I totally agree. You should be able to choose to have things shut down rather go past a limit. I don’t have a single client that would choose $70k versus, say, a couple hours downtime while we figure out what is happening with a resource that is going crazy.
This is where I question the risk of serverless. Although, now that I think about it, while my one client’s EC2 instances are essentially capped in terms of capacity and spend, we also use S3 etc. I suppose it would be entirely possible to accidentally write a huge amount of data to S3. But again I would rather get warnings from the app that writing to S3 has failed due to limits than get a huge bill!
But you also pay for data at rest in S3. Should S3 stop storing that data while you figure things out? Should they bill the customer for the deletion of that data as they normally would?
I don't disagree on wanting this feature, but it's just not something that's possible to implement in totality when you dig into the details.
Could it work based on deviance from norms? Like a setting where you say, I’m okay with up to two standard deviations from our typical write volume but that’s it?
Or, alternatively, just let me set the size of the volume. Treat it like a hard drive?
In terms of your point about the data at rest, part of the issue is that we get a bill once per month, and that’s probably when we would notice it. Of course there is probably a Cloudwatch alarm or something we could set (I assume) but there’s so many damn services…
This is sort of what quotas enforce, and most cloud services have quotas. You ask for an increased quota if you reach the current one, or the system might sometimes automatically increase it if it thinks you will need it.
This all trades off though against the possibility of bringing down one of your customers when they are hitting peak sales on their website, which is a very bad look.
This is where I question the risk of serverless. Although, now that I think about it, while my one client’s EC2 instances are essentially capped in terms of capacity and spend, we also use S3 etc. I suppose it would be entirely possible to accidentally write a huge amount of data to S3. But again I would rather get warnings from the app that writing to S3 has failed due to limits than get a huge bill!