What was the mistake?
The lesson for me was don't trust your internally-hacked-together instance management system. The AWS interface to storage and instances is the base truth. And perhaps more importantly - I'm never getting into another startup which has financial risk like that without being a core expert in that risk/tech. I was focused on the business + client code - and had very little clue about the nitty-gritty of AWS. I should have been more involved with the code on that side, or at least the data-flow architecture.
Scenario 1: Amazon will ask for the payment (if using cc); the bank will respond there are no funds in the account; Amazon deals directly with the company further directly, not with the bank, eventually getting payment order from the court. If the company went bankrupt meanwhile, Amazon might not get their money.
Scenario 2: Amazon will send the invoice; invoice will not get paid. After due date, Amazon will contact the company directly; bank doesn't even enter the picture, until collection order comes from the court. If the company went bankrupt meanwhile, Amazon might not get their money.
There's no scenario where some hypothetical loan would go straight to Amazon, unless Amazon has some instrument, that instruct the bank to pay them. Something like bank guarantee or promisory note, and uses them before declaring bankrupcy.
So for Amazon draining loaned money, they would have to transfer them to a normal account and pay with debit card paired to that account, with no limit set.
It is not wise to transfer them to a normal account; you pay interest for the balance on the loan account; if you move them to your normal account, you are paying interest for money that is sitting on your normal account.
That's why you don't pay large sums with CC, but with invoice + bank transfer, and have a limit set on your cards, when you do.
- control: you are in control, when you do the payment. You can plan your cash flow.
- additional advantages: You also have payment terms, some vendors offer discounts for earlier payments; if your cash flow can handle that, why would you giving up of that?
- liability: with CC, you are getting credit that is drawn at other party leisure. It's you, who is liable for this credit line, even if the other party made a mistake. You are always liable to the bank, never towards the vendors. With bank transfers, every single payment was authorized by you (where by 'you' I mean authorized person at your company) and the liability is towards the vendor, who is not likely to have such a strong position (see Porter's five forces).
- leverage: if another party makes a mistake, they have motivation to correct it. Every company in existence has already received invoices, that are incorrect. Withholding payment until they are corrected is a strong motivator. Without that, you could be left without invoices that can be put into accounting AND without money that you have to account for.
- setting up processes: when you grow beyond certain size, you are going to want to formalize both the procurement, accounts payable and treasury. Having purchasing and payment discipline that are compatible with that already in place will mean less pain from the growth, less things to change.
When we need people in the field purchasing small supplies, we don't want them to handle cash, so they get debit (not credit) cards, with relatively small limits. It is enough for them to get by, but not enough to make any damage of significance. (The exception is fuel and that's what fuel cards are for - basically it has a form factor of a credit or debit card, but works only for fuel, is paired to a license plate and the vendor sends invoice at the end of the month).
Another scenario, where CCs are useful, if you need to pay something right now; you don't or can't want to wait for the order->delivery+invoice->payment cycle. That's fine for consumer impulse purchases, but that should not be a normal way for company purchases.
Of course, if you start a new business relation, some companies would not trust you, that you are going to pay the invoice; sending advance invoice and paying it is fine. In practice, it is quite rare occurrence.
The real world is filled with barbershops, daycares, bars, clinics, PVC manufacturers etc
None of them get VC money.
When they need money, they go to a bank and usually have to place a PG in order to get funds.
Tech startups have it easy. Its all equity. You are not pledging your lifetime earnings on a business idea.
Once tech startups lose their upside potential (prob not anytime soon if ever), you will be sitting with the regular folk, those that pledge their skin and life to their business.
Either way it might be nice to keep your options open, depending on your plans.
Really? Working in Southern California a few years ago, refund requests were refused ALL THE TIME. This is why there's a common belief that what you are charged you simply owe them, period.
It may be more progressive now, but let's not be revisionist.
I know several other companies that had expensive mistakes without refunds. There's probably a complex decision tree for these issues and I doubt anyone really knows outside of AWS.
Being an entitled jerk who blames other people for your own negligence is bad, and you shouldn't change that. But openly giving companies the opportunity to be kind (while admitting that it was entirely your fault) potentially helps both them and you.
In theory ;), you shouldn't have to be a core expert in everything. But yeah... in the real world, things aren't so cut and dry. :/
Easier for both sides to just ask AWS for a refund if there's a reasonable case.
This wouldn't be an issue if it was configurable.
Would you rather make a mistake leading to a big bill with the possibility of a refund or set your max budget and have your resources permanently deleted?
EC2 instances, EBS volumes, S3 data... should AWS delete those when you hit your budget? How do you stop the billing otherwise?
