heyya! Workin' hard, or hardly workin'?" and smiled. He did not smile back. So I went back to the desk I was assigned to, which was already logged in - with the credentials of the previous admin.

While I waited I decided to see what other trouble I could get into. Sure enough, all the old passwords were saved in the old admin's browser with no master password. I couldn't copy-paste the list, so I took a screenshot and began to find a way to print the list out to post on my cube wall. Before I could finish I was asked to leave for the day while they figured out my schedule changes. I should have gotten the hint when they asked me to leave the badge there.

Later I got a voicemail telling me they'd pay me for the time I spent there (about three hours) and they'd no longer require my services. Luckily I got hired soon after to a different company, which was also hiring away all the talented people from the place that had let me go, and the web hosting company eventually went under. So it turned out to be a good thing in the end.

sejje2• 12 years ago

Which part was the screw-up?

peterwwillis• 12 years ago

Well, not making sure all the terms of the job were met, being flippant to somebody I didn't know at a new job, and digging into someone else's credentials were all pretty murky. But overall it was just getting fired on my first day vs working out the issue over time.

jawns• 12 years ago

I run Correlated.org, which is the basis for the upcoming book "Correlated: Surprising Connections Between Seemingly Unrelated Things" (July 2014, Perigee).

I had had some test tables sitting around in the database for a while and decided to clean them up. I stupidly forgot to check the status of my backups; because of an earlier error, they were not being correctly saved.

So, I had a bunch of tables with similar names:

    users_1024
    users_1025
    users_1026

I decided to delete them all in one big swoop.

Guess what got deleted along with them? The actual users table (which I've since renamed to something that does not even contain "users" in it).

So, how do you recover a users table when you've just deleted it and your backup has failed?

Well, I happened to have all of my users' email addresses stored in a separate mailing list table, but that table did not store their associated user IDs.

So I sent them all an email, prompting them to visit a password reset page.

When they visited the page, if their user ID was stored in a cookie -- and for most of them, it was -- I was able to re-associate their user ID with their email address, prompt them to select a new password, and essentially restore their account activity.

There was a small subset of users who did not have their user IDs stored in a cookie, though.

Here's how I tackled that problem:

Because the bulk of a user's activity on the site involves answering poll questions, I prompted them to select some poll questions that they had answered previously, and that they were certain they could answer again in the same way. I was then able to compare their answers to the list of previous responses and narrow down the possibilities. Once I had narrowed it down to a single user, I prompted them to answer a few more "challenge" questions from that user's history, to make sure that the match was correct. (Of course, that type of strategy would not work for a website where you have to be 100% sure, rather than, say, 98% sure, that you've matched the correct person to the account.)

ejk314• 12 years ago

Neat solution.

10feet2• 12 years ago

Ha, nice one. Whenever I start a new job, the first thing I do is create a backup of the database, because I have done something similar before. Backup everything the first day, onto your own machine.

Debugreality• 12 years ago

This one is really embarrassing. I started a new job for a small company as the only developer with the aim of creating a new site for them. So they gave me full access to their very small technology stack that included one Mssql server.

So one of the first things I wanted to do was setup a development db for which I exported the structure from their prod db. I then proceeded to change the name of the create database statement at the top to the new dev db I wanted and ran the script.

Unfortunately the prod db name was still pretended to every drop and create table command in the script so I had just replaced their whole prod db with an empty one.

Owning up to that was one of the most embarrassing moments of my career. It was such a rookie mistake I just wanted to die. Luckily they had daily backups so I only cost their 4 man business about half a day of work but... it was enough for me to be a much more careful developer from that day forward!

maxaf• 12 years ago

I spoiled business users by saying "yes" way too often.

unfunco• 12 years ago

I think everybody has done this at some point, and I'm sure I will not be the last person to have done it; leaving the WHERE clause from DELETE and UPDATE statements when writing SQL, I caused about 45 minutes downtime on our RDS instance the last time I did it, but since we had multi-AZ setup, no data was lost. I also frequently get mixed up between development and production environments.

Every database alias I have now has the MySQL --i-am-a-dummy flag appended. This has been a career-saver in my eyes.

typicat• 12 years ago

mysql> drop database PRODUCTION Do you really want to drop the 'PRODUCTION' database [y/N] y^Hn Database "PRODUCTION" dropped

tadfisher3• 12 years ago

The correct answer is always ^C

edw519• 12 years ago

Boss: We have thousands of bad orders that must be fixed now!

Me: No we don't. We have 121 bad orders.

Boss: There are thousands of them!

Me: No there aren't. There are exactly 121 of them. I'm sure.

Boss: I'm not going to argue with you!

Me: Good. Because you'd lose.

I fixed 121 orders that night. The next day my login & password wouldn't work.

msh• 12 years ago

I might be dense, but I dont get it.

opendais• 12 years ago

He was fired.

sejje2• 12 years ago

Don't call the boss out when he's wrong if it's not important.

noir_lord3• 12 years ago

He got fired for the cardinal sin, proving the idiot boss an idiot.

In theory you are supposed to be able to tell your boss he is wrong (and a good boss will appreciate this) in practice it depends on the boss.

ShaneOG• 12 years ago

I think the "screw up" was disagreeing with his boss, and thus losing his job.

FireBeyond2• 12 years ago

Tangential to this, in the middle of the dot-com boom, a downsizing. Company meeting: "Here's what's happening, layoffs, etc, etc. When you go back to your desk, if you find your login is disabled, take a moment to pack your desk and come back to the Conference Room."

That's how they "notified" people. You got to play a lucky jackpot game of logging into your desktop after the meeting to see if you still had a job. Straight out of Dilbert.

rmc• 12 years ago

When trying to put our webserver-cum-database-server onto nagios, I tried to apt-get install nagios-plugins. For some reason when installing that, apt wanted to remove mysql-server. I just pressed "Y" without thinking (because, hey, it's like 99.9999999% the right thing to do). So apt dutifuly stopped and uninstalled MySQL in the middle of the day.

Within about 2 minutes CTO strolls in asking about the flood of exception emails due to each request being unable to connect to the database.

Thankfully, I was able to apt-get install mysql-server, all the data was still there, and things were back to normal within 5 minutes.

teamcoltra_• 12 years ago

I deleted the entire sales team's sales database (for Canada's second largest cable company) because I was making a minor change and was too lazy to back it up first.

snikch• 12 years ago

Sigh, I cringe even remembering this one.

We were storing payment details sent from a PHP system into a Ruby system, I was responsible for the sending and receiving endpoints. Everything was heavily tested on the Ruby end but the PHP end was a legacy system with no testing framework. Since the details were encrypted on the Ruby end, I didn't do a full test from end to end AND unencrypt the stored results.

Turns out for two months we were storing the string '[Array]' as peoples payment details.

Takeaway: If you're doing an end to end test, make sure you go all the way to the end.

peg_leg• 12 years ago

This was some time ago. I've learned a lot since. I mkfs'd the main disk on our email server. There was no redundancy. There was a new volume that needed to be formatted, my superior told me to do it. I protested that I didn't quite know how. He pushed it. So, I would up wiping it by mistake. since then I've made it a mission to make the entire stack at that place resilient and redundant. Now, it's virtualized, failover file and DB systems, NLB web servers, redundant storage, proper backups. It would take a hell of a lot more than what I did to make the same mistake again.

FireBeyond2• 12 years ago

You could ask your boss if you can test it via re-running that fateful command ...

meifun2• 12 years ago

I once revoked my bosses e-mail and VPN access because his password was 'password123'. It was my job to keep things safe after all and I had asked him nicely a few times.

EDIT: I proposed a new password of: @$tevezA$$ignedPwD@# (Steve's Assigned Password)

He said no to that one.

noir_lord3• 12 years ago

While awesome this is not a screw-up.

The screw-up is the boss in this one.

meifun2• 12 years ago

He didn't view it that way but thankfully his boss did.

jlgaddis3• 12 years ago

Heh, I have a co-worker that recently deployed a production router with a blank administrator password. Fortunately, I've got automated jobs in place that find those kind of things. The password is now a variation of "John is a dumbass" and I'm just waiting for him to ask for the password.

jason_slack• 12 years ago

That's awesome!

cdelsolar• 12 years ago

sudo reboot

patio11• 12 years ago

I've only cried literal tears once in the last ten years, over business. Due to inattention while coding during an apartment move, I pushed a change to Appointment Reminder which was poorly considered. It didn't cause any immediate problems and passed my test suites, but the upshot is it was a time bomb that would inevitably bring down the site's queue worker processes and keep them down.

Lesson #1: Don't code when you're distracted.

Some hours later, the problem manifested. The queue workers came down, and AR (which is totally dependent on them for its core functionality) immediately stopped doing the thing customers pay me money to do. My monitoring system picked up on this and attempted to call me -- which would have worked great, except my cell phone was in a box that wasn't unpacked yet.

Lesson #2a: If you're running something mission critical, and your only way to recover from failure means you have to wake up when the phone rings, make sure that phone stays on and by you.

Later that evening I felt a feeling of vague unease about my change earlier and checked my email from my iPad. My inbox was full of furious customers who were observing, correctly, that I was 8 hours into an outage. Oh dear. I ssh'ed in from the iPad, reverted my last commit, and restarted the queue workers. Queues quickly went down to zero. Problem solved right?

Lesson #3: If at all possible, avoid having to resolve problems when exhausted/distracted. If you absolutely must do it, spend ten extra minutes to make sure you actually understand what went wrong, what your recovery plan is, and how that recovery plan will interact with what went wrong first.

AR didn't use idempotent queues (Lesson #4: Always use idempotent queues), so during the outage, every 5 minutes on a cron job every person who was supposed to be contacted that day got one reminder added to the queue. Fortuitously, AR didn't have all that many customers at the time, so only 15 or so people were affected. Less than fortuitously, those 15 folks had 10 to 100 messages queued, each. As soon as I pressed queues.restart() AR delivered all of those phone calls, text messages, and emails. At once.

