fix per-worker listen directive in after_fork hook

[unicorn.git] / lib / unicorn / http_server.rb

# -*- encoding: binary -*-

# This is the process manager of Unicorn. This manages worker
# processes which in turn handle the I/O and application process.
# Listener sockets are started in the master process and shared with
# forked worker children.
#
# Users do not need to know the internals of this class, but reading the
# {source}[http://bogomips.org/unicorn.git/tree/lib/unicorn/http_server.rb]
# is education for programmers wishing to learn how \Unicorn works.
# See Unicorn::Configurator for information on how to configure \Unicorn.
class Unicorn::HttpServer
  # :stopdoc:
  attr_accessor :app, :request, :timeout, :worker_processes,
                :before_fork, :after_fork, :before_exec,
                :listener_opts, :preload_app,
                :reexec_pid, :orig_app, :init_listeners,
                :master_pid, :config, :ready_pipe, :user
  attr_reader :pid, :logger
  include Unicorn::SocketHelper
  include Unicorn::HttpResponse

  # backwards compatibility with 1.x
  Worker = Unicorn::Worker

  # all bound listener sockets
  LISTENERS = []

  # This hash maps PIDs to Workers
  WORKERS = {}

  # We use SELF_PIPE differently in the master and worker processes:
  #
  # * The master process never closes or reinitializes this once
  # initialized.  Signal handlers in the master process will write to
  # it to wake up the master from IO.select in exactly the same manner
  # djb describes in http://cr.yp.to/docs/selfpipe.html
  #
  # * The workers immediately close the pipe they inherit from the
  # master and replace it with a new pipe after forking.  This new
  # pipe is also used to wakeup from IO.select from inside (worker)
  # signal handlers.  However, workers *close* the pipe descriptors in
  # the signal handlers to raise EBADF in IO.select instead of writing
  # like we do in the master.  We cannot easily use the reader set for
  # IO.select because LISTENERS is already that set, and it's extra
  # work (and cycles) to distinguish the pipe FD from the reader set
  # once IO.select returns.  So we're lazy and just close the pipe when
  # a (rare) signal arrives in the worker and reinitialize the pipe later.
  SELF_PIPE = []

  # signal queue used for self-piping
  SIG_QUEUE = []

  # list of signals we care about and trap in master.
  QUEUE_SIGS = [ :WINCH, :QUIT, :INT, :TERM, :USR1, :USR2, :HUP, :TTIN, :TTOU ]

  # :startdoc:
  # We populate this at startup so we can figure out how to reexecute
  # and upgrade the currently running instance of Unicorn
  # This Hash is considered a stable interface and changing its contents
  # will allow you to switch between different installations of Unicorn
  # or even different installations of the same applications without
  # downtime.  Keys of this constant Hash are described as follows:
  #
  # * 0 - the path to the unicorn/unicorn_rails executable
  # * :argv - a deep copy of the ARGV array the executable originally saw
  # * :cwd - the working directory of the application, this is where
  # you originally started Unicorn.
  #
  # To change your unicorn executable to a different path without downtime,
  # you can set the following in your Unicorn config file, HUP and then
  # continue with the traditional USR2 + QUIT upgrade steps:
  #
  #   Unicorn::HttpServer::START_CTX[0] = "/home/bofh/1.9.2/bin/unicorn"
  START_CTX = {
    :argv => ARGV.map { |arg| arg.dup },
    0 => $0.dup,
  }
  # We favor ENV['PWD'] since it is (usually) symlink aware for Capistrano
  # and like systems
  START_CTX[:cwd] = begin
    a = File.stat(pwd = ENV['PWD'])
    b = File.stat(Dir.pwd)
    a.ino == b.ino && a.dev == b.dev ? pwd : Dir.pwd
  rescue
    Dir.pwd
  end
  # :stopdoc:

  # Creates a working server on host:port (strange things happen if
  # port isn't a Number).  Use HttpServer::run to start the server and
  # HttpServer.run.join to join the thread that's processing
  # incoming requests on the socket.
  def initialize(app, options = {})
    @app = app
    @request = Unicorn::HttpRequest.new
    self.reexec_pid = 0
    options = options.dup
    @ready_pipe = options.delete(:ready_pipe)
    @init_listeners = options[:listeners] ? options[:listeners].dup : []
    options[:use_defaults] = true
    self.config = Unicorn::Configurator.new(options)
    self.listener_opts = {}

    # we try inheriting listeners first, so we bind them later.
    # we don't write the pid file until we've bound listeners in case
    # unicorn was started twice by mistake.  Even though our #pid= method
    # checks for stale/existing pid files, race conditions are still
    # possible (and difficult/non-portable to avoid) and can be likely
    # to clobber the pid if the second start was in quick succession
    # after the first, so we rely on the listener binding to fail in
    # that case.  Some tests (in and outside of this source tree) and
    # monitoring tools may also rely on pid files existing before we
    # attempt to connect to the listener(s)
    config.commit!(self, :skip => [:listeners, :pid])
    self.orig_app = app
  end

  # Runs the thing.  Returns self so you can run join on it
  def start
    inherit_listeners!
    # this pipe is used to wake us up from select(2) in #join when signals
    # are trapped.  See trap_deferred.
    init_self_pipe!