With prioritisation, so the non-steady state services are stopped/killed with plenty of time to leave the needed foundations still running. :)
2) If it's a soft budget then it's no different than the alarms you already have.
3) If you want to stop it before it hits the budget, then you're asking for a forecasted model with a non-deterministic point in time where things will be shutdown.
This just leads to neverending complexity and AWS doesn't want this liability. That's why they provide billing alarms and APIs so you can control what you spend.
Not if I'm busy, or away from work, or asleep. There is a massive difference between getting an alarm (which is probably delayed because AWS is so bad at reporting spent money) versus having low priority servers immediately cut.
Even without a priority system, shutting down all active servers would be a huge improvement over just a warning in many situations.
You want it to selectively turn off only EC2? Does it matter which instance and in which order? What if you're not running EC2 and it's other services? Is there a global priority list of all AWS services? Is it ranked by what's costing you the most? Do you want to maintain your own priority of services?
And what if the budget was a mistake and now you lost customers because your service went down? Do you still blame AWS for that? Or would you rather have the extra bill?
There is no easy solution.
"Everything except for persistent storage" is nowhere near useful enough to work and can cause catastrophic losses. Wipe local disks? What about bandwidth? Shutdown Cloudfront and Lambda? What about queues and SNS topics? What about costs that are inseparable from storage like Kinesis, Redshift, and RDS? Delete all those too? And as I said before, what happens if you set a budget and AWS takes your service down which affects your customers?
It's easy to say it's simple in an HN comment. It's entirely different when you need to implement it at massive scale and that's before even talking about legal and accounting issues. There's a reason why AWS doesn't offer it.
I sometimes for example fiddle with Google APIs. I do not even have customers so don't really care if things will stop working, but I have accidentally spent 100 euros or more. I have alerts, but those alerts arrived way too late.
I make a loop mistake in my code and now I suddenly owe 100 euros...
I literally just explained why this doesn't work with AWS services. You will have data loss.
And it creates a whole new class of mistakes. If people mistakenly overspend then they'll mistakenly delete their resources too. All these complaints that AWS should cover their billing will then be multiplied by complaints that AWS should recover their infrastructure. No cloud vendor wants that liability.
ADDED: A lot of people seem to think it's a simple matter of a spending limit. Which implies that a cloud provider can easily decide:
1.) How badly you care about not exceeding a spending threshold at all
2.) How much you care about persistent storage and services directly related to persistent storage
3.) What is reasonable from a user's perspective to simply shutdown on short notice
Except it wouldn't. This is the 3rd time in this thread explaining that. Edge cases matter, especially when creating leading to new mistakes like setting a budget and deleting data or shutting off service when customers need it most.
If it's not a hard budget but a complex set of rules to disable services... then you already have that today. Use the alarms and APIs to turn off what you don't need.
> If it's not a hard budget but a complex set of rules to disable services... then you already have that today. Use the alarms and APIs to turn off what you don't need.
I have been describing a simple set of rules, not a complex one.
It used to be extremely difficult to get accurate usage data on all their services. Has that been fixed? If not, then the alarms aren't good enough. If the alarms can automate enough right now, in a non-buggy way, then that should be the answer to people "hey, the alarms do more than alarm, use them to trigger shutdowns". Don't say "it can't be done, sorry". If the alarms aren't good enough for that automation, then the argument stands.
And using the APIs means that each company that wants safety is duplicating effort in an almost untested way, a recipe for so many bugs it makes the problem worse. No, this needs to be a feature of AWS itself.
Not so simple.
That's not a he's cap, since turning off services isn't instant and costs continue to accrue. But, yes, there are ways to mitigate the risk of uncapped costs and they are subject to automation.
AWS would rather lose some billings than deal with the fallout of losing data or critical service for customers (and in turn their customers).
In theory I could build something using budget alarms, apis, and iam permissions to make sure everything gets shut down if a developer exceeds their budget, but if I made a mistake it could end up being very expensive. Not that I don't trust developers at my company to use such an account responsibly, but it is very easy to accidentally spend a lot of many on AWS, especially if you aren't an expert in it.
It's impossible for AWS to know how to handle hard caps because there are too many ways to alter what's running and it's too contextual to your business at that moment. That's why they give you tools and calculators and pricing tables so that it's your responsibility (or a potential startup opportunity).
Money is easy to deal with. Alarms work. Bills can be negotiated. But you can't get back lost data, lost service, or lost customers.
Right. In my experience, if you don't understand what's going on beneath your abstractions, you're always in for a world of hurt as soon as something goes sideways.