Very few residential phone systems or cell phones respond in a customer-pleasing manner to 40 simultaneous telephone calls. It was a total DDOS on my customers' customers.

I got that news at 3 AM in the morning Japan time, at my new apartment, which didn't have Internet sufficient to run my laptop and development environment to see e.g. whose phones I had just blown up. Ogaki has neither Internet cafes nor taxis available at 3 AM in the morning. As a result, I had to put my laptop in a bag and walk across town, in the freezing rain, to get back to my old apartment, which still had a working Internet connection.

By the time I had completed the walk of shame I was drenched, miserable, and had magnified the likely impact that this had on customers' customers in my own mind. Then I got to my old apartment and checked email. The first one was, as you might expect, rather irate. And I just lost it. Broke down in tears. Cried for a good ten minutes. Called my father to explain what had happened, because I knew that I had to start making apology calls and wasn't sure prior to talking to him that I'd be able to do it without my voice breaking.

The end result? Lost two customers, regained one because he was impressed by my apology. The end users were mostly satisfied with my apologies. (It took me about two hours on the phone, as many of them had turned off their phones when they blew up.)

You'd need a magnifying glass to detect it ever happened, looking on any chart of interest to me. The software got modestly better after I spent a solid two weeks on improved fault tolerance and monitoring.

Lesson the last: It's just a job/business. The bad days are usually a lot less important in hindsight than they seem in the moment.

d0m• 12 years ago

Can't agree more with Patrick, mine is formulated a bit differently:

Lesson #1: Don't push code on Friday afternoon.

Lesson #2: Beer, Code and Commit is totally fine. Just don't push! Wait until next day to review and push/deploy it..

nextweek2• 12 years ago

This is the difference between a software developer and a senior software developer. I never release anything on a Friday OR the day before a holiday. ALWAYS plan it afterwards, give any technical reason you like (solar flares). Project managers will realise and thank you for it in the long run.

jwdunne• 12 years ago

Lesson 1 rings a mad bell. I distinctively remember a text from a colleague saying "yeah, I'm never launching anything on a Friday".

As a general rule, the best time to launch is first thing on a Tuesday. Why Tuesday morning? We are also well aware of "Monday morning madness" too - pre-planning stress for Monday morning is akin to masochism.

jonalmeida2• 12 years ago

I learned the first lesson the long way. To make myself feel better, my excuse is that it was my first job and I can in on the weekend to fix it.

pkinsky• 12 years ago

I work for a Forex (have USD? Want JPY?) service provider that runs a 24-5 service (At any time of day during a weekday an exchange will be open somewhere, on some continent. They all close on weekends)

We exclusively push code on Fridays.

erobbins2• 12 years ago

at my last job we had an unofficial slogan of "don't fuck with it friday"

hvass• 12 years ago

There is a great lesson in one of pg's essays. If you push you better be there for few hours to monitor what is happening. You do not just push and leave. I hope we all learn from this thread.

BU_student• 12 years ago

Fortuitously, AR didn't have all that many customers at the time, so only 15 or so people were affected. Less than fortuitously, those 15 folks had 10 to 100 messages queued, each.

Excuse me for caviling at your post, but "fortuitously" is a synonym for "accidentally", not "fortunately".

lstamour• 12 years ago

... and apparently people have been saying that about "fortuitously" since the 1920s.

kgermino• 12 years ago

Well, a quick definition says "happening by accident or chance." Well I suspect he didn't know the proper definition (I certainly didn't) it's not that bad a choice.

Thanks for the comment though - I personally always appreciate a gracious correction.

xmrsilentx• 12 years ago

I think what he meant was "It just so happened,"

zem• 12 years ago

oh man. the idempotent queue thing reminds me of the time i was working for a very-early-stage startup, and we'd managed to persuade an exec from a pretty big company to sign up for a trial. our ceo got a very angry call the next morning; the guy had woken up to 300 email messages because our queue had hiccuped and, yeah, not idempotent.

bedhead3• 12 years ago

Who packs their cell phone in a box when moving???

teacup502• 12 years ago

Who ships a product release while in no position to monitor the results?

Who ships a product release immediately after a commit?

bradbatt2• 12 years ago

Thanks for sharing—some good lessons learned in there! I love that you posted this as a comment on HN…it is worthy of it's own blog post.

zamalek• 12 years ago

I've only cried literal tears once in the last ten years, over business.
Don't code when you're distracted.

Same story here, I can't remember the exact scenario but I was concurrently acting under all three of my titles (Developer, Architect, Escalation Engineer/Critical Debugger). The customer (who was 7 hours different to us) had been battling for 3 months and we were getting nowhere (all thanks to a, since fixed, bug in WinDBG which essentially came down to broken stack traces in certain scenarios), for those 3 months I had been working 20 hour shifts (development by day, support by night).

Under that strain I eventually made a screw up with the dev and it cost QA time. The MD of my region had a sit down with me and I

cried literal tears

Needless to say they were impressed that I someone cared so much and sent me home to sleep. The next day I came in and decided to go through the 800kloc codebase line-by-line and see what could be causing the issue - I found it in a few hours.

Lesson #4: Get sleep.

ozten• 12 years ago

Many years ago I was being shown the server room for the first time. They asked me to unplug a certain box. I unplugged everything on one power strip. Panicking at the drop of ambient noise in the room, I quickly plugged it back in, but...

I have no idea why they didn't use UPS, but it took many critical servers offline and caused a few hours of headaches for everyone.

Come to think of it, that was the last time I was allowed in the server room.

Lessons learned - don't let developers in the server room.

benched• 12 years ago

I once cared about a job to the point of damaging my mental health. I haven't made that mistake since. I did, however, rather stupidly accomplish the same thing, years later, by caring too much about an entrepreneurial venture.

findjashua• 12 years ago

In my newbie days of event-driven programming, I forgot to add 'if (err) {...}' in an express application and crashed the server.

smalu• 12 years ago

chmod -R 0777 /* instead of chmod -R 0777

Reply

noir_lord3• 12 years ago

Ouch.

Reply

elf25• 12 years ago

Working Christmas Day at the Liquor store (before cameras were everywhere) and drinking Tanqueray and Mt Dew ALL day. WHEEE!!

Reply

camperman• 12 years ago

dd if=outfile of=infile

Raw unadulterated fear followed by panic.

A full reinstall.

Triple checked dd params ever since.

Reply

xutopia2• 12 years ago

To me `dd` stands for "destroy disk"

Reply

admiraltbags• 12 years ago

Lurker turned member to post this.

Second web related job at an insurance company, I was 20 years old at the time. We were heavy into online advertising, mostly banners at the time (this was right around when adwords started to get big). The company just bought out all of the MSN finance section of their site for the day-- it was a pretty big campaign ($100,000). We drove all the traffic to a landing page I had created with a short form to "Get a quote".

IT had given me permissions to push things live for quick fixes and such, I made a last minute design tweak and, you guessed it, broke something. I was checking click traffic and inbound leads and realized traffic was through the roof but leads were non-existent. This was about 45 minutes after the campaign was turned on. I jumped on the page and tested it out and got an error on submit. FUCK. I literally started to perspiration INSTANTLY.

Jumped into my form and quickly found the bug, can't recall what it was but something small and stupid, then pushed it live without telling a soul. Tested, worked, re-tested, worked. Ran some quick numbers to get a ballpark estimate on the damage I caused... several thousand.

Stood up and walked over to the two IT guys, mentioned I borked things and that I had fixed it... what should I do? I can still see the look on their faces. Shock, then smiles. Walked back to my desk and about 10 minutes later my two bosses show up (I worked for both dev & marketing managers).

They said thanks for catching the problem, not to worry. I did good for finding it myself, fixing it, and pushing it live. I was still sweating and shaking. They walk off and later that day marketing manager informs me MSN will refund us for the 45 minutes of clicks.

It took about a month before I felt competent enough to touch our forms again.

Reply

anilshanbhag• 12 years ago

This is about https://dictanote.co I changed the login flow to use a different package. After pulling the latest changes in server, I restarted apache, opened the website to notice everything working smoothly and went to sleep.

9 hours later I wake up to check my inbox has 800+ emails. Django by default sends out email when an error occurs and a tiny mistake of not installing a package led to a lot of frustrated customers and well a huge pile of email in my inbox !

Moral of the story: Put that pip freeze > requirements.txt and pip install -r requirements.txt into your deployment flow.

Reply

PakG1• 12 years ago

My first real summer job was working for a computer store that also did tech support contracts with local businesses. I'll preface that the boss should never have given me the responsibilities he gave me, or should have gotten me to job shadow more experienced people, but the shop was tiny and I was actually the only full-time employee.

We had the tech support contract for the city's Mexican consulate. One of the things we were doing was patching and updating their server and installing a tape drive backup system. Server was NT4.

I'm in there doing work after 5pm, and wrongly assume that everyone's gone home for the day. Install some patches and the server asks me if I want to reboot. I say yes. Few moments later, a guy sticks his head into the server room and asks if I'd shut down or rebooted the server. Oh, whoops, someone's here. Yeah, I just installed some patches. Oh, OK, see ya.

Next day? Turns out he had been doing some work in their database where they track and manage visa applications. That database got corrupted when I did the server reboot while he was doing his work. That night, the backup process then overwrote the previous good copy database on the tape drive with the newly corrupted database. We had not yet started rolling over multiple tapes to prevent backups of corrupt data, though we were going to purchase some tapes for that purpose shortly.

