Using Docker with RTL-SDR for ADS-B

I bought my first RTL-SDR receiver a little over a year ago and was soon up and running with FlightAware's dump1090. Shortly after that I signed up with Flightradar24, downloaded and installed the x86_64 package and started contributing. ADS-B Exchange has a script to bootstrap Linux installation and that too made it easy to get up and running. It was only when I created a second ADS-B feeder site at a new location that I appreciated how onerous maintaining all the installs and configuration could become.

Somewhat ironically I could not contribute to FlightAware either since PiAware, as the name implies, only supports the Raspberry Pi and not x86_64 Linux. I was similarly disappointed to find out that RadarBox only has Pi and Windows feeders.

While investigating various solutions I stumbled across the work of Mike and his fellow SDR Enthusiasts. They have done the hard yards and bundled the RadarBox rbfeeder and other clients into docker images supporting x86_64, arm32 and arm64 architectures - either by compiling native binaries or emulation through qemu. The docker-compose examples made it all too easy replace to my manual installs with docker-flightradar24 and docker-adsbexchange then ditch dump1090 for docker-readsb.

This was swiftly followed by the addition of docker-radarbox, docker-piaware and docker-planefinder feeders. Maintaining just a single docker-compose.yaml in each location couldn't be more simple.

Deployments have been quite robust with very little intervention required. The PiAware client occasionally loses connectivity to FlightAware and requires a restart. The connection failure is picked up by the healthcheck.sh script so I've added Will Farrell's docker-autoheal into the mix to automate the restarts.

posted by James Gemmell on Sat, 06 May 2023 at 20:47 | permalink | tags: adsb, docker, rtlsdr

Loggers are not constants

Loggers are not constants, they are mutable objects.

I am currently working on a legacy Java/Spring codebase littered with calls to LOG and LOGGER, proudly declared as static final. There appear to be sound justifications for adopting this convention (besides avoiding Checkstyle complaints) but sometimes it is useful to reflect on how we ended up here.

Loggers have traditionally been declared static as their construction was considered expensive and therefore justified a singleton status. Times have changed along with coding conventions.

Most of the applications in this particular codebase are of the vanilla Spring application variety, namely singleton services processing value objects on multiple threads.

Loggers used by service singletons are obviously singletons themselves and clearly don't warrant this premature optimization. Loggers in value objects shouldn't be there.

A grey area exists for short lived domain objects, say some business logic encapsulated in a function that would log input arguments and the result. The solution is to have the object factory inject the logger at object construction time. Dependency injection ensures that logging is just another service, singleton or not.

A further argument against using static final loggers is the increased use of semantic logging. The implications are that if, in order to unit test the logging feature, the object under test requires the static mocking capabilities of PowerMock then you're doing dependency injection incorrectly or not at all.

In another example, this practice was followed to the absurd extreme with SimpleDateFormat. Not only is this not a constant but the methods are not thread-safe. Declaring it as such merely increases the likelihood of multiple threads executing the unsafe methods concurrently, even more so when running in a multi-application container.

The Google Java Style guide suggests some reasonable conventions for constant names.

Forgetting why we started doing it in the first place and blindly following the static final Logger convention is, at best, premature optimization and, at worst, a shining example of donning the coconut headphones.

posted by James Gemmell on Tue, 24 Nov 2015 at 22:37 | permalink | tags: java, spring

A review of Packt Publishing's Spring Data

Packt Publishing's Spring Data is their latest offering in the now familiar cookbook format covering the Spring Data support for JPA and Redis. The Kindle version was well laid out, which can be difficult for technical texts.

The cookbook uses a simple domain model for contact data to demonstrate the creation of a CRUD database and then steps the reader through creating operations using Native SQL, Java Persistence Query Language (JPQL) and the Criteria API. What I found most useful was being able to compare and contrast the pros and cons of each approach.

This section may be less suited to the novice it assumes a working knowledge of JPA. The cookbook could not be treated as a JPA tutorial or reference as other texts have greater coverage using more complex data models.

The chapters on Redis start with step-by-step instructions on configuring the connection to a Redis service using the Jedis, JRedis, RJC and SRP connectors. It then proceeds through to the implementing a CRUD application for the contact data and adding publish/subscribe notifications for updates.

By far the most beneficial thing I found in these chapters was how to use Spring Data Redis to add transparent caching support to a JPA repository.

Spring Data can be bought from Packt Publishing and Amazon.com.

posted by James Gemmell on Sat, 19 Jan 2013 at 13:46 | permalink | tags: java, jpa, redis, spring

A review of the Spring Web Services 2 Cookbook

The Cookbook fills a long-standing vacancy on the Spring bookshelf and plays a useful complementary role to the online Spring Web Services reference documentation. Considering the volume of publications, Spring WS is by far the poorer cousin of the Spring core, persistence, MVC, batch and integration lineup.

Over the last two years I've used Spring Web Services extensively in the implementation of a multi-operation web service facade to a foreign exchange dealing system in a large Australian bank.

I wish I'd had this book when I started the project. The WS examples from SpringSource are great, but they don't come close to demonstrating the richness of the platform's capabilities. The Cookbook's recipes and the accompanying sample projects cover just about every combination of transport and object-XML mapping that you're likely to use (SOAP over HTTP and JMS with JAXB2 and XMLBeans) and some of the more esoteric (SOAP over e-mail and XMPP with JiBX and MooseXML.)

Both XWSS and WSS4J for security are dealt with comprehensively. I was pleasantly surprised to find a chapter dealing with JAX-WS and Apache CXF as well as creating RESTful services, using Spring MVC. Testing using Spring's MockWebService as well as TCPMon and soapUI also get their own much needed chapter.

As you would expect of any cookbook, this one doesn't read easily cover to cover, and the recipes can appear repetitive at times. In-depth coverage has been sacrificed in favour of breadth. However, in my opinion, this is the Cookbook's strongest selling point. The wide range of subject matter allows the reader to easily explore the featured technologies and make educated evaluations and comparisons.

I find the Spring Web Services 2 Cookbook a worthwhile addition to my bookshelf. The book can bought from Packt Publishing and Amazon.com.

posted by James Gemmell on Tue, 15 May 2012 at 21:00 | permalink | tags: java, soap, spring

Using What instead of Why to report errors

During peer code reviews I have sometimes observed that there is a preference for programmers to interpret errors or exceptions as part of the error handling process. Instead of reporting what caused the error, an interpretation is applied and why the error occurred is reported instead.

As an example, an exception such as "connection failed" is reported as "the server is down". This is, quite clearly, a naive interpretation. There may be many other reasons as to why the "connection failed"; the connection credentials may be incorrect, there may be a network fault, the service application may not be running or a solar flare may be affecting your Wifi. Under these circumstances, all you can safely assume about the situation is that, well, the "connection failed".

The example above seem obvious but it becomes even easier to make the mistake when reporting business exceptions from a web service. When required to handle a particular error, the programmer often has to rummage through a toolbox of available business exception codes and apply the one that fits best. More often than not it doesn't.

The importance of getting this right may only become apparent after that prolonged phone call with the irate user who insists that your "server is down" when you know perfectly well that it isn't.

posted by James Gemmell on Wed, 01 Jun 2011 at 21:14 | permalink | tags: java, soap, spring