Fix your Bootsnap
A deep dive into what Bootsnap caches, how its load-path and compile caches work inside, and why having it in your Gemfile does not mean your production processes benefit from it.
How a Rails 7.2.1.1 upgrade caused composite-key queries on large datasets to blow the stack, and what changed inside Arel's OR-node tree to make it happen.
Rails 8.1's Active Job Continuations lets long-running jobs be interrupted and resumed from where they left off, so previously completed work is never lost.
An introduction to queueing theory through the lens of Rails background job systems.
An overview of how Active Record pools and reuses database connections within each process, how the Reaper cleans up idle connections, and how to tune the pool to avoid ConnectionTimeoutError as a Rails application scales.
A practical walkthrough of finding the ideal Puma thread count by measuring a Rails app's I/O time in production with GVL instrumentation and the perfm gem.
Amdahl's Law shows that the serial portion of a workload limits how much speedup extra Puma threads can deliver.
An explanation of Ruby's Global VM Lock, which lets only one thread per process run Ruby code at a time, and how it shapes the way Rails scales.
A look at how Puma handles incoming requests, tracing their path through the socket backlog, reactor, and thread pool before they reach Rails, and explaining why a default process with three threads serves three requests at once.