I am student that currently using azure student, but their free vm has no capacity left in all regions it is the Standard_B1s vm size.<p>I wouldn't mind putting my credit card in azure to access to bypass the capacity issue, and get their paid vm, but I am worried about messing up, and getting an expensive charge, but is there any reason with cloud providers like azure I can't prepay?
Most of their money is coming from larger customers. So the small customer afraid of going over budget isn't a major concern.<p>They'd also rather give refunds for accidental overages than build in billing shutoffs. There are two reasons for this:<p>1) Their infrastructure doesn't have realtime tracking for usage so they can't detect you're over budget until some time after you've surpassed it already.<p>2) There is a decision that most customers would rather go over budget when something happens than have their services get shut off.
I shut down my Google Cloud Billing account a couple of days ago. I only used it for HashBackup testing, and only to test Google Cloud Storage specifically, but there are too many recurring stories about crazy charges at all storage providers, and since trying to actually speak to the Great and Powerful Google is impossible, I just decided to close it. I was a bit anxious about doing anything with it because I didn't want it to screw up my Gmail account, but having moved that to Fastmail a few months ago, Gmail isn't such a risk now.<p>I did check into their Budget screens, but holy cow - there are 1900 services there! And besides, setting a budget doesn't help if someone breaks in while I'm sleeping, racks up $10's of thousands of dollars of charges, and I get a notice the next day. So I never did that.<p>Canceling the billing account lets me sleep at night. And if I absolutely, positively have to get back on, it looked fairly simple to reactivate the billing account.
Used to work for one of the big ones<p>IMO the reason they don't do this is because you would have to design every system with billing as a critical path dependency<p>For example say you have a service that lets people run batch jobs<p>What happens if the pre-pay credit of the account runs out mid way through the job?<p>Does your job service have to constantly check the remaining balance as its running and abort halfway through? What happens if the account balance service goes down?<p>Or would you somehow reserve the credit before the job starts? Well in that case how do you know how long the job will run?<p>Its possible to do in theory but it would be a pain to support
Is there a specific reason you're tied to Azure (or any of the other 'big' providers)?<p>I do all my spin-up/spin-down testing at Digital Ocean, Vultr, or Hetzner ... even if I happen to forget something's up and running, it's never more than ~$5-10 in a month (and I get the billing statement every month, so I can quickly nix stuff if I did happen to forget about it)
It's a lot of trouble and ultimately they lose money on the changes. Right now if someone goes over whatever limit they set for themselves they mostly pay the chargers or on the rare cases they try to dispute them. With a set limits the provider will always lose business. Also trying to make sure the limits are not met is extra CPU power that could otherwise be sold for a profit. Bottom line, it does not make business sense .<p>It doesn't even make sense as a competitive advantage since it costs more to maintain than what it brings in. If I owned the cloud service I would only set the capability if the customers demanded it to the point of them switching to a competitor en masse.
DigitalOcean allows this if you use paypal as a payment method. The interface design makes it quite clear that they'd prefer to have your credit card instead though.
> but their free vm has no capacity left in all regions it is the Standard_B1s vm size.<p>As an aside on this topic, I'm finding that they don't have capacity for some <i>paid</i> VM types as well in some regions. If I'm looking for specific instance types that don't exist in my default region (e.g. GPU optimized compute) I have to go to other regions to find one where those are available.
I would say all the major cloud providers are full of foot guns that can end up charging you frightening amounts of money. I'm undecided if it's actually intended or just a side effect of their complexity, but either way it's not something they're trying to fix. There are tons of other models out there besides them though - traditionally small VPS providers are what you are looking for - from no names to something in between like Linode or Digital Ocean. AWS emulated that model with Lightsail as well.<p>But if you want my actual protip, then check out Oracle's cloud product. I'd mostly steer you away from Oracle, but in this case their cloud has a good always free tier and will not allow you to overrun the free boundaries. It's also basically an AWS clone so what you learn will have some carryover as well.