How to monitor your local weather with Grafana | Hacker News Reader

How to monitor your local weather with Grafana

grafana.com

202 pointsmeysamazad41 comments

I‘m ingesting data from my local weather service into Prometheus and display it in grafana, and what I find most interesting is how the forecast for a certain day changes over time, e.g. seeing how many changes happen until the service lands on the actual precipitation probability for the next hour, as well as correlations between changes across days - something most APIs don’t provide by default but once you feed them into Prometheus you get the historic changes easily.

This is an idea that I’ve long wanted to explore. It’s along the lines of using the GEFS weather model, where the standard deviation model output parameter can be used to predict how settled a model run is.

Anything you can share? I would love to play with this, although my interest is in marine weather forecasts.

I did this a few years ago (sans plugin) and correlated weather to velocity and WIP (Jira) for our development team. I was disappointed to find no meaningful correlation.

Years ago I worked on a live TV show that allowed viewers to call and ask questions to hosts. It was a small local TV station but the show was some what popular. We wanted to know what variables affected how many calls there are per show and we manually tracked a bunch of data points and one of them was the weather. Once enough data was collected it was very clear for our show the weather had a big impact, when we had beautiful weather we had less calls but when the weather was horrible we had a lot more calls. In hindsight it was obvious that when it rains people will stay home and watch TV.

Another fun thing this data collecting showed me was subjective bias of certain managers. One of them was sure that a certain hosts were getting much less calls than the others during the show but the data showed that besides one "star" all the hosts have about the same amount of calls even though it "felt" like they got less.

Not to “captain obvious” your efforts but did you really need to prove this?

Yeah, I feel 2025 will be the year where you need petabytes of data, LLMs and dashboards just to get ‚insight’ like that

Yes, because some hosts were getting unnecessary grief because it “felt” like they weren’t doing as well as others when in fact they were doing just fine.

Or maybe your show was affecting the weather

I monitor my local weather using a windows based system.

I use a combination of windows and doors. People say they help with requirements somehow, however I believe they're most suitable for verification of the monitored state.

I use a rock hanging from a string.

https://en.wikipedia.org/wiki/Weather_rock

I use a FL calendar. Mar-Oct: sucks. Nov-Feb: sucks less.

Sounds like the FL calendar is about 100% inverted from the ME calendar. :)

> I use a rock hanging from a string.

And where is this rock hanging and how to you observe it?

We all try to escape the claws of Doors & Windows Inc, but whatever you do, when you interact with The Outside, they're required...

is it wifi, bt, zigbee? surely matter-compatible, right?

it is a pane to set up

Very happy to see Grafana adopt this plugin to continue it's development among all the various other data sources grafana can plugin in to. I'm happy to bring Grafana and it's stack to the work place and developers are very happy using it with their OpenTelemetry integrations. Their only complaint is that some of the settings for panels can be difficult to find.

My own wish is that I hope something like this https://github.com/knightss27/grafana-network-weathermap is adopted too. I know it maybe possible using Canvas but Canvas is too much of a blank slate and needs a lot of work to set it up the same way.

> Very happy to see Grafana adopt this plugin

we adopted the author along with it :)

> My own wish is that I hope something like this https://github.com/knightss27/grafana-network-weathermap is adopted too. I know it maybe possible using Canvas but Canvas is too much of a blank slate and needs a lot of work to set it up the same way.

good feedback. it's been discussed on and off, and we'll def improve Canvas in the future, but the viz team is fairly small so we have to manage how much stuff we take on to maintain forever.

I really hope grafana adopts a code-based back end so I'm not having to fiddle around with settings in the UI, which are very very very very terrible

Perhaps you can have a go with jsonnet. https://github.com/grafana/grafonnet it's not perfect but I believe there's some progress towards dashboards as code.

it's coming, and high prio. stay tuned for "as code" updates in the next 6mo.

making this experience good (e.g. composable dashboards) required some fundamental schema changes that were difficult to do until recently.

a lot of this re-schematization is happening as "v2" work, e.g. https://github.com/grafana/grafana/pull/98408

You can take this a step further and get one of those 5-in-one/7-in-one weather stations and then listen to its communications to its basestation with an RTL-SDR using rlt_433 and then pipe those datapoints into a Prometheus + Grafana installation.

https://github.com/merbanan/rtl_433

If you're lucky a neighbour may already have a compatible weather station nearby, which you can freely collect data from as long as you're in range ;)

I ended up buying a bunch of temperature/humidity/pm2.5 sensors that communicate over Zigbee, then collect the data with Zigbee2MQTT, which was trivial with the mqtt_consumer in Telegraf.

