puma 4.3.12 → 6.0.2

data/lib/puma/cluster.rb

@@ -1,35 +1,34 @@
 # frozen_string_literal: true
 
-require 'puma/runner'
-require 'puma/util'
-require 'puma/plugin'
-
-require 'time'
+require_relative 'runner'
+require_relative 'util'
+require_relative 'plugin'
+require_relative 'cluster/worker_handle'
+require_relative 'cluster/worker'
 
 module Puma
   # This class is instantiated by the `Puma::Launcher` and used
   # to boot and serve a Ruby application when puma "workers" are needed
   # i.e. when using multi-processes. For example `$ puma -w 5`
   #
-  # At the core of this class is running an instance of `Puma::Server` which
-  # gets created via the `start_server` method from the `Puma::Runner` class
-  # that this inherits from.
-  #
   # An instance of this class will spawn the number of processes passed in
   # via the `spawn_workers` method call. Each worker will have it's own
   # instance of a `Puma::Server`.
   class Cluster < Runner
-    def initialize(cli, events)
-      super cli, events
+    def initialize(launcher)
+      super(launcher)
 
       @phase = 0
       @workers = []
-      @next_check = nil
+      @next_check = Time.now
 
-      @phased_state = :idle
       @phased_restart = false
     end
 
+    # Returns the list of cluster worker handles.
+    # @return [Array<Puma::Cluster::WorkerHandle>]
+    attr_reader :workers
+
     def stop_workers
       log "- Gracefully shutting down workers..."
       @workers.each { |x| x.term }
@@ -37,7 +36,7 @@ module Puma
       begin
         loop do
           wait_workers
-          break if @workers.empty?
+          break if @workers.reject {|w| w.pid.nil?}.empty?
           sleep 0.2
         end
       rescue Interrupt
@@ -46,6 +45,7 @@ module Puma
     end
 
     def start_phased_restart
+      @events.fire_on_restart!
       @phase += 1
       log "- Starting phased worker restart, phase: #{@phase}"
 
@@ -62,143 +62,102 @@ module Puma
       @workers.each { |x| x.hup }
     end
 
-    class Worker
-      def initialize(idx, pid, phase, options)
-        @index = idx
-        @pid = pid
-        @phase = phase
-        @stage = :started
-        @signal = "TERM"
-        @options = options
-        @first_term_sent = nil
-        @started_at = Time.now
-        @last_checkin = Time.now
-        @last_status = '{}'
-        @term = false
-      end
-
-      attr_reader :index, :pid, :phase, :signal, :last_checkin, :last_status, :started_at
-
-      def booted?
-        @stage == :booted
-      end
-
-      def boot!
-        @last_checkin = Time.now
-        @stage = :booted
-      end
-
-      def term?
-        @term
-      end
-
-      def ping!(status)
-        @last_checkin = Time.now
-        @last_status = status
-      end
-
-      def ping_timeout?(which)
-        Time.now - @last_checkin > which
-      end
-
-      def term
-        begin
-          if @first_term_sent && (Time.now - @first_term_sent) > @options[:worker_shutdown_timeout]
-            @signal = "KILL"
-          else
-            @term ||= true
-            @first_term_sent ||= Time.now
-          end
-          Process.kill @signal, @pid
-        rescue Errno::ESRCH
-        end
-      end
-
-      def kill
-        Process.kill "KILL", @pid
-      rescue Errno::ESRCH
-      end
-
-      def hup
-        Process.kill "HUP", @pid
-      rescue Errno::ESRCH
-      end
-    end
-
     def spawn_workers
       diff = @options[:workers] - @workers.size
       return if diff < 1
 
       master = Process.pid
+      if @options[:fork_worker]
+        @fork_writer << "-1\n"
+      end
 
       diff.times do
         idx = next_worker_index
-        @launcher.config.run_hooks :before_worker_fork, idx
 