    # setup signal handlers before writing pid file in case people get
    # trigger happy and send signals as soon as the pid file exists.
    # Note that signals don't actually get handled until the #join method
    QUEUE_SIGS.each { |sig| trap(sig) { SIG_QUEUE << sig; awaken_master } }
    trap(:CHLD) { awaken_master }
    self.pid = config[:pid]

    self.master_pid = $$
    build_app! if preload_app
    spawn_missing_workers
    self
  end

  # replaces current listener set with +listeners+.  This will
  # close the socket if it will not exist in the new listener set
  def listeners=(listeners)
    cur_names, dead_names = [], []
    listener_names.each do |name|
      if ?/ == name[0]
        # mark unlinked sockets as dead so we can rebind them
        (File.socket?(name) ? cur_names : dead_names) << name
      else
        cur_names << name
      end
    end
    set_names = listener_names(listeners)
    dead_names.concat(cur_names - set_names).uniq!

    LISTENERS.delete_if do |io|
      if dead_names.include?(sock_name(io))
        IO_PURGATORY.delete_if do |pio|
          pio.fileno == io.fileno && (pio.close rescue nil).nil? # true
        end
        (io.close rescue nil).nil? # true
      else
        set_server_sockopt(io, listener_opts[sock_name(io)])
        false
      end
    end

    (set_names - cur_names).each { |addr| listen(addr) }
  end

  def stdout_path=(path); redirect_io($stdout, path); end
  def stderr_path=(path); redirect_io($stderr, path); end

  def logger=(obj)
    Unicorn::HttpRequest::DEFAULTS["rack.logger"] = @logger = obj
  end

  # sets the path for the PID file of the master process
  def pid=(path)
    if path
      if x = valid_pid?(path)
        return path if pid && path == pid && x == $$
        if x == reexec_pid && pid =~ /\.oldbin\z/
          logger.warn("will not set pid=#{path} while reexec-ed "\
                      "child is running PID:#{x}")
          return
        end
        raise ArgumentError, "Already running on PID:#{x} " \
                             "(or pid=#{path} is stale)"
      end
    end
    unlink_pid_safe(pid) if pid

    if path
      fp = begin
        tmp = "#{File.dirname(path)}/#{rand}.#$$"
        File.open(tmp, File::RDWR|File::CREAT|File::EXCL, 0644)
      rescue Errno::EEXIST
        retry
      end
      fp.syswrite("#$$\n")
      File.rename(fp.path, path)
      fp.close
    end
    @pid = path
  end

  # add a given address to the +listeners+ set, idempotently
  # Allows workers to add a private, per-process listener via the
  # after_fork hook.  Very useful for debugging and testing.
  # +:tries+ may be specified as an option for the number of times
  # to retry, and +:delay+ may be specified as the time in seconds
  # to delay between retries.
  # A negative value for +:tries+ indicates the listen will be
  # retried indefinitely, this is useful when workers belonging to
  # different masters are spawned during a transparent upgrade.
  def listen(address, opt = {}.merge(listener_opts[address] || {}))
    address = config.expand_addr(address)
    return if String === address && listener_names.include?(address)

    delay = opt[:delay] || 0.5
    tries = opt[:tries] || 5
    begin
      io = bind_listen(address, opt)
      unless Kgio::TCPServer === io || Kgio::UNIXServer === io
        IO_PURGATORY << io
        io = server_cast(io)
      end
      logger.info "listening on addr=#{sock_name(io)} fd=#{io.fileno}"
      LISTENERS << io
      io
    rescue Errno::EADDRINUSE => err
      logger.error "adding listener failed addr=#{address} (in use)"
      raise err if tries == 0
      tries -= 1
      logger.error "retrying in #{delay} seconds " \
                   "(#{tries < 0 ? 'infinite' : tries} tries left)"
      sleep(delay)
      retry
    rescue => err
      logger.fatal "error adding listener addr=#{address}"
      raise err
    end
  end

