puma 6.6.0 → 7.1.0

data/lib/puma/cluster.rb

@@ -22,7 +22,8 @@ module Puma
       @workers = []
       @next_check = Time.now
 
-      @phased_restart = false
+      @worker_max = [] # keeps track of 'max' stat values
+      @pending_phased_restart = false
     end
 
     # Returns the list of cluster worker handles.
@@ -44,10 +45,14 @@ module Puma
       end
     end
 
-    def start_phased_restart
-      @events.fire_on_restart!
+    def start_phased_restart(refork = false)
+      @events.fire_before_restart!
       @phase += 1
-      log "- Starting phased worker restart, phase: #{@phase}"
+      if refork
+        log "- Starting worker refork, phase: #{@phase}"
+      else
+        log "- Starting phased worker restart, phase: #{@phase}"
+      end
 
       # Be sure to change the directory again before loading
       # the app. This way we can pick up new code.
@@ -166,7 +171,7 @@ module Puma
       (@workers.map(&:pid) - idle_timed_out_worker_pids).empty?
     end
 
-    def check_workers
+    def check_workers(refork = false)
       return if @next_check >= Time.now
 
       @next_check = Time.now + @options[:worker_check_interval]
@@ -184,7 +189,12 @@ module Puma
         w = @workers.find { |x| x.phase != @phase }
 
         if w
-          log "- Stopping #{w.pid} for phased upgrade..."
+          if refork
+            log "- Stopping #{w.pid} for refork..."
+          else
+            log "- Stopping #{w.pid} for phased upgrade..."
+          end
+
           unless w.term?
             w.term
             log "- #{w.signal} sent to #{w.pid}..."
@@ -228,7 +238,7 @@ module Puma
     def phased_restart(refork = false)
       return false if @options[:preload_app] && !refork
 
-      @phased_restart = true
+      @pending_phased_restart = refork ? :refork : true
       wakeup!
 
       true
@@ -258,11 +268,14 @@ module Puma
     end
 
     # Inside of a child process, this will return all zeroes, as @workers is only populated in
-    # the master process.
+    # the master process.  Calling this also resets stat 'max' values to zero.
     # @!attribute [r] stats
+    # @return [Hash]
+
     def stats
       old_worker_count = @workers.count { |w| w.phase != @phase }
       worker_status = @workers.map do |w|
+        w.reset_max
         {
           started_at: utc_iso8601(w.started_at),
           pid: w.pid,
@@ -273,7 +286,6 @@ module Puma
           last_status: w.last_status,
         }
       end
-
       {
         started_at: utc_iso8601(@started_at),
         workers: @workers.size,
@@ -342,7 +354,7 @@ module Puma
 
           stop_workers
           stop
-          @events.fire_on_stopped!
+          @events.fire_after_stopped!
           raise(SignalException, "SIGTERM") if @options[:raise_exception_on_sigterm]
           exit 0 # Clean exit, workers were stopped
         end
@@ -359,16 +371,12 @@ module Puma
 
       if preload?
         # Threads explicitly marked as fork safe will be ignored. Used in Rails,
-        # but may be used by anyone. Note that we need to explicit
-        # Process::Waiter check here because there's a bug in Ruby 2.6 and below
-        # where calling thread_variable_get on a Process::Waiter will segfault.
-        # We can drop that clause once those versions of Ruby are no longer
-        # supported.
-        fork_safe = ->(t) { !t.is_a?(Process::Waiter) && t.thread_variable_get(:fork_safe) }
+        # but may be used by anyone.
+        fork_safe = ->(t) { t.thread_variable_get(:fork_safe) }
 
         before = Thread.list.reject(&fork_safe)
 
-        log "*     Restarts: (\u2714) hot (\u2716) phased"
+        log "*     Restarts: (\u2714) hot (\u2716) phased (#{@options[:fork_worker] ? "\u2714" : "\u2716"}) refork"
         log "* Preloading application"
         load_and_bind
 
@@ -386,7 +394,7 @@ module Puma
           end
         end
       else
-        log "*     Restarts: (\u2714) hot (\u2714) phased"
+        log "*     Restarts: (\u2714) hot (\u2714) phased (#{@options[:fork_worker] ? "\u2714" : "\u2716"}) refork"
 
         unless @config.app_configured?
           error "No application configured, nothing to run"
@@ -413,6 +421,7 @@ module Puma
 
       log "Use Ctrl-C to stop"
 