-        pid = fork { worker(idx, master) }
-        if !pid
-          log "! Complete inability to spawn new workers detected"
-          log "! Seppuku is the only choice."
-          exit! 1
+        if @options[:fork_worker] && idx != 0
+          @fork_writer << "#{idx}\n"
+          pid = nil
+        else
+          pid = spawn_worker(idx, master)
         end
 
         debug "Spawned worker: #{pid}"
-        @workers << Worker.new(idx, pid, @phase, @options)
+        @workers << WorkerHandle.new(idx, pid, @phase, @options)
+      end
+
+      if @options[:fork_worker] &&
+        @workers.all? {|x| x.phase == @phase}
 
-        @launcher.config.run_hooks :after_worker_fork, idx
+        @fork_writer << "0\n"
       end
+    end
+
+    # @version 5.0.0
+    def spawn_worker(idx, master)
+      @config.run_hooks(:before_worker_fork, idx, @log_writer)
 
-      if diff > 0
-        @phased_state = :idle
+      pid = fork { worker(idx, master) }
+      if !pid
+        log "! Complete inability to spawn new workers detected"
+        log "! Seppuku is the only choice."
+        exit! 1
       end
+
+      @config.run_hooks(:after_worker_fork, idx, @log_writer)
+      pid
     end
 
     def cull_workers
       diff = @workers.size - @options[:workers]
       return if diff < 1
+      debug "Culling #{diff} workers"
 
-      debug "Culling #{diff.inspect} workers"
-
-      workers_to_cull = @workers[-diff,diff]
-      debug "Workers to cull: #{workers_to_cull.inspect}"
+      workers = workers_to_cull(diff)
+      debug "Workers to cull: #{workers.inspect}"
 
-      workers_to_cull.each do |worker|
-        log "- Worker #{worker.index} (pid: #{worker.pid}) terminating"
+      workers.each do |worker|
+        log "- Worker #{worker.index} (PID: #{worker.pid}) terminating"
         worker.term
       end
     end
 
+    def workers_to_cull(diff)
+      workers = @workers.sort_by(&:started_at)
+
+      # In fork_worker mode, worker 0 acts as our master process.
+      # We should avoid culling it to preserve copy-on-write memory gains.
+      workers.reject! { |w| w.index == 0 } if @options[:fork_worker]
+
+      workers[cull_start_index(diff), diff]
+    end
+
+    def cull_start_index(diff)
+      case @options[:worker_culling_strategy]
+      when :oldest
+        0
+      else # :youngest
+        -diff
+      end
+    end
+
+    # @!attribute [r] next_worker_index
     def next_worker_index
-      all_positions =  0...@options[:workers]
-      occupied_positions = @workers.map { |w| w.index }
-      available_positions = all_positions.to_a - occupied_positions
-      available_positions.first
+      occupied_positions = @workers.map(&:index)
+      idx = 0
+      idx += 1 until !occupied_positions.include?(idx)
+      idx
     end
 
     def all_workers_booted?
       @workers.count { |w| !w.booted? } == 0
     end
 
-    def check_workers(force=false)
-      return if !force && @next_check && @next_check >= Time.now
+    def check_workers
+      return if @next_check >= Time.now
 
-      @next_check = Time.now + Const::WORKER_CHECK_INTERVAL
-
-      any = false
-
-      @workers.each do |w|
-        next if !w.booted? && !w.ping_timeout?(@options[:worker_boot_timeout])
-        if w.ping_timeout?(@options[:worker_timeout])
-          log "! Terminating timed out worker: #{w.pid}"
-          w.kill
-          any = true
-        end
-      end
-
-      # If we killed any timed out workers, try to catch them
-      # during this loop by giving the kernel time to kill them.
-      sleep 1 if any
+      @next_check = Time.now + @options[:worker_check_interval]
 
+      timeout_workers
       wait_workers
       cull_workers
       spawn_workers
@@ -211,104 +170,40 @@ module Puma
         w = @workers.find { |x| x.phase != @phase }
 
         if w
-          if @phased_state == :idle
-            @phased_state = :waiting
-            log "- Stopping #{w.pid} for phased upgrade..."
-          end
-
+          log "- Stopping #{w.pid} for phased upgrade..."
           unless w.term?
             w.term
             log "- #{w.signal} sent to #{w.pid}..."
           end
         end
       end
-    end
 
-    def wakeup!
-      return unless @wakeup
+      t = @workers.reject(&:term?)
+      t.map!(&:ping_timeout)
 