Summer was ending, and I quit a week later to return to school. Horrible timing in terms of quitting! No idea what happened after that, as I was spending the summer in a city that was not my own. I do know that the original database developer contractor was on vacation at the time and so they couldn't reach him. I think the consulate was SOL. I regret rebooting that server without checking if anyone was working to this day.

Lesson learned? Don't assume anything when doing anything. Carried that lesson with me for the rest of my life. And find a boss who knows how to guide you if you don't have much experience in your area. I guess for founding startups, at least get an advisor.

edit: spelling

Reply

onyxraven• 12 years ago

My first deploy at a once-top-10 photo hosting site as a developer was a change to how the DNS silo resolution worked.

Users were mapped into specific silos to separate out each level of the stack from CDN to storage to db. There was a bit of code executed at the beginning of each request that figured out if a request was on the proper subdomain for the resource being requested.

This was a feature that was always tricky to test, and when I joined the codebase didn't have any real automated tests at all. We were on a deploy schedule of every morning, first thing (or earlier, sometimes as early as 4am local time).

By the time the code made it out to all the servers, the ops team was calling frantically saying the power load on the strips and at the distribution point was near critical.

What happened: the code caused every user (well upwards of millions daily) to enter an infinite redirect, very quickly DoSing our servers. It took a second to realize where the problem was, but I quickly committed the fix and the issue was resolved.

Why it happened: a pretty simple string comparison was being done improperly, the fix was at most 1 line (I can't remember the exact fix). There was no automation, and testing it was difficult enough that we just didn't test it.

What I learned: If its complicated enough to not want to test using a browser, at least always build automation to test your assumptions. Or have some damn tests period. We built a procedure for testing those silos with a real browser as well.

I got a good bit of teasing for nearly burning down the datacenter on my very first code deploy, but ever since, its been assumed that if its your first deploy, you're going to break something. Its a rite of passage.

Reply

enthdegree• 12 years ago

I was this close from setting the asset management server's hostname to `ASSMAN'

Reply

jlgaddis3• 12 years ago

Heh, we have a Windows Server that all of our accounting software runs on. It's named "beancounter".

Reply

nostromo• 12 years ago

I was once in charge of running an A/B test at my work. Part of the test involved driving people to a new site using AdWords.

After the test was complete, I forgot to turn off the Adwords. (Such a silly mistake...) Nobody notices until our bill arrives from Google, and it's substantially higher than normal. When my coworker came to ask me about it, "are these your campaigns?!?" I just sank in my chair.

I think it cost the company $30k. I suppose it's not that much money in the grand scheme of things, but I felt very bad.

Reply

preinheimer• 12 years ago

I ended up as the architect for a new live show we were putting on. You could either pre-purchase some number of minutes, or pay per minute, it was like $4.99/minute or something insane.

The billing specs kept changing, as did the specs for the show itself. New price points, more plans, change the show interface, add another option here, etc. The plan had been to do a free preview show the day before to work out the kinks. That didn't happen.

The time leading up to show start was pretty tense, lots of updates, even a few last minute changes! Then the show actually started, brief relief. The chat system built in started deleting messages, one of those last minute feature changes had screwed up automatic old-message deletion. We had a fix though, update the JS, and bounce everyone out of the show and back in so the JS updates. Fixed!

Then the CEO pointed out that the quality just kept getting worse. Turns out that while the video player had both a numeric value and a string description for the different quality levels, it assumed they were in ascending order. So once it confirmed it could stream well at a given level, it automatically tried the next, which worked! Poor quality for everyone. Fixed, and another bounce.

Then it was over, time to go home. Back in the next day to finish off the billing code. I decided to approach it like a time card system. Traverse the logs in order, recording punch in time, when someone punches out, look up their punch-in times and set that user's time spent to the difference. Remove punch-in and out from the current record so they're not used again.

Now two facts from above added up to a pretty serious bug. 1) I _set_ the time spent to the difference between the two times. Not added, set. 2) We bounced everyone from the show twice to update their JS, and video player. So everyone had multiple join/parts.

I under-billed customers by tens of thousands of dollars.

Things I learned:

- Don't just argue that you need a trial run, make sure management understands the benefits. Why, not What.

- Duplicate billing code. After that a co-worker and I wrote two separate billing parsers for things, 1 designed to be different, not efficient.

- Give yourself ways to fix problems after they crop up. The bounce killed my billing code, but not doing it would have damaged the actual product (which later became a regular feature). Wish that thing had been my idea.

Reply

roryokane• 12 years ago

Your “duplicate billing code” strategy is called N-version programming (https://en.wikipedia.org/wiki/N-version_programming).

Reply

donretag• 12 years ago

A long time ago while working on a *nix box logged in as root, I executed a simple "!find". Basically execute the last find. In root's history, the last find command was something like "find ... -exec rm ...". The command was run at the root of the content directory of a CMS, deleting all the content (major media website). CMS was down while backups were restored.

I now never execute ! commands as root. Actually, nowadays I simply use CTRL-r.

Reply

BjoernKW• 12 years ago

Around 2000 my team was responsible for installing and maintaining a larger amount of servers in 19" racks in a data centre.

Most servers had those hot swap drive bays for convenient access from the front while the server was running. You only had to make sure no write operation occurred while you pulled the drive out of the bay.

So, I had to exchange a backup disk on a database server running quite a few rather large forums. The server had two disk bays: One for the live hard disk and one for the backup disk. I was absolutely sure at that time which one was the backup disk so I didn't bother to shut down the database server and incur a minimal downtime. Of course, I was wrong and blithely yanked the live disk from the drive bay.

I spent the rest of the night and most of the following day running various MySQL database table repair magic. It worked out surprisingly well but having to admit this error to our forum users was embarrassing, nonetheless.

Lesson: Appropriately label your servers and devices.

Reply

glazskunrukitis• 12 years ago

Two screw ups come to mind.

1. First day at a job. I need to get familiar with a legacy system and get a SQL dump from it to create a local copy of the database. After some SSHing and MySQLing, I confuse my two split terminal panes and end up importing my local dump to production server. Of course database names and users were the same so I end up dropping the database. No biggie. Backups were available from previous day.

2. Similar story to the first one. I got a new shiny Zend Studio IDE. Want to set up sync with remote server (just a static company website with no version control). Fill all the settings, press the sync button - and what happens? Zend Studio somehow figured that I want to force sync my local folder, which is empty, to the remote site, and it just deletes everything on the web root and uploads my empty folder. Wat. Should have read the settings twice.

Reply

codygman• 12 years ago

This could be a strong example for why "the unix way" is better. You don't overwrite your server with anything unless you explicity rsync/scp it.

Reply

drdeadringer2• 12 years ago

I dropped two units of equipment, ~$1.5Mil a piece. Each unit was dropped in separate incidents. No damage at all, but management didn't care. I blamed myself despite mitigating factors such as impossible schedules, vicious multi-tasking "to compensate", and less-than-ideal support equipment. At the time, I didn't handle it very well but I ended up living through it -- first job//assignment ever in the worst environment I've ever had before or since with the worst coworker I've ever had before or since, and I mess up in the millions of dollars. "Lasting Impressions", tonight at 8/7 Central.

I left that job about 3 years later when the metaphorical train stopped at a nicer place. My name is still known in certain circles for this ["Oh bah, how could I forget?" one former manager recently stated], but I don't plan to go back there at this time.

I learned that life's too short for assholes and working in an environment you don't like. If you don't screw up, your soul will die and you'll become that former coworker you hated so much and who hated you in return. It's worth picking and choosing where you work.

Reply

rajacombinator• 12 years ago

what kind of equipment costs 1.5Mil and can be carried/dropped by a single person?

Reply

drdeadringer2• 12 years ago

The defense kind.

Reply

gbasin• 12 years ago

Added some additional logging for an edge case, rolled it out to production and then went camping in the remote wilderness for a week. Two days in, the edge case got hit and the logging wasn't sufficiently tested. It logged as intended... and kept logging and logging... until out of disk space :(

Oh yea, I run a proprietary trading firm (still at the same spot), as a result of that bug we went down and lost about $250k over the next few hours. Testing is important in automated trading :)

Reply

hcarvalhoalves2• 12 years ago

Happened to a colleague: it was the end of the day, and we were packing up to leave. He used Ubuntu on his notebook, so he typed "shutdown -h now" on his shell prior to closing the lid. Seconds later he's groaning, having noticed it was a SSH session to the production server...

It wouldn't be a big deal, wasn't for the fact it was an EC2 instance, and back then halting the instance was equivalent to deleting it permanently. We then spent the night at the office recovering and testing the server. I think we left 3:00 AM that day.

Lesson #1: it's never a good idea to "shutdown -h now" on a shell. any shell.

Lesson #2: have the process to spin up a new production server fully automated and tested

Reply

brey• 12 years ago

molly-guard is an good way to prevent this - forces you to type in the hostname of the machine you are trying to halt / restart etc.

http://manpages.ubuntu.com/manpages/trusty/man8/molly-guard....

Reply

scott_karana2• 12 years ago

Excellent if you have complete and thorough control of every server you touch, but if you don't, it could be dangerous to rely on. Murphy's Law means it'll be that one dang machine that gets shut down...

Personally, I'd think that training this is a lot more portable.

  # hostname
  <foo>
  # shutdown -h now

Reply

jbb555• 12 years ago

I once did a "shutdown -h now" on a remote server on a customer site with nobody there who knew anything about the server when I'd meant to do a reboot....

Reply

riquito• 12 years ago

Last day of work before moving to the new job: I do some cleanup and rm -fr my home directory. Seconds passed. Minutes passed. I start to think about how can it take so long.

I list the content of my home directory trying to understand which folder was so big. Then I see it. A folder usually empty. Empty because I use it as generic mount point. A mount point that the day before was attached via sshfs to the production server...

I had a strange feeling, like if I was seeing myself from behind, something crumbling inside me. And at that moment someone start to ask "what's happened to <hostname>"?

