Recently I've noticed, and my users have noticed, that Weatherkit is returning dubious values for apparent temperatures. For example, in monitoring S. Florida this week, Weatherkit is returning apparent temps 1 to 2 degrees warmer than actual temps. (85, feels like 86.) All other services are reporting ~10 - 15 degree differences. (85, feels like 96) I believe this is a recent development. Are others having the same issue?

Is anyone else experiencing issues with the REST API today? I'm getting very slow response times (again.)

Once again, WeatherKit is returning a slew of 502's or is being very, very slow. And, once again, the system status is not showing the problems. If Apple actually communicated about the problems with WeatherKit, that would be so much better.

After Apple essentially shut down Dark Sky early (you can't trust it if it's off-line 8 hours a day 2/3 days), I've transitioned production traffic to WeatherKit. Generally slower (higher p50's, significantly higher p90s) Under bursts of traffic, it just chokes. Always. I have many users who have things set up to retrieve the weather on the hour (e.g., 6:00am). I have to have a failover weather source for these peaks, because weather kits just stops returning data under heavy load. It's not really acceptable that I'm paying significant $$$ for an API, and have to have another source for when the API predictably chokes. Dark Sky handled the load without breaking a sweat. Ever. I'm not sure what to do here. It's not an issue with the code, so there's no point in opening a code level support ticket. The feedback assistant refers you to the forums. General performance issues are not addressed by Apple engineers on the forums. So, the whole thing is a bit Kafkaesque. Is stumbling under load expected behavior? Does Apple have any performance expectations for the API? I've never dealt with an API provider that you couldn't get some sort of acknowledgement about issues. It's just weird.

Another day, another slew of issues with WeatherKit. It's not ready. Apple: you said you'd have a replacement for Dark Sky available. People made plans around this. Do the right thing and keep Dark Sky API available until you're able to sort through the issues with WeatherKit.

Trying to transition over to weatherkit for production traffic. However, the pattern is clear: as the load increases, the frequency of slow responses ( > 650ms) and 502 Bad Gateway errors increases. In peak conditions, more than 50% of requests fail. Apple says the API is not rate limited. So just wondering if others have experienced this and found a solution.

So, as we get closer to Dark Sky API closing, Weatherkit it still incredibly unreliable for me, especially under a heavy load. Periods of lots of 502 Bad Gateway errors. They come in waves, then go away, then come again. Has anyone opened a support ticket for this issues? Has anyone gotten a response that would be helpful for the community?

Just curious what people think. As of today, at least, WeatherKit REST still doesn't seem ready for production traffic. Still lots of 502:Bad_Gateway and slow response times. (I've set a timeout of 650ms with 3 retries + backoff and still get timed out frequently.) Will Apple keep Dark Sky going until WeatherKit REST is ready?

As others have pointed out, right now the REST API is slow, especially above p50. I'm hoping that it will settle down by the time Dark Sky shuts down. If we set a reasonable timeout for the API call (e.g., 650ms) are we charged if the call times out?

I'm just posting this for visibility. Like many others, I'm seeing a lot of 502 errors. And, just like the prior threads about the intermittent 401 errors, there's no response from Apple.

I'm also getting occasional 502 errors along with the well-documented 401 errors others are reporting. Retries go through on both, so it sure seems it's (another) WeatherKit issue. I wish someone from Apple would at least say they are looking into it. The silence is disconcerting.

WeaktherKit REST API returns snowfallAmount. DarkSky used to return precipAccumulation (for snow.) Are these measuring the same thing? The Weatherkit Docs say that snowfallAmount represents: "The depth of snow as ice crystals forecasted to occur during the day, in millimeters." I'm not sure I really understand. Where is the "depth" to "occur"? Is it supposed to refer to the actual amount of snow predicted to accumulate (like, on the ground)? (The documentation for transitioning from DarkSky https://developer.apple.com/weatherkit/get-started/ doesn't really clarify, I don't think.)

So, is the weatherAlerts field ever populated? From my (limited experience) it does not populate with current NWS alerts.

Anyone else have worries about the accuracy of WeatherKit so far? For example, current temps do not correspond to (the generally very accurate) Dark Sky. In my admittedly very unscientific tests, the forecasts seem to be weirdly worse than those given by Dark Sky.