+      warn_ruby_mn_threads
       single_worker_warning
 
       redirect_io
@@ -447,14 +456,18 @@ module Puma
               break
             end
 
-            if @phased_restart
-              start_phased_restart
-              @phased_restart = false
-              in_phased_restart = true
+            if @pending_phased_restart
+              start_phased_restart(@pending_phased_restart == :refork)
+
+              in_phased_restart = @pending_phased_restart
+              @pending_phased_restart = false
+
               workers_not_booted = @options[:workers]
+              # worker 0 is not restarted on refork
+              workers_not_booted -= 1 if in_phased_restart == :refork
             end
 
-            check_workers
+            check_workers(in_phased_restart == :refork)
 
             if read.wait_readable([0, @next_check - Time.now].max)
               req = read.read_nonblock(1)
@@ -497,7 +510,7 @@ module Puma
                   end
 
                   if !booted && @workers.none? {|worker| worker.last_status.empty?}
-                    @events.fire_on_booted!
+                    @events.fire_after_booted!
                     debug_loaded_extensions("Loaded Extensions - master:") if @log_writer.debug?
                     booted = true
                   end
@@ -514,7 +527,7 @@ module Puma
             end
 
             if in_phased_restart && workers_not_booted.zero?
-              @events.fire_on_booted!
+              @events.fire_after_booted!
               debug_loaded_extensions("Loaded Extensions - master:") if @log_writer.debug?
               in_phased_restart = false
             end
@@ -570,7 +583,9 @@ module Puma
           #    `Process.wait2(-1)` from detecting a terminated process: https://bugs.ruby-lang.org/issues/19837.
           # 2. When `fork_worker` is enabled, some worker may not be direct children,
           #    but grand children.  Because of this they won't be reaped by `Process.wait2(-1)`.
-          if reaped_children.delete(w.pid) || Process.wait(w.pid, Process::WNOHANG)
+          if (status = reaped_children.delete(w.pid) || Process.wait2(w.pid, Process::WNOHANG)&.last)
+            w.process_status = status
+            @config.run_hooks(:after_worker_shutdown, w, @log_writer)
             true
           else
             w.term if w.term?

data/lib/puma/cluster_accept_loop_delay.rb

@@ -0,0 +1,91 @@
+# frozen_string_literal: true
+
+module Puma
+  # Calculate a delay value for sleeping when running in clustered mode
+  #
+  # The main reason this is a class is so it can be unit tested independently.
+  # This makes modification easier in the future if we can encode properties of the
+  # delay into a test instead of relying on end-to-end testing only.
+  #
+  # This is an imprecise mechanism to address specific goals:
+  #
+  # - Evenly distribute requests across all workers at start
+  # - Evenly distribute CPU resources across all workers
+  #
+  # ## Goal: Distribute requests across workers at start
+  #
+  # There was a perf bug in Puma where one worker would wake up slightly before the rest and accept
+  # all the requests on the socket even though it didn't have enough resources to process all of them.
+  # This was originally fixed by never calling accept when a worker had more requests than threads
+  # already https://github.com/puma/puma/pull/3678/files/2736ebddb3fc8528e5150b5913fba251c37a8bf7#diff-a95f46e7ce116caddc9b9a9aa81004246d5210d5da5f4df90a818c780630166bL251-L291
+  #
+  # With the introduction of true keepalive support, there are two ways a request can come in:
+  # - A new request from a new client comes into the socket and it must be "accept"-ed
+  # - A keepalive request is served and the connection is retained. Another request is then accepted
+  #
+  # Ideally the server handles requests in the order they come in, and ideally it doesn't accept more requests than it can handle.
+  # These goals are contradictory, because when the server is at maximum capacity due to keepalive connections, it could mean we
+  # block all new requests, even if those came in before the new request on the older keepalive connection.
+  #
+  # ## Goal: Distribute CPU resources across all workers
+  #
+  # - This issue was opened https://github.com/puma/puma/issues/2078
+  #
+  # There are several entangled issues and it's not exactly clear what the root cause is, but the observable outcome
+  # was that performance was better with a small sleep, and that eventually became the default.
+  #
+  # An attempt to describe why this works is here: https://github.com/puma/puma/issues/2078#issuecomment-3287032470.
+  #
+  # Summarizing: The delay is for tuning the rate at which "accept" is called on the socket.
+  # Puma works by calling "accept" nonblock on the socket in a loop. When there are multiple workers
+  # (processes), they will "race" to accept a request at roughly the same rate. However, if one
+  # worker has all threads busy processing requests, then accepting a new request might "steal" it from
+  # a less busy worker. If a worker has no work to do, it should loop as fast as possible.
+  #
+  # ## Solution: Distribute requests across workers at start
+  #
+  # For now, both goals are framed as "load balancing" across workers (processes) and achieved through
+  # the same mechanism of sleeping longer to delay busier workers. Rather than the prior Puma 6.x
+  # and earlier behavior of using a binary on/off sleep value, we increase it an amount proportional
+  # to the load the server is under, capping the maximum delay to the scenario where all threads are busy
+  # and the todo list has reached a multiplier of the maximum number of threads.
+  #
+  # Private: API may change unexpectedly
+  class ClusterAcceptLoopDelay
+    attr_reader :max_delay
+
+    # Initialize happens once, `call` happens often. Perform global calculations here.
+    def initialize(
+      # Number of workers in the cluster
+      workers: ,
+      # Maximum delay in seconds i.e. 0.005 is 5 milliseconds
+      max_delay:
+    )
+      @on = max_delay > 0 && workers >= 2
+      @max_delay = max_delay.to_f
+
+      # Reach maximum delay when `max_threads * overload_multiplier` is reached in the system
+      @overload_multiplier = 25.0
+    end
+
+    def on?
+      @on
+    end
+
+    # We want the extreme values of this delay to be known (minimum and maximum) as well as
+    # a predictable curve between the two. i.e. no step functions or hard cliffs.
+    #
+    # Return value is always numeric. Returns 0 if there should be no delay.
+    def calculate(
+      # Number of threads working right now, plus number of requests in the todo list
+      busy_threads_plus_todo:,
+      # Maximum number of threads in the pool, note that the busy threads (alone) may go over this value at times
+      # if the pool needs to be reaped. The busy thread plus todo count may go over this value by a large amount.
+      max_threads:
+    )
+      max_value = @overload_multiplier * max_threads
+      # Approaches max delay when `busy_threads_plus_todo` approaches `max_value`
+      return max_delay * busy_threads_plus_todo.clamp(0, max_value) / max_value
+    end
+  end
+end