-      begin
-        @wakeup.write "!" unless @wakeup.closed?
-      rescue SystemCallError, IOError
-        Thread.current.purge_interrupt_queue if Thread.current.respond_to? :purge_interrupt_queue
-      end
+      @next_check = [t.min, @next_check].compact.min
     end
 
     def worker(index, master)
-      title  = "puma: cluster worker #{index}: #{master}"
-      title += " [#{@options[:tag]}]" if @options[:tag] && !@options[:tag].empty?
-      $0 = title
-
-      Signal.trap "SIGINT", "IGNORE"
-
       @workers = []
+
       @master_read.close
       @suicide_pipe.close
+      @fork_writer.close
 
-      Thread.new do
-        Puma.set_thread_name "worker check pipe"
-        IO.select [@check_pipe]
-        log "! Detected parent died, dying"
-        exit! 1
-      end
-
-      # If we're not running under a Bundler context, then
-      # report the info about the context we will be using
-      if !ENV['BUNDLE_GEMFILE']
-        if File.exist?("Gemfile")
-          log "+ Gemfile in context: #{File.expand_path("Gemfile")}"
-        elsif File.exist?("gems.rb")
-          log "+ Gemfile in context: #{File.expand_path("gems.rb")}"
-        end
+      pipes = { check_pipe: @check_pipe, worker_write: @worker_write }
+      if @options[:fork_worker]
+        pipes[:fork_pipe] = @fork_pipe
+        pipes[:wakeup] = @wakeup
       end
 
-      # Invoke any worker boot hooks so they can get
-      # things in shape before booting the app.
-      @launcher.config.run_hooks :before_worker_boot, index
-
-      server = start_server
-
-      Signal.trap "SIGTERM" do
-        @worker_write << "e#{Process.pid}\n" rescue nil
-        server.stop
-      end
-
-      begin
-        @worker_write << "b#{Process.pid}\n"
-      rescue SystemCallError, IOError
-        Thread.current.purge_interrupt_queue if Thread.current.respond_to? :purge_interrupt_queue
-        STDERR.puts "Master seems to have exited, exiting."
-        return
-      end
-
-      Thread.new(@worker_write) do |io|
-        Puma.set_thread_name "stat payload"
-        base_payload = "p#{Process.pid}"
-
-        while true
-          sleep Const::WORKER_CHECK_INTERVAL
-          begin
-            b = server.backlog || 0
-            r = server.running || 0
-            t = server.pool_capacity || 0
-            m = server.max_threads || 0
-            payload = %Q!#{base_payload}{ "backlog":#{b}, "running":#{r}, "pool_capacity":#{t}, "max_threads": #{m} }\n!
-            io << payload
-          rescue IOError
-            Thread.current.purge_interrupt_queue if Thread.current.respond_to? :purge_interrupt_queue
-            break
-          end
-        end
-      end
-
-      server.run.join
-
-      # Invoke any worker shutdown hooks so they can prevent the worker
-      # exiting until any background operations are completed
-      @launcher.config.run_hooks :before_worker_shutdown, index
-    ensure
-      @worker_write << "t#{Process.pid}\n" rescue nil
-      @worker_write.close
+      server = start_server if preload?
+      new_worker = Worker.new index: index,
+                              master: master,
+                              launcher: @launcher,
+                              pipes: pipes,
+                              server: server
+      new_worker.run
     end
 
     def restart
@@ -316,8 +211,8 @@ module Puma
       stop
     end
 
-    def phased_restart
-      return false if @options[:preload_app]
+    def phased_restart(refork = false)
+      return false if @options[:preload_app] && !refork
 
       @phased_restart = true
       wakeup!
@@ -333,7 +228,7 @@ module Puma
     def stop_blocked
       @status = :stop if @status == :run
       wakeup!
-      @control.stop(true) if @control
+      @control&.stop true
       Process.waitall
     end
 
