Ruby Garbage Collection Tuning for Older Versions

By USEO Engineering Team November 5, 2025

BLUF (Bottom Line Up Front): Default Ruby Garbage Collection (GC) settings are optimized for small scripts, not enterprise web applications. By tuning environment variables like RUBY_GC_HEAP_INIT_SLOTS, you can reduce the frequency of major GC pauses and significantly improve response times in older Ruby versions (2.5 - 2.7).

Phase 1: The Cost of Default GC

Ruby uses a mark-and-sweep garbage collector. A “minor GC” cleans up short-lived objects quickly. A “major GC” scans the entire heap, stopping the world and halting application execution.

Synthetic Engineering Context: The GC Bottleneck

If you profile a slow legacy request, you might find that 30% of the execution time is spent allocating memory and triggering GC.

# Profiling output (GC.stat)
{
  :count=>145, 
  :heap_allocated_pages=>4500, 
  :heap_sorted_length=>4500, 
  :heap_allocatable_pages=>0, 
  :heap_available_slots=>1834520, 
  :heap_live_slots=>1830000, 
  :heap_free_slots=>4520, 
  :major_gc_count=>42, # High number of stop-the-world pauses
  :minor_gc_count=>103
}

The low heap_free_slots indicates Ruby is constantly out of memory slots, forcing frequent major GC runs.

Phase 2: Tuning the Environment

Instead of letting Ruby slowly expand its heap during runtime (which causes latency spikes), you can pre-allocate memory.

Execution: Setting ENV Variables

Add these variables to your production environment (e.g., in your Dockerfile, Heroku config, or systemd service).

# Pre-allocate a large number of slots (e.g., 1 million) to avoid expansion overhead
export RUBY_GC_HEAP_INIT_SLOTS=1000000

# Increase the threshold before a major GC is forced
export RUBY_GC_HEAP_FREE_SLOTS=500000

# Allow the heap to grow faster when it does need to expand
export RUBY_GC_HEAP_GROWTH_FACTOR=1.25

# For Ruby 2.7+, tune the malloc limits to reduce minor GC pressure
export RUBY_GC_MALLOC_LIMIT=64000000
export RUBY_GC_MALLOC_LIMIT_MAX=128000000

By pre-allocating RUBY_GC_HEAP_INIT_SLOTS, the application boots with enough memory to handle typical requests without immediately halting to ask the OS for more RAM.

Phase 3: Next Steps & Risk Mitigation

Tuning GC is a trade-off: you are trading RAM for CPU cycles (lower latency). If you set these values too high on a server with limited memory, you will trigger swap thrashing or the OOM killer.

Need Help Stabilizing Your Legacy App? GC tuning requires precise monitoring and load testing. Our team at USEO analyzes memory profiles to find the exact GC parameters suited for your specific workload.