The Architecture of Oban Web

Oban Web, the live dashboard for Oban, is fully open source! Now we can properly share the internals of a complex, battle tested, real-world, and formerly commercial LiveView app.

After a tour of foundational elements such as page structure and the table system, we'll delve into exotic features like real-time charts, integrated combo boxes, autocompletion, complex queries that don't grind the database to a halt, and more.

Additionally, we'll investigate Oban Met, an in-memory, compacting, time series, distributed database purpose built to power the Oban Web dashboard. Come and see what is possible with the spectactular combination of LiveView, Oban, and OTP.

The Oban Murders

Join us on a thrilling detective tale as we investigate three mysterious Oban failures that haunted users and stumped our finest agents. Meet the victims, follow the clues, and uncover the motivations behind these job deaths. Why did they die, and is it true that they came back to life?

We'll utilize all the forensic tools at our disposal, including:

* Logging review to retrace the job's footsteps
* Telemetry investigation to decode internal patterns
* Query profiling to uncover performance abnormalities
* Dashboard analysis to manipulate the clues

Together, let's crack the deepest Oban mysteries and unravel the web of deceit. Will we uncover the killer(s)? The answer lies within the data.

https://www.youtube.com/watch?v=WjxU2fBk3cY

Union of the Snake

How do you port a highly concurrent framework from Elixir—a platform with arguably the best concurrency story in the industry—to Python, a language with a flawed one at best? In this talk, we'll show how we tried, and along the way managed to prove "Virding's First Rule of Programming".

This goes deeper than wrestling with mutability or the emotional burden of imperative loops. We'll dig into the techniques we used to work around Python's concurrency limitations in order to implement a BEAM-inspired architecture, including:

* Supervising autonomous objects
* Monitoring object lifecycles
* Replicating GenServers with async loops
* Communicating without native distribution

You'll have a renewed appreciation for Elixir: an expressive syntax, fantastic tooling, and a community that converges on a shared set of fundamental abstractions. Nothing clarifies the strengths of a system quite like trying to rebuild it elsewhere.

Come and learn how we blended Elixir's sensibilities with Python's semantics to build the best async, concurrent, distributed background job framework we could manage—and what we learned in the process.

https://oban.pro/articles/introducing-oban-python
https://oban.pro/docs/py/0.5.0/index.html

Scaling Oban Applications

Perhaps you've heard the common refrain that relational database backed queues don't scale. Well, times have changed, databases have improved, and for Elixir the BEAM gives us a leg up. Every day real-world applications use Oban to process tens-of-millions of a jobs a day across clusters of all sizes.

Come learn about the common pain-points impacting the growth of Oban applications and how you can avoid them. Together, we'll identify the underlying causes of common problems and explore step-by-step strategies to mitigate them (spoiler, it's bloat).

Armed with a better understanding of how Oban operates at scale, you can prevent common issues before you encounter them and plan for the future as your application booms.

This talk was given by Shannon & Parker Selbert in Lisbon, Portugal at ElixirConf EU; April 2024. 40 mins in duration, the video of this talk: https://www.youtube.com/watch?v=1v52vTYlblM