Solid Queue v1.6.0: Now Supporting Fiber Workers

Released on July 31 by rosa, the latest update to solid_queue (v1.6.0) introduces a highly anticipated shift in how jobs can be executed.

🚀 The Headline Feature: Fiber Execution Mode

Thanks to the contributions of @crmne, users are no longer limited to using a standard thread pool to handle multiple jobs per worker. Instead, you can now leverage fibers running on a single fiber reactor thread.

Why use Fibers?

This mode is specifically designed to optimize CPU-heavy I/O-bound workloads. A prime example would be applications making frequent calls to Large Language Models (LLMs), where the system spends most of its time waiting for a network response.

Mathematically, the efficiency gain in I/O-bound scenarios can be thought of as reducing the overhead of context switching: $\text{Efficiency} \propto \frac{\text{Concurrent Tasks}}{\text{Memory Overhead}}$

🛠️ Implementation Guide

To transition from threads to fibers, you must update your worker configuration.

Configuration Change: threads: 5 $\rightarrow$ fibers: 100

workers:
  - queues: " api* "
    fibers: 100
    polling_interval: 0.05

Prerequisites Checklist

Before this will function, ensure the following are completed:

• Add the Async gem to your project dependencies (as it powers the reactor under the hood).
• Enable fiber isolation within your Rails configuration: config.active_support.isolation_level = :fiber

---

🔄 Architectural Shift

The following diagram illustrates the transition from the traditional multi-threaded approach to the new fiber-based reactor model:

---

📝 Other Notable Changes

Beyond the fiber implementation, v1.6.0 includes several refinements:

Note: This release marks the first contributions from both @crmne and @wintan1418 to the project.

---

👥 Contributors & Community

The community response has been overwhelmingly positive, with dozens of reactions including ❤️, 👍, and 🚀.

Key Contributors to this Release:

• rosa
• crmne
• wintan1418

For a complete list of modifications, you can view the Full Changelog: v1.5.1...v1.6.0.