Hacker News Reader

Comment by yeahdef | Hacker News Reader

willing to bet that 99.999% uptime is just static text, because 5+ hours out of 2190 hours is not 99.999%

From @koopajah on discord: "We don't really drop nines if one specific feature is having issues". Pretty convenient payment processing is just a "feature" of a payment infrastructure SaaS

This is a bullshit and untenable position. This “one specific feature” takes out the core feature. Shameful, deceitful and reputation tarnishing position.

To be fair, it only takes out the core feature for people using Stripe Tax, which is apparently a small fraction of customers.

In a OKR culture, the way to make your KR good is to add lots of useless and very simple microservices, which are always up, so the overall metric is 100%.

If your KRs are the uptimes of your simple microservices, you're not in an OKR culture.

I guess 99.999% uptime is no longer an engineer's well-earned badge of honor at Stripe. Now it's just sales puffery, like a LOWEST PRICES IN THE CITY!! sign at a discount store.

Sorry, but true class isn't just elegant rectangles decorated by various subtle shades of gray sans serif.

Hate it when companies do it. It should be the min of the availabilities of your APIs.

a.k.a. "my probed availability ignores user journeys that are actually an important part of my product"

the service doesn’t look to be down for everyone.

if their status is a percentage of requests, then a relatively small number of 500s compared to a big total volume will be tiny, even if ongoing.

Stripe has only existed for 2190 hours? /s

It is not static text. Calcuated automatically.

It's disingenuous to report "partial degradation" as "uptime" when some customers are unable to process any payments.

Yes, you either need to have a per-customer uptime number (that is visible to those customers) or you have one unified uptime number that takes a hit if any customers experience downtime. You can't have it both ways.

Why can't you maintain and report an average uptime across all service usage? So if you have an outage that affects 1% of traffic it moves your figure 1/10th as one on 10% of traffic?

That's what I'd expect a reported number to be, since that's what a client experiences on average.

Reply on news.ycombinator.com