This may be something that is 'unstated', but unless you actually had access to fix something that was wrong, as well, being an expert in that wouldn't really help all that much. I've been in situations where I have explicit/expert knowledge of XYZ, but when the people responsible for XYZ do not take your input, and/or don't provide you the ability to fix a problem, expert knowledge is useless (or worse, it's like having to watch a train wreck happen when you know you could have stopped it).
https://docs.aws.amazon.com/AmazonCloudWatch/latest/monitori....
If you don't want to pay for PD, you can patch together any number of ways to get your phone to scream and holler when it gets an email from ohshit@amazonasws.com. It's also good to have clear expectations as to whose responsibility it is to deal with problem x between the hours of y and z and exactly what they are supposed to do.
Keep the alerts restricted to the really important stuff, because if your team becomes overloaded with useless alerts they will 1) dislike you and 2) be more prone to accidentally mistaking a five alarm fire for a burnt casserole.
There are more complex systems you could build, but that's a start.
I remember PagerDuty was advertising (a lot) on Leo Laporte's podcasts a few years back.
A clause in the contract: if monthly bill reaches $Xk amount then:
(a) seek written approval by client, and
(b) continue until $Yk or approval is given with a new ceiling price.
I imagine AWS would have 0 problems suspending all my services if I can't pay, so why can't it do the same thing when it reaches my arbitrary cap?
That might make sense for some particular services (e.g., capping the cost on active EC2 instances) but lots of AWS costs of data storage costs, and you probably don't want all your data deleted because you ran too many EC2 instances and hit your budget cap.
Where exactly you are willing to shut off to avoid excess spend and what you don't want to sacrifice automatically varies from customer to customer, so there's no good one-size-fits-all automated solution.
They have a good track record of cancelling huge bills the first time they happen
Does AWS update the billing console per day or upon request? I get charged per month, but I should add a habit in my habit tracker to learn more about my expenses...
In my experience AWS had very stringent limits on the amount of active instances of each type (starts around 10 for new accounts, 2 for the more expensive instances). It takes tickets to support then days of waiting to raise these limits.
That should have prevented your company from creating tens of instances, let alone hundreds, unless that's already your typical daily usage.
It's horrifying how many places treat writing tests for services as critical, but then completely fail to write tests for their operational tooling. Including tools responsible for scaling up and down infrastructure, deleting objects etc.
The problem with billing is that often these charges are not calculated instantly, and others are not trivial to deal with. For example what happens if you go over budget on bandwidth or bucket storage, but still within quota? What do you kill? Do you immediately shut down everything? Do you lose data? There are lots of edge cases.
You can normally write your own hooks to monitor billing alerts and take action appropriately.
It's the very very first thing I set when setting up my GCloud hobby project. I was like, this is fun and all, but I don't care about this enough so I limited it to 3$ per day and 50$ per month. If it goes above, I'm very happy to let it die, and it also gives me a warning so I know something is up. The 2 times it triggered, there was something I managed to fix so the tool is still up and running costing pennies.
The account I did it on was tied to my "junk" email, so I didn't catch amazon banging on my door saying my payment info needed to be updated. Well until I did happen upon one of the emails. Nearly had a heart attack.
Talked to aws support and they full refunded me. Very very kind of them, but now I'm terrified to touch anything aws.
Greed, I'm assuming.
I also don't understand why everyone is assuming
"if I hit threshold X do A, if I hit threshold Y do B" where A and B are some combination of shutting down and deleting resources,
is as difficult as solving NP complete.
Now I have learned, _always_ refresh the page and instance list prior to shutting anything down and _always_ confirm the shutdown was successful.
I set the lifecycle rule on all objects in the bucket, for as soon as possible (24 hours).
About 2 days later first thing in the morning I get a bunch of frantic messages from my manager that whatever script I was running, please stop it, before I'd even done anything for the day.
The lifecycle rule had taken effect near the end of the previous day, and he was just getting all the billing alerts from overnight, it was all done.
I read about glacier pricing, but didn't realize there was a lifecycle transfer fee per 1000 objects (I forget the exact price, maybe $0.05 per 1000 objects). That section was a lot further down the pricing page.
The bucket contained over 700 million small files.
I'd just blown $42,000.
That was over a month's AWS budget for us, in the end AWS gave us 10% back.
On the plus side, I didn't get in too much trouble, and given we'd break even in 4 years on S3 costs, upper management was gracious enough to see it as an unplanned investment.
TLDR: My company spent 42k for me to learn to read to the bottom of every AWS pricing page.