I take my courage and I say "I know it"...

That was really hard. The worst day at work in years, and during the last day too. Luckily we had a good enough backup strategy and the damage was mostly solved in a couple hours.

There I realized how much of an idiot I was to have mounted the production server on my home and I grow a little.

Reply

Beltiras2• 12 years ago

I work at a newspaper as a programmer for the website. Mostly my job is backend programming, some HTML and CSS work (mostly left to designers). I run our local computer infrastructure as well as manage a cluster for our online presence and assist in technology related journalism as well as assisting our CEO in managing the IT budget.

I inherited a mess of an architecture and am finally getting around to rewriting our deployment process. We buy VM services from a local outfit and the prices are basically an arm and a leg for rather small machines. Due to this my predecessor put in place an insane deployment script. It pulls the new version from github then reloads code on the running dynos, one after another. Reverting is out of the question with our current approach to VCS (something I am also fixing). Most of the time this is no problem, all we are changing really is some template code, or introducing new models and their views.

Thinking back I am quite happy we don't run into more problems than we do, but also happy that this type of insanity is soon in the rearview mirror.

The worst mistake was recently, cost us about 4 hours of downtime during the busiest time of the day.

A big feature on all news sites are lists of stories to present to the user to look at after they have read what you put in front of them at the moment. They may take the form of most viral, most read, most commented, sliced by time or category or many other factors. My predecessor had written all those lists statically, which made maintenance a nightmare and extension very fragile.

I made a function that was a generic list of items. You supply basic parameters, amongst them a QuerySet for what would construct the list and my function would check to see if it was cached and if it wasn't, generate it and cache it.

The framework I use (Django) generally uses lazy evaluation for all QuerySets and I rarely have to think about the size of the list I generate, I just take care to limit the query before I list() it. During development nothing showed up as a problem and I deployed this and all seemed to be good with the world.

A week passes by where I made at least 2 minor deploys (small changes to templates, minor tweaks to list filters) and all seemed to be good with the world.

Designer sends me a pull request, I look over the code, just some garden-variety template changes, nothing that should raise an eyebrow. Make the merge, plan to deploy and then go to lunch. Deployment done, all seems well for 2 minutes but then suddenly servers lit on fire. Pages spewed 404's and 500's like there was no tomorrow.

For 4 hours I tear my hair out, examine every piece of code I was deploying that day, call in the big gun support (the kind that costs more money than I care to think about). Everything I was looking at pointed to the caching agent not working. Too many pageviews requesting the database, too much load on the servers, reboots made them work fine for about a minute but then everything became bogged down.

The big gun support pointed something out finally that I had missed: Traffic from the database to the dynos was abnormally high. Made me take a look at code that had been there for a while and lo and behold: For some reason when you pass a QuerySet as a parameter, it seems to be evaluated for the receiving function! 2 lines of code added, one deploy, problem fixed.

I have no idea to this day how this code could be live for a week without causing problems but an unrelated change triggers the bad behavior. This is not be the first time I've seen strange behavior from code, having seen a Heisenbug in Java code.

There's a happy ending to this. I made a big mea culpa slideshow where I pointed out all the flaws and what we needed to do to prevent a re-occurence. I got support to make the changes needed and my new cluster goes live day after tomorrow. Now I can carefully change NEW dynos for a deployment, keeping the old one's around if the shit hits the fan. I got some changes instituted in how we approach VC, something that's hampered work for a while. And we save money in the long run because we will no longer be paying an arm and a leg for the VMs (AND I got to learn about clustering machines with HA, goodstuff with gravy).

Reply

jlgaddis3• 12 years ago

... and my new cluster goes live day after tomorrow.

On a Friday? :-)

Reply

Beltiras2• 12 years ago

End-of-month.

Reply

leothekim• 12 years ago

Not the worst, but certainly most infamous thing I've done: I was testing a condition in a frontend template which, if met, left a  comment in the header HTML of all the sites we served. Unfortunately the condition was always met and I pushed the change without thinking. This was back in the day when bandwidth was precious and extraneous HTML was seriously frowned upon. We didn't realize it was in production for a week, at which point several engineers actually decided to leave it in as a joke. Then someone higher up found out and browbeat me into removing it, citing bandwidth and disk space costs.

Now, if you go to a CNET site and view source, there's a  comment. I like to think of that as an homage to my original fuckup.

Reply

strozykowski• 12 years ago

Haha! I had to go out and check, and sure enough, there it is in CNET.com's source.

Reply

adw• 12 years ago

I've screwed up countless things, many much more expensive than this, but those stories aren't entirely mine to tell.

But this was one of my first. Years ago, making boot floppies for a physics lab where I was reinstalling all the servers:

I meant: dd if=/dev/zero of=/dev/fd0

I did: dd if=/dev/zero of=/def/hda

Oops. Bye, partition table.

(Always double-check everything you type as root.)

Reply

cjensen• 12 years ago

I meant to dd a partition to tape... and ended up doing the reverse on our fileserver. Tape is slow, so I caught it after a few kilobytes.

Somehow I managed to salvage the situation using the man pages for the filesystem, the C compiler, and making sure I did not reboot. Really don't remember much about that late night...

Reply

slowmover• 12 years ago

system("tar --delete-files -czf archive.tar.gz $datadir/");

What could possibly go wrong?

Reply

slilo• 12 years ago

$datadir could be emtpy

Reply

Ecio78• 12 years ago

I can't decide between these two:

1) after few months working in a bank, I was doing some simple admin check task via RDP to a Windows 2003 (no, maybe 2000) server, when I right-clicked the network icon and instead of clicking the properties options i clicked "disable". Just the time to say "oh sh!t" and to realise that it was the production Trading On Line machine, on a remote datacenter, during market hours, and to discover couple of minutes later that the KVM over IP was crappy and was not working. We had to call the datacenter operators to go back to the local KVM and re-enable the NIC.

Lesson 1: Better move slowly when you're on a production machine (and also have plan B and C to reach your machines is a good idea)

2) same bank, one or two years later, I was doing some testing on a new mail system that integrated also VoIP (SIP). Mail/SIP System running in a VM (I think Vmware Server at that time) in the same remote datacenter as above. So, I enable the SIP feature and after few seconds, bum, we lose the whole (production) datacenter and the connection between the local server room and the datacenter. Panic, I look at my colleague, WTF in stereo, everything come back for few sec, bum again down. Long story short, the issue was that that version of Netscreen firewall ScreenOS had a buggy ALG implementation for SIP that lead to core dumps. The fun thing is that we had two of those in HA, same version of course, so they were bouncing between core dumping, rebooting slave becoming master and then core dumping again etc.. We had to ask a datacenter operator to reach the rack, disconnect one of the cables from the firewall (the one that was managing the traffic of the DMZ where that machine was hosted) and then reach the virtual host to kill the machine.

Lesson 2: you can segment your network but if everything is connected through the same device(s), sh!t can still hit the fan...

Reply

wpietri2• 12 years ago

Long ago when I was, I think, a sophomore in college and worked for the university IT group, I was trying to add an external drive to an early NeXT machine [1]. I wanted to try out their fancy GUI development stuff, you see. I was at best a modestly competent Unix admin, and this was circa NextStep 1.0, so the OS was... rough. It was in the dark days of SCSI terminators, so just telling if the drive was properly connected and, if so, how to address it was challenging.

After a couple hours of swearing, instead of working from a root shell in my own account, I just logged into the GUI as root. And there was a pretty interface showing the disks. I could just click on one and format it. Hooray!

Well either the GUI was buggy or I clicked on the wrong disk, because as the format was going, I realized the external drive wasn't doing anything. I was formatting the internal boot hard drive. And since nobody but me gave a crap about this weird free box somebody had given them, they had repurposed it. As a file server. For the home directories of a bunch of my colleagues. Who were now collecting around me wondering what was going on. Oops.

No problem, says I. I'll just restore from backups. But this thing used a weird magneto-optical drive [2]. The only boot media we had was on an MO disk. The backups were on another. And there was only one of these drives, probably only one in the whole state. The drives were, of course, incredibly slow, especially if you needed to swap disks. Which, I eventually discovered, I would have to do about a million times to have a hope of recovery.

Long story short, I spent 28 hours in a row in that chair. It was my immersion baptism [2] in the ways of being a sysadmin. The things I learned:

Fear the root shell. It should be treated with as much caution as a live snake.

Have backups. People will do dumb things; be ready.

A backup plan where you have never tried restoring anything may lead to more excitement than you want.

Be suspicious of GUI admin tools. Avoid new GUI admin tools if at all possible. Let somebody else be the one to discover the dangerous flaws.

If you were smart enough to break something, you're smart enough to fix it. Don't give up.

When some young idiot fucks up, check to make sure that they are sufficiently freaked out. If they are, no need to yell at them. Instead support them in solving the problem.

Seriously, my colleagues were awesome about this. I went on to become an actual paid sysadmin, and spent many years enjoying the work. The experience taught me fear, and a level of care that sticks with me today. I'm sure at the time I was wishing somebody would wave a magic wand and make it the problems go away, but working through it gave me a level of comfort in apparent disasters that has been helpful many times since.

[1] http://en.wikipedia.org/wiki/NeXTcube [2] http://en.wikipedia.org/wiki/Magneto-optical_drive [3] http://en.wikipedia.org/wiki/Immersion_baptism

Reply

georgemcbay• 12 years ago

root shell, plus rm with any sort of wildcard matching, plus a bit too much of a delay before you get your shell prompt back results in a very specific kind of panicked anxiety that almost anyone who has been programming or sysadmining for a while can easily relate to.

Reply

jrabone• 12 years ago

I'm sure zfs on Linux has a random sleep built in just to increase my anxiety levels...

Reply

sj26• 12 years ago

alias rm="rm -v" helps.

