puma 4.1.1 → 5.0.0

data/lib/puma/cluster.rb

@@ -24,9 +24,8 @@ module Puma
 
       @phase = 0
       @workers = []
-      @next_check = nil
+      @next_check = Time.now
 
-      @phased_state = :idle
       @phased_restart = false
     end
 
@@ -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
@@ -73,13 +72,15 @@ module Puma
         @first_term_sent = nil
         @started_at = Time.now
         @last_checkin = Time.now
-        @last_status = '{}'
-        @dead = false
+        @last_status = {}
         @term = false
       end
 
       attr_reader :index, :pid, :phase, :signal, :last_checkin, :last_status, :started_at
 
+      # @version 5.0.0
+      attr_writer :pid, :phase
+
       def booted?
         @stage == :booted
       end
@@ -89,25 +90,24 @@ module Puma
         @stage = :booted
       end
 
-      def dead?
-        @dead
-      end
-
-      def dead!
-        @dead = true
-      end
-
       def term?
         @term
       end
 
       def ping!(status)
         @last_checkin = Time.now
-        @last_status = status
+        require 'json'
+        @last_status = JSON.parse(status, symbolize_names: true)
       end
 
-      def ping_timeout?(which)
-        Time.now - @last_checkin > which
+      # @see Puma::Cluster#check_workers
+      # @version 5.0.0
+      def ping_timeout
+        @last_checkin +
+          (booted? ?
+            @options[:worker_timeout] :
+            @options[:worker_boot_timeout]
+          )
       end
 
       def term
@@ -118,14 +118,14 @@ module Puma
             @term ||= true
             @first_term_sent ||= Time.now
           end
-          Process.kill @signal, @pid
+          Process.kill @signal, @pid if @pid
         rescue Errno::ESRCH
         end
       end
 
       def kill
-        Process.kill "KILL", @pid
-      rescue Errno::ESRCH
+        @signal = 'KILL'
+        term
       end
 
       def hup
@@ -139,27 +139,44 @@ module Puma
       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)
+      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)
+      @launcher.config.run_hooks :before_worker_fork, idx, @launcher.events
 
-      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
+
+      @launcher.config.run_hooks :after_worker_fork, idx, @launcher.events
+      pid
     end
 
     def cull_workers
@@ -188,26 +205,12 @@ module Puma
       @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
-
+      timeout_workers
       wait_workers
       cull_workers
       spawn_workers
@@ -220,15 +223,18 @@ 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..."
+          log "- Stopping #{w.pid} for phased upgrade..."
+          unless w.term?
+            w.term
+            log "- #{w.signal} sent to #{w.pid}..."
           end
-
-          w.term
-          log "- #{w.signal} sent to #{w.pid}..."
         end
       end
+
+      @next_check = [
+        @workers.reject(&:term?).map(&:ping_timeout).min,
+        @next_check
+      ].compact.min
     end
 
     def wakeup!
@@ -247,12 +253,19 @@ module Puma
       $0 = title
 
       Signal.trap "SIGINT", "IGNORE"
+      Signal.trap "SIGCHLD", "DEFAULT"
+
+      fork_worker = @options[:fork_worker] && index == 0
 
       @workers = []
-      @master_read.close
-      @suicide_pipe.close
+      if !@options[:fork_worker] || fork_worker
+        @master_read.close
+        @suicide_pipe.close
+        @fork_writer.close
+      end
 
       Thread.new do
+        Puma.set_thread_name "worker check pipe"
         IO.select [@check_pipe]
         log "! Detected parent died, dying"
         exit! 1
@@ -270,16 +283,49 @@ module Puma
 
       # Invoke any worker boot hooks so they can get
       # things in shape before booting the app.
-      @launcher.config.run_hooks :before_worker_boot, index
+      @launcher.config.run_hooks :before_worker_boot, index, @launcher.events
+
+      server = @server ||= start_server
+      restart_server = Queue.new << true << false
 