The benefit (or drawback I guess) with your approach is that you can also ingest your neighbors data, if you can reach it :)

I'm gonna stay that's a little more than "a step further"

Davis Instruments is on the pricey end but their WeatherLink Live gateway has a simple local JSON API. For years I scraped it with Prometheus for use in Grafana but I have since switched to HomeAssistant for the analytics/visualization just for simplicity.

this is just marketing spam for: install this plugin.

the Infinity plugin is pretty awesome, free, and oss; i'm not sure this qualifies as spam, per se. nothing there requires signing up for cloud/SaaS, and you can 100% run this locally.

i mean, you can always roll your own UI if that's something you prefer / can do. (grafana uses uPlot [0] if you need to render millions of points).

disclaimer: i work on dataviz at grafana

[0] https://github.com/leeoniya/uPlot

I'm guessing that because the article/blogpost only talks about Grafana Cloud and doesn't even mention that there are other ways of running it, it ends up reading kind of like marketing post rather than just instructions/engineering post.

> it ends up reading kind of like marketing post

well, that's not untrue either :)

Yeah, it also has high "draw the rest of the owl" levels. It just tells you how to import someone else's dashboard as opposed to telling you how one might configure that from scratch.

All the steps that promise slightly more details just mention "Coming Soon" [0]

[0]: https://github.com/scarolan/grafana-cloud-tutorial/blob/main...

This article's primary usefulness is in reminding me that there are some great public APIs out there that I could use to feed into a personal dashboard in Grafana.

Please for the love of god be gentle on the NWS infrastructure - and also ask yourself "do I really need to set up something that hits NWS public APIs every X minutes, or can I just get a $5 dongle to decode ISM band transmissions and read a $10 temp and humidity sensor stuck to a tree outside my house?"

Private weather companies have been lobbying Congress for decades to keep the NWS underfunded and hamstrung in what weather services they can offer. That's why we've seen an explosion in the number of companies offering things like forecast "products" for farmers, fleet operators, and so on.

The NWS is like the post office- invaluable, something the government should provide as a service to all, and constantly under attack and hamstrung by people with privatization agendas.

Coincidentally, I was just looking for an easy way to visualize temperature data from my thermostat and outdoor sensors, and found an InfluxDB logger plugin for Hubitat. It only took a few clicks to get it to send all of my switch and sensor data to InfluxDB. Now I just need to finish creating some dashboards and alerts.

https://hubitat.com/

https://community.hubitat.com/t/using-he-with-with-influxdb-...

“Grafana Cloud” sounds a little limiting in the context of a weather app, no?

Is there a list of projects that have used (or expanded) upon this? Curious to see how far people have taken this.

I have build a grafana weather dashboard with the data from an enviro+ some years ago. It monitors a lot of weather data from my home. You can see it here in German: https://umwelt-olching.de

Is there a good simple Docker containerized weather app? I want to run it locally and Grafana Cloud plus a bunch of required plugins is way to heavy.

Today was the first day I ever used Grafana and what a terrible day its been. Grafana really needs to simplify their interface. There is no point having so many features when you can't even use the most basic functionality without pulling your hair out.

If there is going to be an editable JSON panel at least separate out the presentation and data layers. I don't even know where to start with the query editor, NewRelic allows autocomplete using the text editor so instead of clicking it will use context to figure out what to show. If it weren't were all the available dashboards and pretty looking things I would just use chronograf.

Try an OTel based monitoring tool like KloudMate or Honeycomb, maybe

To be honest, building a JSON feed consumer with a web frontend is a weekend project nowadays.

Hosting it costs money and requires more work but… deploying a simple app is something that you should have in your skillset anyway.

With a cloudflare tunnel, you can run it on a Mac Mini at home.

Reply on news.ycombinator.com