Reply

wicker• 12 years ago

"When some young idiot fucks up, check to make sure that they are sufficiently freaked out. If they are, no need to yell at them. Instead support them in solving the problem."

My first job was as a Marine avionics technician and I had the misfortune to work for bosses who reassured themselves by blaming you at full volume and making you feel like shit. Usually with the 'you could have killed somebody you useless idiot' which is actually not that often true. Not a great environment.

Last week I'm just starting my second year at this company as a real proper electrical engineer and I neglected to run all the steps of the pre-tests I said I ran and I therefore missed a (in retrospect, really cool) bug that showed up in front of the certifying examiner on a Friday afternoon. Said examiner costs a lot of money per hour, it's a government agency, etc. etc. I spent a weekend and Monday finding the bug and squashing it. It took til Friday before the new code was released and all the paperwork up to the certifying agency was fixed.

The whole time my boss was there and clearly frustrated but also very clearly keeping quiet. I definitely went and thanked him on Friday afternoon when it was all over for being so chill about it. He said it was no problem as he could see I had sufficiently embarrassed myself and was working hard to fix it. It's so true. I was keeping my head down, being calm but intent on the problem, and I don't know how I could've solved anything if he'd lost his shit.

Fear and care are definitely with me now. As is proper testing and better logging of my own day-to-day dev process. (Note to self: power cycle in the test when it tells you to power cycle. You're also testing the non-obvious functionality.)

Reply

highace• 12 years ago

Changed the default RDP port on a remote Windows box, but didn't open the port on the firewall and couldn't get back in. Whoopsie.

Reply

a3n• 12 years ago

Connected leads on an expensive piece of equipment, power live, being very careful with a pair of needle nose pliers. Because the power switch was way in the other room, tag out procedure took time and I was late.

Poof. Equipment electronics fried and useless.

I was chewed out. Could have been way worse.

Follow your safety procedures.

Reply

derwiki• 12 years ago

I accidentally brought down yelp.fr by typoing the timezone field in the database.

Reply

seanhandley2• 12 years ago

Wrote an article on the company blog and linked it on HN. Traffic brought down the server -_-

Reply

jhonnycano2• 12 years ago

that shouldn't be a screw up, but a fire's baptism for your company's infrastructure

Reply

arethuza• 12 years ago

I led an engineering team that almost sent out a demo on tens of thousands of IBM CDs (this was 1998) that contained test data that included some that had been sourced from the worst possible alt.* newgroup.

As it turned out the only data that did go out was the single word "sheep" in the search index.

Reply

CurtMonash• 12 years ago

Short version:

I was a stock analyst, for a firm with dozens of institutional salesmen and thousands of retail brokers. Some of my recommendations were very, very wrong.

The right thing to do is stand up, take the heat, and explain what you now know as best you can. I learned that watching a colleague who I thought was otherwise an unserious ass.

Reply

Smerity• 12 years ago

I was testing disaster recovery for the database cluster I was managing. Spun up new instances on AWS, pulled down production data, created various disasters, tested recovery.

Surprisingly it all seemed to work well. These disaster recovery steps weren't heavily tested before. Brilliant! I went to shut down the AWS instances. Kill DB group. Wait. Wait... The DB group? Wasn't it DB-test group...

I'd just killed all the production databases. And the streaming replicas. And... everything... All at the busiest time of day for our site.

Panic arose in my chest. Eyes glazed over. It's one thing to test disaster recovery when it doesn't matter, but when it suddenly does matter... I turned to the disaster recovery code I'd just been testing. I was reasonably sure it all worked... Reasonably...

Less than five minutes later, I'd spun up a brand new database cluster. The only loss was a minute or two of user transactions, which for our site wasn't too problematic.

My friends joked later that at least we now knew for sure that disaster recovery worked in production...

Lesson: When testing disaster recovery, ensure you're not actually creating a disaster in production.

Reply

grecy2• 12 years ago

I think you just started a new Agile-like trend:

'hyper-committed disaster recovery testing'

Reply

_mikelcelestial• 12 years ago

I accidentally deleted all data from live database, thought it was our beta database server. Good thing it is synchronized on our beta servers so was able to bring it back in no time. The moment I clicked that delete button I was like face palming myself all over. I learned from thereon to double check every time especially when working on between production and test servers.

Reply

jonathanjaeger• 12 years ago

I often turn on my performance-based ad campaigns before going into the office, as they are very predictable at the beginning (slow ramp up to spend). However this time the CTR was through the roof for something new and spent $15,000 by the time I could turn them off when I got to the office. It only brought in about $5,000 in revenue. Not the end of the world in the grand scheme of the monthly P&L, but still not something to replicate.

Reply

joncooper• 12 years ago

There's a saying in the rates market: "don't counter-trend trade the front end".

I lost $7 million dollars in minutes by being short $700 million of US 2yr notes when the levees failed during the hurricane Katrina disaster.

Although my bet that the 2y point would be under pressure in the intermediate term turned out to be true, I got carried out by fund flows as folks spazzed out to cut risk by rolling into short duration high quality paper.

To his credit, my boss, who sat across from me, said only: "wouldn't want to be short 2 years." He let me make the call, which I did, and I covered my position. (Ouch.)

My book was up considerably on the year already, but this was a huge hit, and nearing year-end. I dialed back the risk of my portfolio and traded mostly convex instruments (options) for the remainder of the year.

Reply

andy_thorburn• 12 years ago

My worst screw up was causing a fire that destroyed one of the two prototype 3D printers my company had built.

I was working at a startup that was trying to create an affordable 3D printer. We had two working prototypes that were used for everything - demos, print testing, software testing, PR shoots, everything. Each prototype had cost hundreds of man hours to build and debug and quite a bit of cash as well.

Among other things I had done all the work on the thermal control system for the printer, it kept the print heads and build chamber at the correct temperature. One night while working on one of the printers I hit an edge case that my control code didn't handle well and the printer turned all of the heaters on full-bore. Half an hour all the plastics in the prototype had either melted or burned and I was left with a room full of smoke and a pile of scrap aluminum.

Reply

schmichael2• 12 years ago

I unknowingly reset serial number counters in a bicycle part's database, so now there are a few hundred people in the world with high end bike hubs that overlap each other.

Lesson: Keep the code that touches production databases as simple as possible so it's easy to verify exactly what it does. I was using a framework's database tooling incorrectly because I never dreamed what I used would touch the databases's counters.

(Not my worst mistake in terms of people affected, but it's the only mistake that was literally laser etched in metal forever.)

Reply

tadfisher3• 12 years ago

Chris King?

Reply

schmichael2• 12 years ago

I plead the fifth.

Reply

tsaoutourpants2• 12 years ago

Back in my younger days, I once had a project manager who was asking me to make a significant network infrastructure change but refused to tell me why the change was necessary and basically told me to do as I was told. I messaged a coworker to see if he knew what was going on, and dropped in that the PM was being a "fucking cunt." I was unaware, however, that the co-worker and the PM were troubleshooting an issue together and the PM was staring at his screen as my message came through.

The PM brought the issue to the CTO, but somehow I didn't get fired. Ended up apologizing (obviously a poor choice of words :)) and moved on. Never made that infrastructure change.

Key takeaway: if you're going to talk shit, don't do so in writing. ;)

Reply

tsaoutourpants2• 12 years ago

Also, I had a friend with a similar (perhaps worse) story. His company sent every employee an e-mail about being on time, to which he pushed Reply to All and typed "FUCK YOU!" He laughed to himself and went to push the Discard button, but accidentally hit Send in the process. He was about to try to ExMerge it out of Exchange when he heard a BlackBerry vibrate and realized that no amount of ExMerge would get it off the BBs. He spent the next hour going door-to-door apologizing, and also managed to not get fired.

Reply

lostlogin2• 12 years ago

Damn iMessage is a shocker for this. Coworker sends out group message saying big boss wants xyz and includes big boss in the group. Someone always misses that its a group message and sends an expletive or sarcastic reply. I am also guilty.

Reply

m3mnoch• 12 years ago

way back in the late 90s stone age of interactive ad agencies, we were doing our first really big gig for hp. it was a demo shipping out to retail stores showcasing one of their products -- a run of 30,000 stamped cd roms.

i was the one developing the macromedia director app running on the cd.

we were on-time.

we were ready to send them out the door.

it was awesome.

and then we tested the rom outside of our network...

in some far-off corner of code, i had baked in a hard reference to one of our file servers on our network for some streaming assets. the cd failed as soon as you put it in the drive due to that reference to the missing file.

by the time we discovered this, we'd already glass-mastered and stamped 30,000 discs to the tune of $40k or so. or, about $6k per employee. in a company that booked about $50k the previous year. where i worked for free for 9 months.

so, my line of code cost our little company the equivalent of almost all of our previous year's revenue -- not profit, but revenue.

we, of course, had to make the run again -- only this time at the emergency rush prices. and this time, we were running late.

we managed to book some time in the middle of the night at the stamping plant. it was 4am. i had a courier standing over my shoulder watching me run the final build again, this time without the dreaded line of code -- which broke other things i had to fix when i removed it -- before he could take it.

i finished testing. ejected the disc. handed it to the courier, who started running as he was placing it into its case. he drove like hell to make it to the airport where we counter-to-countered it on a 2-hour, 6am flight to vegas for stamping.

oh, and it almost got even worse from there. almost.

we didn't know if they would be able to stuff the cds into the packaging because this was an emergency run and they didn't have the people available.

so...

we were actually on our way to rent a uhaul which we calculated we could drive to vegas just in time for the stamping run to finish. from there, we would load the discs on their spindles, and 4 of us were going to sit in the back of the van, stuffing 30,000 discs while we drove the uhaul to palo alto. from vegas. yes, stuffing discs in the back of a traveling uhaul.

we even had the patio furniture from one of the employees yards already picked out to sit in while we were in the back of the truck.

luckily, the plant managed to squeeze in our packaging (at rush pricing, of course) and all we needed to do was have one of our guys take them as luggage on a later flight that day to the bay area instead.

as to a couple, big lessons learned?

