Memory Dump Analyzer

README.md

This is a adaptation of some of the ObjectSpace.dump_all analyisis scripts found in Sam Saffron's article on "Debugging memory leaks in Ruby":

https://samsaffron.com/archive/2015/03/31/debugging-memory-leaks-in-ruby

The input files are also generated in a similar fashion that is described in the blog post:

io=File.open("/tmp/my_dump", "w")
ObjectSpace.dump_all(output: io); 
io.close

And assumes that you are currently running with ObjectSpace.trace_object_allocations_start (see the post for more details).

That said, this script supports working with gzipped files, which is highly recommend that you zip up the output files that you do dump, as they are quite large (500MB a piece in my case).

Usage:

There are two modes for the script:

Summary mode

Without any flag, the script will keep track of all object allocations and print out the counts, grouped by GC generation.

Specific generation analysis

The -g/--generation flag allows for displaying the more versbose output described in the post, which displays the counts of the currently allocated objects for a specific generation, grouped by line and line number.

memory_dump_analyzer.rb

require 'optparse'

options = { :generation => nil }

OptionParser.new do |opt|
  opt.banner = "Usage: #{File.basename $0} [options] DUMPFILE"

  opt.separator ""
  opt.separator "Parses the given DUMPFILE (output from ObjectSpace.dump_all)"
  opt.separator "and either gives the generation statisics or the specific"
  opt.separator "memory locations of the objects in a given generation."
  opt.separator ""
  opt.separator "Files can either be the raw json data, or their gzipped"
  opt.separator "equivalents, and the parser will figure out how to handle"
  opt.separator "them accordingly."
  opt.separator ""
  opt.separator "Options"

  opt.on("-gNUM", "--generation=NUM", Integer, "Method to analyze") do |val|
    options[:generation] = val.to_i
  end

  opt.on("-h",    "--help",                    "Show this message") do
    puts opt
    exit
  end
end.parse!

require 'json'
require 'zlib'

class Analyzer
  def initialize filename, options = {}
    @filename   = filename
    @generation = options[:generation]
    @data       = []
  end

  def analyze
    io_klass.open(@filename) do |f|
      f.each_line do |line|
        parsed = JSON.parse(line)
        @data << parsed if @generation.nil? || parsed["generation"] == @generation
      end
    end

    @generation ? print_for_speicific_generation : print_generation_count
  end

  private

  def print_generation_count
    @data.group_by { |row| row["generation"] }
         .sort     { |a,b| a[0].to_i <=> b[0].to_i }
         .each     { |k,v| puts "generation #{k} objects #{v.count}" }
  end

  def print_for_speicific_generation
    puts "Generation #{@generation}:\n"

    @data.group_by { |row| "#{row["file"]}:#{row["line"]}" }
         .sort     { |a,b| b[1].count <=> a[1].count }
         .each     { |k,v| puts "#{k} * #{v.count}" }
  end

  def io_klass
    File.extname(@filename) == ".gz" ? Zlib::GzipReader : File
  end
end

Analyzer.new(ARGV[0], options).analyze