data/lib/puma/commonlogger.rb

@@ -29,13 +29,13 @@ module Puma
 
     CONTENT_LENGTH       = 'Content-Length' # should be lower case from app,
                                             # Util::HeaderHash allows mixed
-    HTTP_VERSION         = Const::HTTP_VERSION
     HTTP_X_FORWARDED_FOR = Const::HTTP_X_FORWARDED_FOR
     PATH_INFO            = Const::PATH_INFO
     QUERY_STRING         = Const::QUERY_STRING
     REMOTE_ADDR          = Const::REMOTE_ADDR
     REMOTE_USER          = 'REMOTE_USER'
     REQUEST_METHOD       = Const::REQUEST_METHOD
+    SERVER_PROTOCOL      = Const::SERVER_PROTOCOL
 
     def initialize(app, logger=nil)
       @app = app
@@ -70,7 +70,7 @@ module Puma
         env[REQUEST_METHOD],
         env[PATH_INFO],
         env[QUERY_STRING].empty? ? "" : "?#{env[QUERY_STRING]}",
-        env[HTTP_VERSION],
+        env[SERVER_PROTOCOL],
         now - began_at ]
 
       write(msg)
@@ -87,7 +87,7 @@ module Puma
         env[REQUEST_METHOD],
         env[PATH_INFO],
         env[QUERY_STRING].empty? ? "" : "?#{env[QUERY_STRING]}",
-        env[HTTP_VERSION],
+        env[SERVER_PROTOCOL],
         status.to_s[0..3],
         length,
         now - began_at ]

data/lib/puma/configuration.rb

@@ -3,6 +3,7 @@
 require_relative 'plugin'
 require_relative 'const'
 require_relative 'dsl'
+require_relative 'events'
 