1) i can honestly tell you, i've never, ever had a hard-coded, local network link in anything i've shipped since and never will again. always test off-network. especially these days with mobile apps and their on-off-network states.

2) a strong, non-finger-pointing team is where you need to be. i felt appropriately awful, but we handled it as a team and proceeded to grow that little company to about $40 million a year before a merger.

p.s. oh, and next time, remind me to tell you about the time i ran a database query on production that nuked the entire website for the publicly-traded software company which relied on -- wait for it -- the website to do all its commerce.

Reply

hrrsn• 12 years ago

There is no way you can leave a juicy line like that at the end and not expect me to ask for the story!

Reply

codfrantic• 12 years ago

Hey m3mnoch, you asked me to remind you to tell us about the time you ran a database query on production that nuked the entire website for the publicly-traded software company which relied on -- wait for it -- the website to do all its commerce!

Reply

jjindev• 12 years ago

Once, as a relative UNIX newbie, I "cleaned up" a Sun box, until I had moved things I needed for boot off the boot partition. I got it all back, manually mounting partitions, and etc. but I was certainly in a cold sweat for about 15 minutes.

Perhaps the only lesson is "slow down."

Reply

famousactress• 12 years ago

Accidental sudo chown www-data:www-data /. on the production server.

Thoughtful pause "Why is this taking so long!?"

"OH FUCK"

Reply

it200219• 12 years ago

I had installed "osCommerce - Open source E-Commerce platform" just like Magento on one of our client who had > 500 transactions a day.

Some how in settings, we had flag "Store Credit Card Info" as "Plain Text" enabled. The Admin/Staff of that client could have use this information to make transactions (As in Backend it would show Full CC info into order details)

We didnt realized untill we worked on it again for some bug fixes and adding new features.

Lesson Learned :- When transitioning from DEV to PROD env, make sure to check all these critical flags and correctly set

Luckily, the client didnt had any idea about what was wrong in backend.

Reply

slipangel• 12 years ago

sudo chown -R myname:myname /

Learned: Learning on the job as you hack away on problems is great, but recognize that it's one part enthusiasm and one part risk management. Also learned to never try anything on the command line that wouldn't want to see pulled from my bash history and stuck on the breakroom fridge. Also learned to cope with humiliation well.

Reply

laxk• 12 years ago

Put space in front of any command to avoid history.

Reply

quackerhacker• 12 years ago

I messed up epically on an interview. It was a 3 part interview for a JS/RoR coder.

1. I passed the resume and chat portion

2. I passed the telephone questionnaire and got along great with the interviewer

3. (Fail) I scheduled my interview on a Friday at 4:30pm and there is a 30 min travel time. I left 1hr early...still it was Memorial Day weekend, so I thought the streets would be quicker than the freeway since it was at a stand still. I was so stressed that I literally had an anxiety attack and couldn't even find the address. Never happened to me before, so I'll never forget it.

Reply

Zikes• 12 years ago

I think most interviewers would be understanding if you called to reschedule in that situation.

Reply

alexmarcy2• 12 years ago

My worst would have been catastrophic if I had waited one minute to make my mistake.

I was commissioning a new control system at a power plant's water treatment facility. I was fairly new to the industry and had mostly looked over the guy who did the bulk of the work's shoulder as on the job training.

This particular day the guy was out sick and we had to finalize a couple of things before we ran through the final tests.

There was an instruction to open a valve to fill a tank and it had the wrong variable linked to it. The problem was to maintain the naming standards I had to do a download to the processor to make the change. When I had been doing work in the office this was not a big deal, download the program to the processor, it stops running for a moment while it loads the new logic into memory and starts back up.

Not thinking through the implications of the processor shutting down while the process was up and running I made the code changes, hit download and about 30 seconds later an operator came running over looking like he had seen a ghost and he was pissed.

While I was making my code changes the operator was hooking up a hose to drain a rail car of some chemicals. The way the valves were configured before I made my changes was correct and would have had no consequence it I didn't touch anything. The way the valves were configured when the processor restarted would have routed the rail car's contents to the wrong tank resulting in a reaction which would have created a huge plume of highly toxic gas. The way the wind was blowing this plume would have blown directly to the largest town in the area and could have killed a ton of people.

The operator heard the valves in question changing position before he opened the valve on his hose to empty the rail car and figured something was up. When he saw the whole process had shut down he got really angry because I had ignored the protocol in place to avoid such a disaster.

I got chewed out and kicked off the site. My boss attributed my mistake to inexperience and I had to give a safety presentation on what I did wrong.

Lessons learned: Be sure you are aware of any implications your actions have. If you are unsure or guessing about something stop what you are doing and go ask someone first.

Don't give people mission critical work on their first project and have them work unsupervised. Training is important.

Always be aware of safety requirements, especially when you are working with machinery, automated processes, chemicals or anything else that can hurt, maim or kill you.

Reply

micro-ram• 12 years ago

Rule #1 - Don't make things worse by guessing.

Reply

chrislomax• 12 years ago

Simple one really and probably most common. Realise a data integrity issue on DB, try to load from backups and notice that the backups have the same integrity issue. Find a backup from about 2 weeks previous where data is intact and piece together the good pages from the daily backups and the 2 weeks old one.

All in, took 4 days and a new server where the hard drive had stored bad pages on the DB. We lost 2 days of orders (they were processed through to the internal systems though so not really lost)

Lesson learned, validate backups and check page integrity when backing up

Reply

jamesbrownuhh• 12 years ago

Demonstrated SQL injection to a colleague on the live website. Bringing a sample URL up into the address bar, I explain, "You see, that ASP script takes the value of ?urlparameter and updates the record - but what if I modify urlparameter so that instead of 1, it is... (types) semicolon dash dash DROP TABLE usermaster (presses enter)"

"Shit. Well, as I have just demonstrated, it becomes possible to wipe out a million user login credentials at the touch of a button. So now we'll be needing to restore that from the backups which we don't have." Luckily, and ONLY BY CHANCE, I happened to have a copy of that table exported for other reasons from a few days back.

Lessons learned: Never press enter.

Reply

Phlarp3• 12 years ago

The problem was the service being susceptible to injection in the first place.

This wasn't a mistake; just a hilariously successful penetration test!

Reply

alexmarcy2• 12 years ago

The worst one I ever heard about was while I was at a potato processing plant in Idaho where they make McDonald's hashbrowns.

After the potatoes are peeled and washed they are run through a pipe with blades to slice the potatoes into french fries. These blades are sharpened with lasers and are insanely sharp because they need to cut a lot of potatoes before being changed.

One day they were shutdown and it was time to change the blades. The lady doing the change placed the new blades on the table and bumped the table when she turned to grab a wrench from her toolbox. The new blades started to fall and she instinctively reached out to grab them to prevent them from falling to the floor.

She ended up not grabbing anything because the blades sliced her fingers clean off. They took her to the hospital and due to the blades extreme sharpness, the cut was so clean reattachment was a pretty easy procedure. I don't know if she had any long-term negative effects from the incident.

Safety is important, be aware of your surroundings and don't instinctively grab things you shouldn't be touching in the first place.

Reply

jboggan• 12 years ago

I love these topics.

~ 2007, working in a large bioinformatics group with our own very powerful cluster, mainly used for protein folding. Example job: fold every protein from a predicted coding region in a given genome. I was mostly doing graph analysis on metabolic and genetic networks though, and writing everything in Perl.

I had a research deadline coming up in a month, but I was also about to go on a hunting trip and be incommunicado for two weeks. I had to kick off a large job (about 75,000 total tasks) but I figured spread over our 8,000 node cluster it would be okay (GPFS storage, set up for us by IBM). I kicked off the jobs as I walked out the door for the woods.

Except I had been doing all my testing of those jobs locally, and my Perl environment was configured slightly differently on the cluster, so while I was running through billions of iterations on each node I was writing the same warning to STDOUT, over and over. It filled up the disks everywhere and caused an epic I/O traffic jam that crashed every single long-running protein folding job. The disk space issues caused some interesting edge cases and it was basically a few days before the cluster would function properly and not lose data or crash jobs. The best part was that I was totally unreachable and thus no one could vent their ire, causing me to return happy and well-rested to an overworked office brimming with fermented ill-will. And I didn't get my own calculations done either, causing me to miss a deadline.

Lessons learned:

1) PRODUCTION != DEVELOPMENT ever ever ever ever 2) Big jobs should be proceeded by small but qualitatively identical test jobs 3) Don't launch any multi-day builds on a Friday 4) Know what your resource consumption will mean for your colleagues in the best and worst cases 5) Make sure any bad code you've written has been aired out before you go on vacation 6) Don't use Perl when what you really needed was Hadoop

Reply

lambdaphage• 12 years ago

Nice. I once needed to do reciprocal blast for the complete genomes of about 300 bacterial species. That's on the order of half a billion queries, but the work was embarrassingly parallel, and each discrete job only took about 90 seconds. I wrote a little shell script to kick them off on the cluster, and went home.

I woke up the next morning to several inbox screens' worth of messages from angry people I didn't know, demanding explanations for what I did to their jobs and their cluster. I don't think I have ever biked to the lab faster.

After multiple rounds of palm-drenching emails with the cluster sysadmins and the computational mathematics group PI (and my own boss agonizingly cc'ed), we determined the cause. The cluster sysadmins, lacking imagination for the destructive naivete of their users, had not foreseen that anyone would want to submit more than 10^4 jobs at once. That broke the scheduler, preventing other people from running jobs and me from canceling them. Meanwhile the blast jobs blew past the disk quota, leading to a Hellerian impasse where I somehow lacked the space to delete files so I could create space. I still don't fully understand it.

