Issue #19122 has been updated by ioquatix (Samuel Williams). On the latest Linux kernel, the constants have changed: ``` MADV_DONTNEED = 4 MADV_FREE = 8 ``` I did a quick comparison, creating 10,000 fibers x10 times in a loop: ``` samuel@aiko ~/D/s/memory-leak (main)> time RUBY_SHARED_FIBER_POOL_FREE_STACKS=8 bundle exec test.rb >/dev/null <main>: warning: Setting RUBY_SHARED_FIBER_POOL_FREE_STACKS to a value greater than 1 is operating system specific, and may cause crashes. ________________________________________________________ Executed in 841.07 millis fish external usr time 373.88 millis 441.00 micros 373.44 millis sys time 463.22 millis 168.00 micros 463.05 millis samuel@aiko ~/D/s/memory-leak (main)> time RUBY_SHARED_FIBER_POOL_FREE_STACKS=16 bundle exec test.rb >/dev/null <main>: warning: Setting RUBY_SHARED_FIBER_POOL_FREE_STACKS to a value greater than 1 is operating system specific, and may cause crashes. ________________________________________________________ Executed in 582.77 millis fish external usr time 357.66 millis 420.00 micros 357.24 millis sys time 220.90 millis 164.00 micros 220.74 millis samuel@aiko ~/D/s/memory-leak (main)> time RUBY_SHARED_FIBER_POOL_FREE_STACKS=0 bundle exec test.rb >/dev/null ________________________________________________________ Executed in 444.61 millis fish external usr time 312.92 millis 503.00 micros 312.41 millis sys time 128.99 millis 0.00 micros 128.99 millis ``` Also memory usage as we expected: ``` samuel@aiko ~/D/s/memory-leak (main)> RUBY_SHARED_FIBER_POOL_FREE_STACKS=16 bundle exec test.rb <main>: warning: Setting RUBY_SHARED_FIBER_POOL_FREE_STACKS to a value greater than 1 is operating system specific, and may cause crashes. PID: 70235 Memory usage: 150.82 MB Clearing fibers and garbage collecting... Memory usage: 150.82 MB samuel@aiko ~/D/s/memory-leak (main)> RUBY_SHARED_FIBER_POOL_FREE_STACKS=8 bundle exec test.rb <main>: warning: Setting RUBY_SHARED_FIBER_POOL_FREE_STACKS to a value greater than 1 is operating system specific, and may cause crashes. PID: 70279 Memory usage: 151.63 MB Clearing fibers and garbage collecting... Memory usage: 73.63 MB ``` Using the following test script: ``` #!/usr/bin/env ruby require "memory/leak/system" def print_memory_usage size = Memory::Leak::System.memory_usage(Process.pid) units = %w(B KB MB GB TB) unit = 0 while size > 1024 && unit < units.size size /= 1024.0 unit += 1 end puts "Memory usage: %.2f %s" % [size, units[unit]] end puts "PID: #{Process.pid}" 10.times do |i| puts "Iteration: #{i}" # Create 10,000 fibers: fibers = 10_000.times.map do Fiber.new do Fiber.yield Fiber.current end.resume end # Print memory usage: print_memory_usage puts "Clearing fibers and garbage collecting..." # Clear fibers: fibers.clear GC.start # Print memory usage: print_memory_usage end ``` ---------------------------------------- Misc #19122: Use MADV_DONTNEED instead of MADV_FREE when freeing a Fiber's stack https://bugs.ruby-lang.org/issues/19122#change-112186 * Author: smcgivern (Sean McGivern) * Status: Assigned * Assignee: ioquatix (Samuel Williams) ---------------------------------------- I'd like to propose that Ruby stops using MADV_FREE when freeing a Fiber's stack, and switches to using MADV_DONTNEED even when MADV_FREE is supported. MADV_FREE is used in one place in the Ruby codebase, when freeing the stack of a freed Fiber: https://git.ruby-lang.org/ruby.git/tree/cont.c#n683 The comment for `fiber_pool_stack_free` says: ```c // We advise the operating system that the stack memory pages are no longer being used. // This introduce some performance overhead but allows system to relaim memory when there is pressure. ``` Where possible (i.e. on Linux 4.5 and later), `fiber_pool_stack_free` uses `MADV_FREE` over `MADV_DONTNEED`. This has the side effect that memory statistics such as RSS will not reduce until and unless the OS actually reclaims that memory. If that doesn't happen, then the reported memory usage via RSS will be much higher than the 'real' memory usage. If this was pervasive throughtout the Ruby codebase then that would be one thing, but currently this is just for Fiber. This means that: 1. A program that doesn't use Fiber will have somewhat reliable RSS statistics on recent Linux. 2. A program that heavily uses Fiber (such as something using Async::HTTP) will see an inflated RSS statistic. Go made a similar change to the one I'm proposing here for similar reasons: https://github.com/golang/go/issues/42330
While `MADV_FREE` is somewhat faster than `MADV_DONTNEED`, it doesn't affect many of the statistics that `MADV_DONTNEED` does until the memory is actually reclaimed. This generally leads to poor user experience, like confusing stats in `top` and other monitoring tools; and bad integration with management systems that respond to memory usage. [...] I propose we change the default to prefer `MADV_DONTNEED` over `MADV_FREE`, to favor user-friendliness and minimal surprise over performance. I think it's become clear that Linux's implementation of `MADV_FREE` ultimately doesn't meet our needs.
As an aside, MADV_FREE was not used in Ruby 3.1 (https://bugs.ruby-lang.org/issues/19101), and I haven't found any bugs filed about this behaviour other than that one. -- https://bugs.ruby-lang.org/