 module Puma
   # A class used for storing "leveled" configuration options.
@@ -112,7 +113,7 @@ module Puma
   #     config = Configuration.new({}) do |user_config, file_config, default_config|
   #       user_config.port 3003
   #     end
-  #     config.load
+  #     config.clamp
   #     puts config.options[:port]
   #     # => 3003
   #
@@ -125,10 +126,13 @@ module Puma
   # is done because an environment variable may have been modified while loading
   # configuration files.
   class Configuration
+    class NotLoadedError < StandardError; end
+    class NotClampedError < StandardError; end
+
     DEFAULTS = {
       auto_trim_time: 30,
       binds: ['tcp://0.0.0.0:9292'.freeze],
-      clean_thread_locals: false,
+      fiber_per_request: !!ENV.fetch("PUMA_FIBER_PER_REQUEST", false),
       debug: false,
       enable_keep_alives: true,
       early_hints: nil,
@@ -140,19 +144,18 @@ module Puma
       io_selector_backend: :auto,
       log_requests: false,
       logger: STDOUT,
-      # How many requests to attempt inline before sending a client back to
-      # the reactor to be subject to normal ordering. The idea here is that
-      # we amortize the cost of going back to the reactor for a well behaved
-      # but very "greedy" client across 10 requests. This prevents a not
-      # well behaved client from monopolizing the thread forever.
-      max_fast_inline: 10,
+      # Limits how many requests a keep alive connection can make.
+      # The connection will be closed after it reaches `max_keep_alive`
+      # requests.
+      max_keep_alive: 999,
       max_threads: Puma.mri? ? 5 : 16,
       min_threads: 0,
       mode: :http,
       mutate_stdout_and_stderr_to_sync_on_write: true,
       out_of_band: [],
       # Number of seconds for another request within a persistent session.
-      persistent_timeout: 20,
+      persistent_timeout: 65, # PUMA_PERSISTENT_TIMEOUT
+      prune_bundler: false,
       queue_requests: true,
       rackup: 'config.ru'.freeze,
       raise_exception_on_sigterm: true,
@@ -176,24 +179,31 @@ module Puma
     def initialize(user_options={}, default_options = {}, env = ENV, &block)
       default_options = self.puma_default_options(env).merge(default_options)
 
-      @options     = UserFileDefaultOptions.new(user_options, default_options)
+      @_options    = UserFileDefaultOptions.new(user_options, default_options)
       @plugins     = PluginLoader.new
-      @user_dsl    = DSL.new(@options.user_options, self)
-      @file_dsl    = DSL.new(@options.file_options, self)
-      @default_dsl = DSL.new(@options.default_options, self)
-
-      if !@options[:prune_bundler]
-        default_options[:preload_app] = (@options[:workers] > 1) && Puma.forkable?
-      end
+      @events      = @_options[:events] || Events.new
+      @hooks       = {}
+      @user_dsl    = DSL.new(@_options.user_options, self)
+      @file_dsl    = DSL.new(@_options.file_options, self)
+      @default_dsl = DSL.new(@_options.default_options, self)
 
       @puma_bundler_pruned = env.key? 'PUMA_BUNDLER_PRUNED'
 
       if block
         configure(&block)
       end
+
+      @loaded = false
+      @clamped = false
     end
 
-    attr_reader :options, :plugins
+    attr_reader :plugins, :events, :hooks
+
+    def options
+      raise NotClampedError, "ensure clamp is called before accessing options" unless @clamped
+
+      @_options
+    end
 
     def configure
       yield @user_dsl, @file_dsl, @default_dsl
@@ -206,7 +216,7 @@ module Puma
     def initialize_copy(other)
       @conf        = nil
       @cli_options = nil
-      @options     = @options.dup
+      @_options     = @_options.dup
     end
 
     def flatten
@@ -214,7 +224,7 @@ module Puma
     end
 
     def flatten!
-      @options = @options.flatten
+      @_options = @_options.flatten
       self
     end
 
@@ -227,6 +237,7 @@ module Puma
     def puma_options_from_env(env = ENV)
       min = env['PUMA_MIN_THREADS'] || env['MIN_THREADS']
       max = env['PUMA_MAX_THREADS'] || env['MAX_THREADS']
+      persistent_timeout = env['PUMA_PERSISTENT_TIMEOUT']
       workers = if env['WEB_CONCURRENCY'] == 'auto'
         require_processor_counter
         ::Concurrent.available_processor_count
@@ -237,24 +248,27 @@ module Puma
       {
         min_threads: min && min != "" && Integer(min),
         max_threads: max && max != "" && Integer(max),
+        persistent_timeout: persistent_timeout && persistent_timeout != "" && Integer(persistent_timeout),
         workers: workers && workers != "" && Integer(workers),
         environment: env['APP_ENV'] || env['RACK_ENV'] || env['RAILS_ENV'],
       }
     end
 
     def load
+      @loaded = true
       config_files.each { |config_file| @file_dsl._load_from(config_file) }
-
-      @options
+      @_options
     end
 
     def config_files
-      files = @options.all_of(:config_files)
+      raise NotLoadedError, "ensure load is called before accessing config_files" unless @loaded
+
+      files = @_options.all_of(:config_files)
 
       return [] if files == ['-']
       return files if files.any?
 