I believe it took a full day to get the cluster back online.

Reply

gadders• 12 years ago

My team did something similar once. We pushed a version to UAT with all the dev debug logging still turned on. It filled up a solaris disk so badly the SA had to get a bus to the data centre and fix it in person.

Reply

vacri• 12 years ago

Still feeling my way around in the new job, I was fiddling with a backup script, got distracted, turned back, and dropped the production database. Two minutes later "Hey, is the website down"? Then I look at the prompt...

I run around like a headless chicken trying to find who knows the right backup to use and so forth, and I can't figure out why everyone is so calm and collected about it. Production was down/shit, I hope I still have a job. Turns out we had no active clients at the time - no-one was accessing the site. We'd finished one run and were in 'dead time' before the next. My next project involved implementing coloured prompts and I no longer leave production ssh sessions lying around when I've finished with them.

My CTO still has me listed as "database [vacri]" in his phone...

Reply

aryastark• 12 years ago

This wasn't me, but a coworker.

We were rearranging the layout of the office. Coworker was moving in to his new space, setting up his desk. He boots up his computer, wonders why he has no network. Looks around, discovers the ethernet cable isn't plugged in. Plugs it in to the wall, still has no network.

A few minutes pass, and the entire office is running around wondering why the hell the network isn't working. Maybe an hour passes, the network guys are losing their shit trying to hunt down what is wrong. I'll give you a hint: the router was lit up like a Christmas tree, and the aforementioned coworker had both ends of his ethernet cable plugged in--but neither end was attached to his computer.

Reply

trustfundbaby• 12 years ago

rm -rf .

yup. that really happened. it was 4-5am in the morning and I'd been working all night. I was on the server trying to set something up and was trying to blow away a folder ... I did a normal rm and that didn't work (obviously) because there was crap in the folder. So I pulled out my nuclear weapon to nuke the folder but left off the preceding ./ (which still wasn't that smart anyway) ... I sat there for a second wondering why the deletion was taking so long ... then another 30 then a minute ... then I looked at what I'd just typed again ... then I realized what had happened.

ctrl-c'ed (or d, can't remember now) out of it. then tried to find root folders

cd /etc => folder not found

cd /var => folder not found

I'm from a third world country where we laugh at Americans (sorry) for throwing up when they're nervous or having panic attacks, but at that moment, I had a full blown panic attack. I'll never forget it.

The work was a subcontract for a client who was doing work for Nike, and it was a decently sized project that was critical to the success of the firm, and I'd just blown away their live production server ...

Afer freaking out and almost crying for 5 minutes. I decided to call media temple support (we were using one of their vps servers) ... and by the biggest absolute stroke of luck they'd just backed up the entire server ... not even 2 hours prior to my madness. $100 for a full restore (I don't recall why) and would I like to do that?

HECK YES I WOULD!

so they restored the server for me. I wrote an email to the head of the small company I was doing all the work for, explaining what I had happened and telling him I'd sent over a check for $100 to cover the backup because it was my fault. He was obviously very relieved and never cashed the check I sent.

I still get chills thinking about that exact moment when I thought I'd fucked up my career and reputation for good.

Reply

alok-g2• 12 years ago

The following was not actually me, but worth sharing.

They had ASIC design runs for research purposes once every three months, yielding your design on Silicon as ten 6" wafers. It gives enough parts for testing the first revision of your design. The person was carrying the wafers to a vendor for cutting into separate ICs and packaging or something. Gets to the parking lot, and where are the keys. Puts the wafers on the top of the car, finds the keys in his pockets and starts driving. Boom, the box of wafers was still on the top of the car, now on the ground. All broken. Some $100K in wafers + three months lost + bad face before the customer + ... Lesson: Don't put stuff on the top of the car!

Reply

10feet2• 12 years ago

I don't think that is the lesson, I think the lesson is that it is clearly a two person job. One person to carry the wafers, the other person to remove and hazards/ open doors/ double check everything.

Reply

alok-g2• 12 years ago

It's easy to say that after the fact. There are many delicate tasks people handle as a part of their day-to-day jobs, and not every task can afford more people to help without increasing the costs.

Reply

rosser• 12 years ago

An UPDATE statement without a WHERE clause.

In production.

I'm the DBA.

Reply

taf2• 12 years ago

I hacked our development machines using a rooted rpm, we only had access to the sudo rpm command so I decided to deploy our rails app using capistrano. to work around the sudo rpm only access I decided to add some install scripts to the rpm because these run as root. This allowed me to re-configure sshd making it possible to do a local capistrano deploy. I was smart about it by reverting the ssh changes back after the deploy completed - bash has a kind of ensure that allows you to roll things back like a transaction. The cool thing about the whole thing was that our ops team was on the ball and detected the changes to the sshd configuration even though I restored them. Mind you this was all in a staging development environment. The issue was just how immature it was of me to go this far to cap deploy instead of rpm install our rails app. For me, I looked at it then like a good learning experience in hacking rpms and in security. When you run sudo rpm -Uhv package.rpm - you better trust package.rpm it can execute any shell scripts it wants as root. Also, in the future I would walk away from a company like this much sooner. I enjoyed everyone there I worked with and would work with any of them again, but just would not want to work in such a stress filled environment for so long again.

Reply

kirkthejerk• 12 years ago

I mixed up the meanings of "debit" and "credit", and wrote a credit card processing app that ended up PAYING $75K to our customers instead of charging them.

I'm still not sure how this bug slipped past the bank's tough app certification process, though.

Reply

Tloewald• 12 years ago

In terms of feeling bad, I once had a client who wanted to demo a multimedia project that we currently had in alpha on his Windows 3.11 laptop, but the sound drivers weren't working properly (everything else was fine). He had about an hour before he had to leave for the airport. I started monkeying with the four horsemen of the apocalypse (Windows.ini, System.ini, Autoexec.bat, and Config.sys) as I had many times before but I screwed up saving backups, bricked his machine, and couldn't fix it). In the end it was more embarrassing than anything else, but it was a facepalm stupid mistake.

The lesson from this is pretty obvious. Backup. Make sure your backup is good and safe.

My worst work-related mistake was getting into business with a friend. It cost me the friendship, a very valuable client, and a good portion of my retirement savings. I'm not sure how related it was, but a few years later my (former) friend killed himself.

And the lesson here is not to go into business with friends. Or at least to set up the business as if you're not friends.

Reply

ClayFerguson• 12 years ago

I didn't do this, but was the one who figured out what happened. I guy wrote an installation utility for internal use to automate certain software setups. Part of the program had to clear out a certain directory, where you had to enter the name of the directory. Problem was, if you leave that field blank (the default), it converted into c:\ and people would run it, and it would wipe out their hard drive. After finding the problem, I told only the guy who did it, and no one else. I didn't have the heart to destroy his reputation by telling everyone what had done it. I SHOULD HAVE let the chips fall where they may, because I needed to be sure NO ONE ever ran that EXE "utility" again. They figured it out pretty quick, but nobody really knew the true problem but me and the guy who wrote the bug!

Reply

rokhayakebe• 12 years ago

I built a content site, worked on it for two years and a few months. While updating the entire codebase to make the site faster and easier to work on for future updates, I accidentally deleted my database. 2 years gone, SEO traffic gone.

Takeaway: Sometimes, it takes a disaster to realize you were in another disaster anyways.

Reply

deanly• 12 years ago

Not me, but a co-worker at an internship I held:

Said person entered the number of metric tons of concrete 3 magnitudes higher than it should have been. Imagine the cost difference between 1.0 * 10^6 and 1.0 * 10^9 metric tons... Our boss was not pleased, to say the least.

But imagine how easy it is to enter a few extra zeros in an excel data cell. Yikes!

Reply

coherentpony• 12 years ago

I shouted at someone.

Reply

tptacek2• 12 years ago

I once accidentally ruined the Internet.

https://www.google.com/search?q=ptacek+kaminsky+leak

Reply

erobbins2• 12 years ago

my first day ever using unix. Left with a root shell. Trying things out, learning, made a few junk files somewhere or other. I was done with that and decided to delete them... "delete everything from that directory" I think: rm * /path/towhateveritwas

now on to my tasks.. had some files to print out. Where did they g...... FUCK.

I found a box of tapes and some sunos manuals. Spent the next several hours figuring out how tar and tape drives worked. Got everything back. Never told a soul.

1992. I've never done anything so careless since.

Reply

pasbesoin• 12 years ago

Believing the CFO when he made a point of telling me, "If you ever need anything, let me know."

Gratitude is demonstrated through actions, not vague verbal commitments.

Reply

ufmace• 12 years ago

When I first started my professional career, I was a field engineer in the oilfield, working on drilling rigs around Texas. There was some amount of computer stuff, but a lot of hardware work too. One of the things that we had to do was install a pressure sensor on the drilling mud line, which is normally pressurized to around 2k psi with water or oil-based drilling fluid.

This sounds like a simple task, but it gets complicated by the variety of pipe fittings and adapters available. Our sensors are a particular thread type, and we have to find a free slot to install them, and come up with any pipe connection converters necessary to install them there. Another tricky part is that the rig workers who actually know about all of this stuff are often not particularly eager to help out.

So on one particular job, the only free slot to install the sensor is a male pipe fitting, capped with some sort of female plug. Our sensors are male in that pipe size, so I need a female-female adapter to install it. I go looking around and come up with one, not paying too much attention to it. I install it, and everything seems to go more or less smoothly. We go on drilling with this installed for like a week or two.