  # monitors children and receives signals forever
  # (or until a termination signal is sent).  This handles signals
  # one-at-a-time time and we'll happily drop signals in case somebody
  # is signalling us too often.
  def join
    respawn = true
    last_check = Time.now

    proc_name 'master'
    logger.info "master process ready" # test_exec.rb relies on this message
    if @ready_pipe
      @ready_pipe.syswrite($$.to_s)
      @ready_pipe = @ready_pipe.close rescue nil
    end
    begin
      reap_all_workers
      case SIG_QUEUE.shift
      when nil
        # avoid murdering workers after our master process (or the
        # machine) comes out of suspend/hibernation
        if (last_check + @timeout) >= (last_check = Time.now)
          sleep_time = murder_lazy_workers
        else
          # wait for workers to wakeup on suspend
          sleep_time = @timeout/2.0 + 1
        end
        maintain_worker_count if respawn
        master_sleep(sleep_time)
      when :QUIT # graceful shutdown
        break
      when :TERM, :INT # immediate shutdown
        stop(false)
        break
      when :USR1 # rotate logs
        logger.info "master reopening logs..."
        Unicorn::Util.reopen_logs
        logger.info "master done reopening logs"
        kill_each_worker(:USR1)
      when :USR2 # exec binary, stay alive in case something went wrong
        reexec
      when :WINCH
        if Process.ppid == 1 || Process.getpgrp != $$
          respawn = false
          logger.info "gracefully stopping all workers"
          kill_each_worker(:QUIT)
          self.worker_processes = 0
        else
          logger.info "SIGWINCH ignored because we're not daemonized"
        end
      when :TTIN
        respawn = true
        self.worker_processes += 1
      when :TTOU
        self.worker_processes -= 1 if self.worker_processes > 0
      when :HUP
        respawn = true
        if config.config_file
          load_config!
        else # exec binary and exit if there's no config file
          logger.info "config_file not present, reexecuting binary"
          reexec
        end
      end
    rescue => e
      Unicorn.log_error(@logger, "master loop error", e)
    end while true
    stop # gracefully shutdown all workers on our way out
    logger.info "master complete"
    unlink_pid_safe(pid) if pid
  end

  # Terminates all workers, but does not exit master process
  def stop(graceful = true)
    self.listeners = []
    limit = Time.now + timeout
    until WORKERS.empty? || Time.now > limit
      kill_each_worker(graceful ? :QUIT : :TERM)
      sleep(0.1)
      reap_all_workers
    end
    kill_each_worker(:KILL)
  end

  def rewindable_input
    Unicorn::HttpRequest.input_class.method_defined?(:rewind)
  end

  def rewindable_input=(bool)
    Unicorn::HttpRequest.input_class = bool ?
                                Unicorn::TeeInput : Unicorn::StreamInput
  end

  def client_body_buffer_size
    Unicorn::TeeInput.client_body_buffer_size
  end

  def client_body_buffer_size=(bytes)
    Unicorn::TeeInput.client_body_buffer_size = bytes
  end

  def trust_x_forwarded
    Unicorn::HttpParser.trust_x_forwarded?
  end

  def trust_x_forwarded=(bool)
    Unicorn::HttpParser.trust_x_forwarded = bool
  end

  private

  # wait for a signal hander to wake us up and then consume the pipe
  def master_sleep(sec)
    IO.select([ SELF_PIPE[0] ], nil, nil, sec) or return
    SELF_PIPE[0].kgio_tryread(11)
  end

  def awaken_master
    SELF_PIPE[1].kgio_trywrite('.') # wakeup master process from select
  end

  # reaps all unreaped workers
  def reap_all_workers
    begin
      wpid, status = Process.waitpid2(-1, Process::WNOHANG)
      wpid or return
      if reexec_pid == wpid
        logger.error "reaped #{status.inspect} exec()-ed"
        self.reexec_pid = 0
        self.pid = pid.chomp('.oldbin') if pid
        proc_name 'master'
      else
        worker = WORKERS.delete(wpid) and worker.close rescue nil
        m = "reaped #{status.inspect} worker=#{worker.nr rescue 'unknown'}"
        status.success? ? logger.info(m) : logger.error(m)
      end
    rescue Errno::ECHILD
      break
    end while true
  end

  # reexecutes the START_CTX with a new binary
  def reexec
    if reexec_pid > 0
      begin
        Process.kill(0, reexec_pid)
        logger.error "reexec-ed child already running PID:#{reexec_pid}"
        return
      rescue Errno::ESRCH
        self.reexec_pid = 0
      end
    end

    if pid
      old_pid = "#{pid}.oldbin"
      begin
        self.pid = old_pid  # clear the path for a new pid file
      rescue ArgumentError
        logger.error "old PID:#{valid_pid?(old_pid)} running with " \
                     "existing pid=#{old_pid}, refusing rexec"
        return
      rescue => e
        logger.error "error writing pid=#{old_pid} #{e.class} #{e.message}"
        return
      end
    end

    self.reexec_pid = fork do
      listener_fds = LISTENERS.map { |sock| sock.fileno }
      ENV['UNICORN_FD'] = listener_fds.join(',')
      Dir.chdir(START_CTX[:cwd])
      cmd = [ START_CTX[0] ].concat(START_CTX[:argv])