-      server = start_server
+      if fork_worker
+        restart_server.clear
+        worker_pids = []
+        Signal.trap "SIGCHLD" do
+          wakeup! if worker_pids.reject! do |p|
+            Process.wait(p, Process::WNOHANG) rescue true
+          end
+        end
+
+        Thread.new do
+          Puma.set_thread_name "worker fork pipe"
+          while (idx = @fork_pipe.gets)
+            idx = idx.to_i
+            if idx == -1 # stop server
+              if restart_server.length > 0
+                restart_server.clear
+                server.begin_restart(true)
+                @launcher.config.run_hooks :before_refork, nil, @launcher.events
+                nakayoshi_gc
+              end
+            elsif idx == 0 # restart server
+              restart_server << true << false
+            else # fork worker
+              worker_pids << pid = spawn_worker(idx, master)
+              @worker_write << "f#{pid}:#{idx}\n" rescue nil
+            end
+          end
+        end
+      end
 
       Signal.trap "SIGTERM" do
+        @worker_write << "e#{Process.pid}\n" rescue nil
         server.stop
+        restart_server << false
       end
 
       begin
-        @worker_write << "b#{Process.pid}\n"
+        @worker_write << "b#{Process.pid}:#{index}\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."
@@ -287,17 +333,13 @@ module Puma
       end
 
       Thread.new(@worker_write) do |io|
-        base_payload = "p#{Process.pid}"
+        Puma.set_thread_name "stat payload"
 
         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
+            require 'json'
+            io << "p#{Process.pid}#{server.stats.to_json}\n"
           rescue IOError
             Thread.current.purge_interrupt_queue if Thread.current.respond_to? :purge_interrupt_queue
             break
@@ -305,11 +347,11 @@ module Puma
         end
       end
 
-      server.run.join
+      server.run.join while restart_server.pop
 
       # 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
+      @launcher.config.run_hooks :before_worker_shutdown, index, @launcher.events
     ensure
       @worker_write << "t#{Process.pid}\n" rescue nil
       @worker_write.close
@@ -352,20 +394,62 @@ module Puma
       Dir.chdir dir
     end
 
+    # Inside of a child process, this will return all zeroes, as @workers is only populated in
+    # the master process.
     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: 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,
+        }
+      end
+
+      {
+        started_at: @started_at.utc.iso8601,
+        workers: @workers.size,
+        phase: @phase,
+        booted_workers: worker_status.count { |w| w[:booted] },
+        old_workers: old_worker_count,
+        worker_status: worker_status,
+      }
     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
+    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
+          @launcher.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
@@ -449,12 +533,11 @@ 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"
 
       redirect_io
 
@@ -466,7 +549,8 @@ module Puma
 
       @master_read, @worker_write = read, @wakeup
 
-      @launcher.config.run_hooks :before_fork, nil
+      @launcher.config.run_hooks :before_fork, nil, @launcher.events
+      nakayoshi_gc
 
       spawn_workers
 
@@ -477,8 +561,6 @@ module Puma
       @launcher.events.fire_on_booted!
 
       begin
-        force_check = false
-
         while @status == :run
           begin
             if @phased_restart
@@ -486,31 +568,39 @@ module Puma
               @phased_restart = false
             end
 
-            check_workers force_check
-
-            force_check = false
+            check_workers
 
-            res = IO.select([read], nil, nil, Const::WORKER_CHECK_INTERVAL)
+            res = IO.select([read], nil, nil, [0, @next_check - Time.now].max)
 
             if res
               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
+                  @next_check = Time.now
+                when "e"
+                  # external term, see worker method, Signal.trap "SIGTERM"
+                  w.instance_variable_set :@term, true
                 when "t"
-                  w.dead!
-                  force_check = true
+                  w.term unless w.term?
                 when "p"
                   w.ping!(result.sub(/^\d+/,'').chomp)
+                  @launcher.events.fire(:ping!, w)
                 end
               else
                 log "! Out-of-sync worker list, no #{pid} worker"
@@ -537,6 +627,7 @@ module Puma
     # `#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
@@ -545,9 +636,36 @@ module Puma
             nil
           end
         rescue Errno::ECHILD
-          true # child is already terminated
+          begin
+            Process.kill(0, w.pid)
+            false # child still alive, but has another parent
+          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
+          log "! Terminating timed out worker: #{w.pid}"
+          w.kill
         end
       end
     end
+
+    # @version 5.0.0
+    def nakayoshi_gc
+      return unless @options[:nakayoshi_fork]
+      log "! Promoting existing objects to old generation..."
+      4.times { GC.start(full_mark: false) }
+      if GC.respond_to?(:compact)
+        log "! Compacting..."
+        GC.compact
+      end
+      log "! Friendly fork preparation complete."
+    end
   end
 end