One day, the rig manager comes to find me and ask me about this adapter that I used. He tells me that it is meant for drinking water lines, and is only rated to 200 psi. And had been installed on a 2000 psi line for weeks. My jaw dropped in shock - I have no idea how that adapter didn't fail, and it's entirely possible it could have hurt or killed somebody if it did.

They sent one of their guys to find an adapter that was actually rated for the pressure and replace it, and never said much else of it. No telling how much trouble I could have been in there if anything else had happened. It did make me a lot more safety-conscious.

Reply

krishnasrinivas• 12 years ago

I had done "rm" of a big log file to free up space on the customer's server but our process kept on filling the log file. I assumed that the disk has got enough free space and I got busy with something else. The space was never freed up as the file descriptor was kept open by the process. Ultimately the entire disk got filled up by the opened log file and their server came to a grinding halt. I think the customer stopped using our product after that because we never heard from them again.

learning from this experience: never do an "rm" on the log file, instead do "truncate -s 0" on the log file.

Reply

embarrassed99• 12 years ago

Leading a group working in an underground bunker on a live military radar site in the Australian outback, where it rains every few years. We had to open a rooftop cable duct and when the job ran overtime we closed it up with some rags that were to hand. That night it rained.

The next morning, the bunker was full to ground level and the automatic power cutoff had failed, as the float switch was directly under the cable duct and the water pressure of the deluge and kept the float depressed. By the time the water stopped flowing the float was under a foot of mud. The powered circuits were undergoing electrolysis and eating themselves away, made worse the the site managers refusing to drain the bunker or turn off the power until a week long arse-covering evaluation had been completed.

A few hundred million dollars of front line radar was out of action for several months.

Being a naive newly graduated engineer, I wrote a completely honest report and analysis. My boss said it was one of the best reports he had read and there was no impact on my career (if anything it got me noticed by the upper echelons of the organisation).

Lessons:

1. If you tell the truth you will be respected, even if it is incriminating.

2. If there is a way for something to go wrong it can do so (slight variation of Murphy's Law). Even if it's judged to be uneconomic to take preventative action, be aware of the possibilities, so you can make a conscious decision about the risk.

Reply

mattwritescode• 12 years ago

Deleted a database table and not the temporary table I was working on.

Reply

blueskin_2• 12 years ago

Meant to reboot my desktop...

  [root@importantServer]# reboot

"Hmm... This is taking a while..."

Reply

saganus• 12 years ago

So the post is not on the front page anymore but I guess confessing feels good juging by all the people that contributed.

My screwup was at my first "real" job, fresh out of college. I was asked to free up some space from the production server at $BIGCOMPANY, because it was already at 99% capacity (it managed to get to a 100% for a few minutes before I "solved" the problem). The thing is, at this $BIGCOMPANY, for some reason the budget for disk drives was non-existant, and this meant that whenever the disk usage was at or below 95%, we were happy because we still had free space... figure that.

So here I come, armed with the most dangerous tool a newbie can wield... root access and the drive to impress your boss. I said to myself, "I've used root at my home machines plenty of times and nothing bad happened because I've been using Linux for several years by now and I know I need to be careful... so I don't get why everyone says you should never log in as root". Oh boy, how I learned the hard way.

To continue my story, it turns out that the easiest/fastest way to free up some space was to delete the log files for pretty much everything(except the last 5 or 10 logs... because we were "careful", in case we ever needed them). We usually deleted things under certain directories known to hold "useless" logs. So here comes Mr. Newbie-guy-with-the-need-to-shine, and I thought to myself, "why keep deleting the logs from the same directories over and over if that only buys us about 1 or 2 percentage points, instead of cleaning as much logs as possible for the system and freeing up a lot more space?"

After thinking about it for like 10 seconds the most genius thought of my career materializes: do an rm -rf *.log on the topmost level directory of where we used to store everything (webserver, webservices, databases, etc). I happily pressed enter, and a couple of minutes later, hooray! I got the disk usage down to a whooping 90%! I was a hero! that meant we had bought enough time to keep on working without worrying about the disk space for at least another one or one and a half months. This was a clear victory and an testament to my superb sysadmin skills.

Fast forward 4 hours, and the phone starts ringing like crazy as every other employee (non-IT ones) started wondering and then calling us to try to figure out why was their data gone. They did not understand how come they have been working A-OK so far, and then suddenly ALL data from sales team, admin team, the bosses, etc was gone. And then a few minutes later... the whole intranet came down crashing and burning.... then a full stop... nothing was working.

So we went to the logs directory... oops... no logs there!. Ok, let's try to ping the DB. Dead. It's not running and it's responding with an unknown error. When I tried to connect to it would do so, but then some cryptic ORA-xxxxx error came up. No problem says I, I'll just google it and fix it.

Not so fast young grasshopper. That error meant that the DB was out of sync with its own files used for, ironically, data corruption prevention and rollback (or something like that... to this day I still don't fully undersand what those files were used for).

As far as I can remember those logs where a sort of pre-commit place, where all changes would be stored on those files and every X amount of hours the changes would get commited to the actual DB tables. It was some functionality that supposedly was used to correct corrupted entries and to recover (figure that.. ) and rollback data when lost, or something like that. And unfortunately bringing the system back in-sync was way out my league (did I failed to mention that I was by no means a DBA?).

However a struck of good luck came down on me, as the company had a support contract with Oracle and it was the Platinum-covered-diamonds level or something. That meant that after creating a support ticket at like 1AM, I got a call from one of the support guys like less than 20-30 minutes later. This guy seems calm and tells me I should not panick, it was just as easy as doing $crypticOracleStep1, $crypticOracleStep2, $crypticOracleStep3 and voilá! all would be good again. Except for the fact that I had NO IDEA what those steps actually required me to do. Almost in tears I ask the rep to pretty please SPELL every command I needed to execute, letter by letter. I did not want to screw up again.

So there I was, at close to 2AM, with my boss breathing down my neck asking me what every frigging letter of the command I was typing did (which I had no idea...), all the while trying to keep up with this supper friendly guy that was patient enough to spell everything two times.

After a couple of commands later, behold! the DB could be brought up again! oh boy, did I felt relieved. I was jumping up and down because I had fixed my stupid mistake... or so I thought. After almost causing the support guy to go deaf due to my loud cheering, he says "however...". wait... what? there's a "however"?!?. Then he continues saying, "since you deleted the pre-commit file of the last day, the DB is back in-sync... up to yesterday". My jaw dropped to the floor. That meant that the ENTIRE previous day was utterly lost.... sales data, contracts, customer's info, etc.

I thanked the guy for his help, hanged up the phone and turned to my boss telling him that I was ready to turn in my resignation letter just after helping capture what availabe data was actually there (in papers, by calling customers and asking them again, etc).

My boss then turns to me and says, don't worry. We've all been through this at least once in our careers. Even I made a mistake that is terribly similar... however when I brought down the database, it took us one full week instead of one day... and rest assured that as I learned my lesson, you did as well. And I need guys like you, that have the initiative to solve things... and the ability to learn from mistakes. So don't worry, you are not losing your job. However you can't go home until you help everyone get as much data back as you can.

Aw shoot... well.. I guess it could've been worse. So after having lunch with my boss and the other teammates at like 6-7 AM, I went to the sales dept and started asking around how could I help them get their data back.

Those were the longest 38 continous work hours I've ever had to resist. I did not go back home until more than a full day and a half later. I was tired as hell to say the least... but to this day I think it was a blessing that I got to learn such a hard lesson but being backed up by a boss that was very cool and progressive about it.

Lessons learned:

0) Never ever ever ever use root, especially for deleting files and ESPECIALLY with the -f flag.

1) Do not assume that something you know will hold. Confirm it in the particular system you are going to be working with. (i.e. do not assume .log files are always log files because in your laptop that holds true)

2) Be ready and willing to assume the consequences of your actions. Most of the time if you assume responsability for your mistakes, people will forgive you and even give you a piece of advice.

3) Never ever ever ever use root.

Reply

danellis• 12 years ago

I performed a two-minute manual query on a MySQL database I was told was a backup. What I didn't realize was that it was a live backup, and that it would stop the production database from responding to queries for those two minutes, meaning that authentications failed for two minutes. Several colleagues called for my employment to be terminated immediately, but luckily they got distracted by other issues.

Reply

bobdvb• 12 years ago

I used to work at a major European telecoms company... Unlike everyone else I am not talking programming mistakes, these are generally physical fumbles:

One day I was doing a change control, I was scheduled to change some settings on the modulator of a satellite system providing internet access to a portion of the Middle East. I called the satellite operator and told them I wanted to do the scheduled work, as they would have to confirm I was outputting new configuration within the constraints of the contract.

I entered the change and re-initiated the modulation, the operative said he was seeing nothing. Now, because this signal was going to the middle east I couldn't see it in Europe and without substantial plumbing I couldn't tap in to the antenna. My heart started going, I was checking amplifiers, up-converters, everything and I couldn't see anything wrong. After a few minutes the client called panicking because the action should have taken moments, not minutes. After more confirming with the satellite company that I wasn't transmitting I checked back through my steps and eventually saw I had missed one crucial thing: when you changed certain parameters it muted the output! A quick few button presses and the patient man on the phone said "there it is!" and I can relax again.

Lessons learnt: 1) I should have noticed a critical LED on the modulator was not lit! 2) This is the reason change controls say "working period 10 minutes" (time taken 9m50s). 3) A good boss will defend you if you recognise your mistakes and don't f*ck up too badly. 4) Don't go for a quick drink with your brother before a night shift.

In the broadcast industry they say you aren't a real engineer until you've taken a TV network off air. Lets say I am very experienced, but my employers have never had a problem with me. This is probably because I have also seen people meltdown under the pressure of delivering live services to millions or even billions of viewers, but if you can keep cool you can deal with it.

Reply

Ask HN: What's the worst you've ever screwed up at work?

Related Stories