-      first_default_file = %W(config/puma/#{@options[:environment]}.rb config/puma.rb).find do |f|
+      first_default_file = %W(config/puma/#{@_options[:environment]}.rb config/puma.rb).find do |f|
         File.exist?(f)
       end
 
@@ -262,9 +276,16 @@ module Puma
     end
 
     # Call once all configuration (included from rackup files)
-    # is loaded to flesh out any defaults
+    # is loaded to finalize defaults and lock in the configuration.
+    #
+    # This also calls load if it hasn't been called yet.
     def clamp
-      @options.finalize_values
+      load unless @loaded
+      set_conditional_default_options
+      @_options.finalize_values
+      @clamped = true
+      warn_hooks
+      options
     end
 
     # Injects the Configuration object into the env
@@ -283,11 +304,11 @@ module Puma
     # Indicate if there is a properly configured app
     #
     def app_configured?
-      @options[:app] || File.exist?(rackup)
+      options[:app] || File.exist?(rackup)
     end
 
     def rackup
-      @options[:rackup]
+      options[:rackup]
     end
 
     # Load the specified rackup file, pull options from
@@ -296,9 +317,9 @@ module Puma
     def app
       found = options[:app] || load_rackup
 
-      if @options[:log_requests]
+      if options[:log_requests]
         require_relative 'commonlogger'
-        logger = @options[:logger]
+        logger = options[:custom_logger] ? options[:custom_logger] : options[:logger]
         found = CommonLogger.new(found, logger)
       end
 
@@ -307,7 +328,7 @@ module Puma
 
     # Return which environment we're running in
     def environment
-      @options[:environment]
+      options[:environment]
     end
 
     def load_plugin(name)
@@ -315,18 +336,19 @@ module Puma
     end
 
     # @param key [:Symbol] hook to run
-    # @param arg [Launcher, Int] `:on_restart` passes Launcher
+    # @param arg [Launcher, Int] `:before_restart` passes Launcher
     #
     def run_hooks(key, arg, log_writer, hook_data = nil)
       log_writer.debug "Running #{key} hooks"
 
-      @options.all_of(key).each do |b|
+      options.all_of(key).each do |hook_options|
         begin
-          if Array === b
-            hook_data[b[1]] ||= Hash.new
-            b[0].call arg, hook_data[b[1]]
+          block = hook_options[:block]
+          if id = hook_options[:id]
+            hook_data[id] ||= Hash.new
+            block.call arg, hook_data[id]
           else
-            b.call arg
+            block.call arg
           end
         rescue => e
           log_writer.log "WARNING hook #{key} failed with exception (#{e.class}) #{e.message}"
@@ -336,7 +358,7 @@ module Puma
     end
 
     def final_options
-      @options.final_options
+      options.final_options
     end
 
     def self.temp_path
@@ -346,6 +368,12 @@ module Puma
       "#{Dir.tmpdir}/puma-status-#{t}-#{$$}"
     end
 
+    def self.random_token
+      require 'securerandom' unless defined?(SecureRandom)
+
+      SecureRandom.hex(16)
+    end
+
     private
 
     def require_processor_counter
@@ -386,22 +414,40 @@ module Puma
       rack_app, rack_options = rack_builder.parse_file(rackup)
       rack_options = rack_options || {}
 
-      @options.file_options.merge!(rack_options)
+      options.file_options.merge!(rack_options)
 
       config_ru_binds = []
       rack_options.each do |k, v|
         config_ru_binds << v if k.to_s.start_with?("bind")
       end
 
-      @options.file_options[:binds] = config_ru_binds unless config_ru_binds.empty?
+      options.file_options[:binds] = config_ru_binds unless config_ru_binds.empty?
 
       rack_app
     end
 
-    def self.random_token
-      require 'securerandom' unless defined?(SecureRandom)
+    def set_conditional_default_options
+      @_options.default_options[:preload_app] = !@_options[:prune_bundler] &&
+        (@_options[:workers] > 1) && Puma.forkable?
+    end
 
-      SecureRandom.hex(16)
+    def warn_hooks
+      return if options[:workers] > 0
+      return if options[:silence_fork_callback_warning]
+
+      log_writer = LogWriter.stdio
+      @hooks.each_key do |hook|
+        options.all_of(hook).each do |hook_options|
+          next unless hook_options[:cluster_only]
+
+          log_writer.log(<<~MSG.tr("\n", " "))
+            Warning: The code in the `#{hook}` block will not execute
+            in the current Puma configuration. The `#{hook}` block only
+            executes in Puma's cluster mode. To fix this, either remove the
+            `#{hook}` call or increase Puma's worker count above zero.
+          MSG
+        end
+      end
     end
   end
 end

data/lib/puma/const.rb

@@ -100,13 +100,11 @@ module Puma
   # too taxing on performance.
   module Const
 
-    PUMA_VERSION = VERSION = "6.6.0"
-    CODE_NAME = "Return to Forever"
+    PUMA_VERSION = VERSION = "7.1.0"
+    CODE_NAME = "Neon Witch"
 
     PUMA_SERVER_STRING = ["puma", PUMA_VERSION, CODE_NAME].join(" ").freeze
 
-    FAST_TRACK_KA_TIMEOUT = 0.2
-
     # How long to wait when getting some write blocking on the socket when
     # sending data back
     WRITE_TIMEOUT = 10
@@ -125,9 +123,9 @@ module Puma
       # Indicate that we couldn't parse the request
       400 => "HTTP/1.1 400 Bad Request\r\n\r\n",
       # The standard empty 404 response for bad requests.  Use Error4040Handler for custom stuff.
-      404 => "HTTP/1.1 404 Not Found\r\nConnection: close\r\n\r\n",
+      404 => "HTTP/1.1 404 Not Found\r\nconnection: close\r\n\r\n",
       # The standard empty 408 response for requests that timed out.
-      408 => "HTTP/1.1 408 Request Timeout\r\nConnection: close\r\n\r\n",
+      408 => "HTTP/1.1 408 Request Timeout\r\nconnection: close\r\n\r\n",
       # Indicate that there was an internal error, obviously.
       500 => "HTTP/1.1 500 Internal Server Error\r\n\r\n",
       # Incorrect or invalid header value
@@ -230,6 +228,7 @@ module Puma
     RACK_INPUT = "rack.input"
     RACK_URL_SCHEME = "rack.url_scheme"
     RACK_AFTER_REPLY = "rack.after_reply"
+    RACK_RESPONSE_FINISHED = "rack.response_finished"
     PUMA_SOCKET = "puma.socket"
     PUMA_CONFIG = "puma.config"
     PUMA_PEERCERT = "puma.peercert"
@@ -252,14 +251,14 @@ module Puma
     KEEP_ALIVE = "keep-alive"
 
     CONTENT_LENGTH2 = "content-length"
-    CONTENT_LENGTH_S = "Content-Length: "
+    CONTENT_LENGTH_S = "content-length: "
     TRANSFER_ENCODING = "transfer-encoding"
     TRANSFER_ENCODING2 = "HTTP_TRANSFER_ENCODING"
 
-    CONNECTION_CLOSE = "Connection: close\r\n"
-    CONNECTION_KEEP_ALIVE = "Connection: Keep-Alive\r\n"
+    CONNECTION_CLOSE = "connection: close\r\n"
+    CONNECTION_KEEP_ALIVE = "connection: keep-alive\r\n"
 
-    TRANSFER_ENCODING_CHUNKED = "Transfer-Encoding: chunked\r\n"
+    TRANSFER_ENCODING_CHUNKED = "transfer-encoding: chunked\r\n"
     CLOSE_CHUNKED = "0\r\n\r\n"
 
     CHUNKED = "chunked"

data/lib/puma/control_cli.rb

@@ -124,11 +124,15 @@ module Puma
         end
 
         if @config_file
+          # needed because neither `Puma::CLI` or `Puma::Server` are loaded
+          require_relative '../puma'
+
           require_relative 'configuration'
           require_relative 'log_writer'
 
           config = Puma::Configuration.new({ config_files: [@config_file] }, {} , env)
-          config.load
+          config.clamp
+
           @state              ||= config.options[:state]
           @control_url        ||= config.options[:control_url]
           @control_auth_token ||= config.options[:control_auth_token]
@@ -248,7 +252,7 @@ module Puma
           @stdout.flush unless @stdout.sync
           return
         elsif sig.start_with? 'SIG'
-          if Signal.list.key? sig.sub(/\ASIG/, '')
+          if Signal.list.key? sig.delete_prefix('SIG')
             Process.kill sig, @pid
           else
             raise "Signal '#{sig}' not available'"