@@ -350,20 +245,61 @@ module Puma
 
     # Inside of a child process, this will return all zeroes, as @workers is only populated in
     # the master process.
+    # @!attribute [r] stats
     def stats
       old_worker_count = @workers.count { |w| w.phase != @phase }
-      booted_worker_count = @workers.count { |w| w.booted? }
-      worker_status = '[' + @workers.map { |w| %Q!{ "started_at": "#{w.started_at.utc.iso8601}", "pid": #{w.pid}, "index": #{w.index}, "phase": #{w.phase}, "booted": #{w.booted?}, "last_checkin": "#{w.last_checkin.utc.iso8601}", "last_status": #{w.last_status} }!}.join(",") + ']'
-      %Q!{ "started_at": "#{@started_at.utc.iso8601}", "workers": #{@workers.size}, "phase": #{@phase}, "booted_workers": #{booted_worker_count}, "old_workers": #{old_worker_count}, "worker_status": #{worker_status} }!
+      worker_status = @workers.map do |w|
+        {
+          started_at: utc_iso8601(w.started_at),
+          pid: w.pid,
+          index: w.index,
+          phase: w.phase,
+          booted: w.booted?,
+          last_checkin: utc_iso8601(w.last_checkin),
+          last_status: w.last_status,
+        }
+      end
+
+      {
+        started_at: utc_iso8601(@started_at),
+        workers: @workers.size,
+        phase: @phase,
+        booted_workers: worker_status.count { |w| w[:booted] },
+        old_workers: old_worker_count,
+        worker_status: worker_status,
+      }.merge(super)
     end
 
     def preload?
       @options[:preload_app]
     end
 
+    # @version 5.0.0
+    def fork_worker!
+      if (worker = @workers.find { |w| w.index == 0 })
+        worker.phase += 1
+      end
+      phased_restart(true)
+    end
+
     # We do this in a separate method to keep the lambda scope
     # of the signals handlers as small as possible.
     def setup_signals
+      if @options[:fork_worker]
+        Signal.trap "SIGURG" do
+          fork_worker!
+        end
+
+        # Auto-fork after the specified number of requests.
+        if (fork_requests = @options[:fork_worker].to_i) > 0
+          @events.register(:ping!) do |w|
+            fork_worker! if w.index == 0 &&
+              w.phase == 0 &&
+              w.last_status[:requests_count] >= fork_requests
+          end
+        end
+      end
+
       Signal.trap "SIGCHLD" do
         wakeup!
       end
@@ -392,7 +328,7 @@ module Puma
 
           stop_workers
           stop
-
+          @events.fire_on_stopped!
           raise(SignalException, "SIGTERM") if @options[:raise_exception_on_sigterm]
           exit 0 # Clean exit, workers were stopped
         end
@@ -404,15 +340,25 @@ module Puma
 
       output_header "cluster"
 
-      log "* Process workers: #{@options[:workers]}"
-
-      before = Thread.list
+      # This is aligned with the output from Runner, see Runner#output_header
+      log "*      Workers: #{@options[:workers]}"
 
       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) }
+
+        before = Thread.list.reject(&fork_safe)
+
+        log "*     Restarts: (\u2714) hot (\u2716) phased"
         log "* Preloading application"
         load_and_bind
 
-        after = Thread.list
+        after = Thread.list.reject(&fork_safe)
 
         if after.size > before.size
           threads = (after - before)
@@ -426,14 +372,14 @@ module Puma
           end
         end
       else
-        log "* Phased restart available"
+        log "*     Restarts: (\u2714) hot (\u2714) phased"
 
-        unless @launcher.config.app_configured?
+        unless @config.app_configured?
           error "No application configured, nothing to run"
           exit 1
         end
 
-        @launcher.binder.parse @options[:binds], self
+        @launcher.binder.parse @options[:binds]
       end
 
       read, @wakeup = Puma::Util.pipe
@@ -447,12 +393,13 @@ module Puma
       #
       @check_pipe, @suicide_pipe = Puma::Util.pipe
 
