Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

> Clearly something is going on with AWS's user experience.

But that's not the case. It's a false positive.

Pick a DownDetector service and open the page every day for a few days. You'll see it most of the time just reflects people waking up in the US timezones.



Is it a false positive, though? The data shows there was an outage. We would need more evidence to conclude hundreds of users, at that 1 spike, weren't actually having issues.

In other words, we have hundreds of people saying there was an outage, and 1 person saying there wasn't.

That's a problem AWS needs to resolve, regardless of what they think might be the root cause. If the users weren't experiencing any issues with AWS, I doubt they'd be reporting it.

Your comment about timing is a good point: if people are working with AWS early in the day, and AWS is giving them problems, then they will probably report problems with AWS early in the day. I wouldn't expect them to report problems while they're sleeping.


> Is it a false positive, though?

Yes. AWS was not down this morning.

> In other words, we have hundreds of people saying there was an outage, and 1 person saying there wasn't.

We have hundreds of millions using AWS and AWS-backed services successfully this morning.

I'm out.


Hundreds of users, representing more users who didn't bother reporting, say they experienced issues when interacting with AWS this morning, so we'll need better evidence to the contrary to conclude otherwise.

The fact that some people accessed AWS without reporting issues does not mean that all people did. For those who had issues, AWS is responsible for dealing with those perceptions.

Indeed, it could have been a fault that affected a subset of users, for example 1 service in 1 availability zone. That's still an outage in the eyes of users, which AWS is responsible for managing. It could have been an issue with a route from 1 ISP. That's still an outage in the eyes of users, which AWS is responsible for managing.

An even better example is the DownDetector page for Facebook, with hundreds of thousands of reports. Do we really think there's no correlation between what DownDetector reports and what users experience?

tl;dr: what users think about your site is more important than both what you think about your site and the reality of your site, and you should be tracking it.




Consider applying for YC's Summer 2026 batch! Applications are open till May 4

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: