Connect with attendees on Slack in #grafanacon channel
About this session
Speakers
Speaker 1 (00:05): Hey everyone. I’m Mateusz and I’d like to tell you a short story. It all starts with this device. So when I was a teenager, the USB sticks that allowed you to watch television were all the rage among some of the nerdy folks. As TV use started to shrink, it was discovered that many of these USB sticks have a chipset that’s surprisingly versatile, and that’s how low cost Software Defined Receivers started to become easily available. About a month ago. You could buy one that would do wonders for as low as about 15 bucks. I was always curious about aircraft and radio reception. When I discovered that people are building receivers based on the TV tuners and that it’s fully legal to receive these signals as they are public, I knew that I want one, so I used the Raspberry Pi and I was pretty amazed about the results.
(00:55): With a fairly inexpensive setup of about $70, I could receive signals from planes hundreds of kilometers away. For those that are here, it’s hundreds of miles away as well. I set it up and kept that running for about a year until the memory card broke, and later I also found out that you shouldn’t connect the power input into the audio output. That breaks the Raspberry Pi, it seems. Alright. Sometime ago I came back to the topic and also found out that it’s possible to receive ships using the same type of a setup. Why do people do this? Well, most of them probably just want to see what’s around and get a free premium account at one of the tracking websites for feeding the data back to them. For some, it’s also pretty fascinating to see what’s around you, but there are uses that go beyond the personal gain.
(01:45): You might have seen one of them in the media in the recent months when there were news of the undersea cable cuts in the Baltic Sea region. One of the points was the data from hobbyist AIS receivers who spotted a suspicious turn of one of the ships. The data from these receivers are also sometimes used in the air crash investigations. I wanted to set up a home server, so I recently bought a mini PC for this purpose. I wanted it to host the radio receivers for planes, ships as well as some application monitoring. I also thought I need a mobile one, so I also bought a Raspberry Pi and this specific device. So at work I’m writing charms that wrap observability tools such as Grafana, Tempo, Prometheus or Mimir as part of the Canonical Observability Stack. So I obviously wanted to dogfood the thing that I’m implementing, and there is a saying that you already heard from this stage today that if you own a hammer, everything looks like a nail.
(02:44): So I thought that it might be an interesting experiment to try to see if I can show the data from the radio reception using Prometheus and Grafana and a bit of custom exporters. So for planes, I adjusted an existing open source Prometheus exporter for one of the most common tools, which is called dump1090 to expose metrics for each of the planes. For ships, I gutted the same codebase to parse a JSON file with currently received ships that’s provided by the AIS-catcher. I also looked at ways to introduce some other radio communication using Prometheus exporters. All of them were scraped by Prometheus deployed using Canonical Juju on my home server together with Grafana. The main problem is the Prometheus metric format. So if I want to show something that has a latitude, longitude, and potentially a few other fast changing characteristics, I cannot use the multiple values of a single metric.
(03:38): I probably also don’t want to use labels because of the cardinality, and in retrospect, I probably should have used InfluxDB, but I wanted to use Prometheus. It’s always for fun. So I thought let’s create separate metrics for latitude, longitude, altitude, and potentially one that would have all the metadata together with the last seen time. So I needed to write or adapt a few existing Prometheus exporters out there. That, together with a few quite simple manipulations using the Transform tab in Grafana and a Geomap component has worked surprisingly well and as I realized, the same approach is pretty generic. It can be used to show: aircraft, obviously. Meshtastic devices. I didn’t know anything about Meshtastic until about six months ago when I heard about a decentralized mesh radio system that is using extremely cheap devices, saying about $15 in order to be able to send text and position messages in a mesh.
(04:36): It turns out that there are thousands of devices active all over the world, and I was able to receive messages from dozens of kilometers away. The same approach can also be used for ships or virtually anything that has a latitude, longitude, and a presence. There is something interesting in this picture. I will get back to it later. Since I had access to all this data, I thought that I’d like to focus more on the trends and not on the real time state that I could get from the tracking apps. So I created a map of all the last locations of planes that have been recorded by my receivers. It was quite impressive to see that in some cases it’s literally hundreds of kilometers. If you are fortunate and have access to a place where you can put even a very cheap antenna outside of the window with lots of unobstructed horizon, you can receive signals that can be countries away.
(05:25): If that’s not the case, but you keep the antenna close to the window inside, you could still get a few dozen kilometers. In some places, this would be enough to still receive quite a lot. It was also quite curious to see which countries of origin were the most often seen by my receivers. The codes that you see here are airplane registration prefixes, so it’s not the same as ISO. SP in this case means Poland. As most of my data was coming from Gdańsk, which is roughly 90 minutes by car from the Russian border there was one especially curious omission in the data. I did see the Russian aircraft, but they were not recognized as Russian by my exporter. It looks like it was related to the database that I was using for the planes that either did not contain their data or purged it in 2022, as this is when the last database update was done.
(06:14): In some cases I was able to see sudden shifts on some planes’ location in the area of Gdańsk. It’s a very good example of GPS jamming that is happening in this area and has intensified in the recent years. You don’t see it on Flightradar24 because they filter out the noise and approximate, but for instance, ADSB Exchange, as you can see in the picture on the right side, doesn’t do that. An overview in Grafana helped me grasp the scale of the issue much better. An example that you can see here on the left is a Polish Airlines plane that’s apparently near Moscow. At the time it was descending to land in Gdańsk about a thousand kilometers away. I also promised I will get back to this dashboard. As you can see for ships, the range is much lower. Yet, I received a couple of signals that were apparently from the Russian coast.
(07:00): It prompted me to check out my other maps that were offered by the tool that was gathering the data and there I saw a bunch of ships that were seemingly stuck on the shore right next to a small enclosure. If you are familiar with OpenStreetMap, you can notice that the military structures are marked with the red color. This one that we see here even has an antenna in it. The ships also mysteriously teleported between there and Gdańsk Bay within minutes a few times. I learned a lot while doing all these experiments. All that wouldn’t be anywhere new that easy without the modern observability tooling such as Grafana and Prometheus. And who said that observability is just for applications? Right, thank you. In case you want to discuss more, you can find me at the Ask The Experts booth after the presentations. Thank you so much.