2 ms·
This gets brought up a fair bit. A major issue is that hard bill caps cause not just an availability concern, but also a durability risk (data loss). From the c
by ericpauley 4y ago
This gets brought up a fair bit. A major issue is that hard bill caps cause not just an availability concern, but also a durability risk (data loss). From the cloud provider’s perspective, it’s probably easier to save face by comping accidental overages than risk deleting user data.
One solution here would be to identify SKUs that are not durability-critical, such as just instance runtime (though even this has caveats), and enforce caps by killing these.
- Havoc 4y agoIt's definitely a tricky question once you get to proper production context & I think their stance there is overall reasonable given the thorny questions. However I don't see why they can't provide a sandbox mode with associated terms. They delete user data & impose restrictions all the time with trial accounts after all. If they want to get people to try new parts of the million cloud products providing a safer mode to experiment would likely do a lot to encourage that. Hell I mostly steer clear from learning BigQuery precisely because I've observed it seems to feature disproportionately frequently in posts like these. If I could designate a section of the account as testing/non-prod and limit it to $100 I'd definitely explore more freely. >easier to save face by comping accidental overages than risk deleting user data. With the added benefit that you get to keep the money from the likely substantial group that writes off minor mistakes as school fees / not worth the hassle. Probably immaterial but worth keeping in mind that comp cost flow through 100% to bottom line profit/loss. Granted all the clouds seem very generous with their comp'ing accidents, but still feels like a very intentional power dynamic - rather than giving the user control the clouds reserve unilateral power to decide the users fate. The fact that they're magnanimous doesn't change that.