-      if daemon?
-        log "* Daemonizing..."
-        Process.daemon(true)
-      else
-        log "Use Ctrl-C to stop"
-      end
+      # Separate pipe used by worker 0 to receive commands to
+      # fork new worker processes.
+      @fork_pipe, @fork_writer = Puma::Util.pipe
+
+      log "Use Ctrl-C to stop"
+
+      single_worker_warning
 
       redirect_io
 
@@ -464,7 +411,7 @@ module Puma
 
       @master_read, @worker_write = read, @wakeup
 
-      @launcher.config.run_hooks :before_fork, nil
+      @config.run_hooks(:before_fork, nil, @log_writer)
 
       spawn_workers
 
@@ -472,51 +419,65 @@ module Puma
         stop
       end
 
-      @launcher.events.fire_on_booted!
-
       begin
-        force_check = false
+        booted = false
+        in_phased_restart = false
+        workers_not_booted = @options[:workers]
 
         while @status == :run
           begin
             if @phased_restart
               start_phased_restart
               @phased_restart = false
+              in_phased_restart = true
+              workers_not_booted = @options[:workers]
             end
 
-            check_workers force_check
+            check_workers
 
-            force_check = false
-
-            res = IO.select([read], nil, nil, Const::WORKER_CHECK_INTERVAL)
-
-            if res
+            if read.wait_readable([0, @next_check - Time.now].max)
               req = read.read_nonblock(1)
 
+              @next_check = Time.now if req == "!"
               next if !req || req == "!"
 
               result = read.gets
               pid = result.to_i
 
+              if req == "b" || req == "f"
+                pid, idx = result.split(':').map(&:to_i)
+                w = @workers.find {|x| x.index == idx}
+                w.pid = pid if w.pid.nil?
+              end
+
               if w = @workers.find { |x| x.pid == pid }
                 case req
                 when "b"
                   w.boot!
-                  log "- Worker #{w.index} (pid: #{pid}) booted, phase: #{w.phase}"
-                  force_check = true
+                  log "- Worker #{w.index} (PID: #{pid}) booted in #{w.uptime.round(2)}s, phase: #{w.phase}"
+                  @next_check = Time.now
+                  workers_not_booted -= 1
                 when "e"
                   # external term, see worker method, Signal.trap "SIGTERM"
-                  w.instance_variable_set :@term, true
+                  w.term!
                 when "t"
                   w.term unless w.term?
-                  force_check = true
                 when "p"
                   w.ping!(result.sub(/^\d+/,'').chomp)
+                  @events.fire(:ping!, w)
+                  if !booted && @workers.none? {|worker| worker.last_status.empty?}
+                    @events.fire_on_booted!
+                    booted = true
+                  end
                 end
               else
                 log "! Out-of-sync worker list, no #{pid} worker"
               end
             end
+            if in_phased_restart && workers_not_booted.zero?
+              @events.fire_on_booted!
+              in_phased_restart = false
+            end
 
           rescue Interrupt
             @status = :stop
@@ -534,10 +495,20 @@ module Puma
 
     private
 
+    def single_worker_warning
+      return if @options[:workers] != 1 || @options[:silence_single_worker_warning]
+
+      log "! WARNING: Detected running cluster mode with 1 worker."
+      log "! Running Puma in cluster mode with a single worker is often a misconfiguration."
+      log "! Consider running Puma in single-mode (workers = 0) in order to reduce memory overhead."
+      log "! Set the `silence_single_worker_warning` option to silence this warning message."
+    end
+
     # loops thru @workers, removing workers that exited, and calling
     # `#term` if needed
     def wait_workers
       @workers.reject! do |w|
+        next false if w.pid.nil?
         begin
           if Process.wait(w.pid, Process::WNOHANG)
             true
@@ -546,7 +517,29 @@ module Puma
             nil
           end
         rescue Errno::ECHILD
-          true # child is already terminated
+          begin
+            Process.kill(0, w.pid)
+            # child still alive but has another parent (e.g., using fork_worker)
+            w.term if w.term?
+            false
+          rescue Errno::ESRCH, Errno::EPERM
+            true # child is already terminated
+          end
+        end
+      end
+    end
+
+    # @version 5.0.0
+    def timeout_workers
+      @workers.each do |w|
+        if !w.term? && w.ping_timeout <= Time.now
+          details = if w.booted?
+                      "(worker failed to check in within #{@options[:worker_timeout]} seconds)"
+                    else
+                      "(worker failed to boot within #{@options[:worker_boot_timeout]} seconds)"
+                    end
+          log "! Terminating timed out worker #{details}: #{w.pid}"
+          w.kill
         end
       end
     end