      # avoid leaking FDs we don't know about, but let before_exec
      # unset FD_CLOEXEC, if anything else in the app eventually
      # relies on FD inheritence.
      (3..1024).each do |io|
        next if listener_fds.include?(io)
        io = IO.for_fd(io) rescue next
        IO_PURGATORY << io
        io.fcntl(Fcntl::F_SETFD, Fcntl::FD_CLOEXEC)
      end
      logger.info "executing #{cmd.inspect} (in #{Dir.pwd})"
      before_exec.call(self)
      exec(*cmd)
    end
    proc_name 'master (old)'
  end

  # forcibly terminate all workers that haven't checked in in timeout seconds.  The timeout is implemented using an unlinked File
  def murder_lazy_workers
    t = @timeout
    next_sleep = 1
    now = Time.now.to_i
    WORKERS.dup.each_pair do |wpid, worker|
      tick = worker.tick
      0 == tick and next # skip workers that are sleeping
      diff = now - tick
      tmp = t - diff
      if tmp >= 0
        next_sleep < tmp and next_sleep = tmp
        next
      end
      logger.error "worker=#{worker.nr} PID:#{wpid} timeout " \
                   "(#{diff}s > #{t}s), killing"
      kill_worker(:KILL, wpid) # take no prisoners for timeout violations
    end
    next_sleep
  end

  def after_fork_internal
    @ready_pipe.close if @ready_pipe
    Unicorn::Configurator::RACKUP.clear
    @ready_pipe = @init_listeners = @before_exec = @before_fork = nil

    srand # http://redmine.ruby-lang.org/issues/4338

    # The OpenSSL PRNG is seeded with only the pid, and apps with frequently
    # dying workers can recycle pids
    OpenSSL::Random.seed(rand.to_s) if defined?(OpenSSL::Random)
  end

  def spawn_missing_workers
    worker_nr = -1
    until (worker_nr += 1) == @worker_processes
      WORKERS.value?(worker_nr) and next
      worker = Worker.new(worker_nr)
      before_fork.call(self, worker)
      if pid = fork
        WORKERS[pid] = worker
      else
        after_fork_internal
        worker_loop(worker)
        exit
      end
    end
    rescue => e
      @logger.error(e) rescue nil
      exit!
  end

  def maintain_worker_count
    (off = WORKERS.size - worker_processes) == 0 and return
    off < 0 and return spawn_missing_workers
    WORKERS.dup.each_pair { |wpid,w|
      w.nr >= worker_processes and kill_worker(:QUIT, wpid) rescue nil
    }
  end

  # if we get any error, try to write something back to the client
  # assuming we haven't closed the socket, but don't get hung up
  # if the socket is already closed or broken.  We'll always ensure
  # the socket is closed at the end of this function
  def handle_error(client, e)
    msg = case e
    when EOFError,Errno::ECONNRESET,Errno::EPIPE,Errno::EINVAL,Errno::EBADF
      Unicorn::Const::ERROR_500_RESPONSE
    when Unicorn::RequestURITooLongError
      Unicorn::Const::ERROR_414_RESPONSE
    when Unicorn::RequestEntityTooLargeError
      Unicorn::Const::ERROR_413_RESPONSE
    when Unicorn::HttpParserError # try to tell the client they're bad
      Unicorn::Const::ERROR_400_RESPONSE
    else
      Unicorn.log_error(@logger, "app error", e)
      Unicorn::Const::ERROR_500_RESPONSE
    end
    client.kgio_trywrite(msg)
    client.close
    rescue
  end

  # once a client is accepted, it is processed in its entirety here
  # in 3 easy steps: read request, call app, write app response
  def process_client(client)
    status, headers, body = @app.call(env = @request.read(client))

    if 100 == status.to_i
      client.write(Unicorn::Const::EXPECT_100_RESPONSE)
      env.delete(Unicorn::Const::HTTP_EXPECT)
      status, headers, body = @app.call(env)
    end
    @request.headers? or headers = nil
    http_response_write(client, status, headers, body)
    client.close # flush and uncork socket immediately, no keepalive
  rescue => e
    handle_error(client, e)
  end

  # gets rid of stuff the worker has no business keeping track of
  # to free some resources and drops all sig handlers.
  # traps for USR1, USR2, and HUP may be set in the after_fork Proc
  # by the user.
  def init_worker_process(worker)
    QUEUE_SIGS.each { |sig| trap(sig, nil) }
    trap(:CHLD, 'DEFAULT')
    SIG_QUEUE.clear
    proc_name "worker[#{worker.nr}]"
    START_CTX.clear
    init_self_pipe!
    WORKERS.clear
    LISTENERS.each { |sock| sock.fcntl(Fcntl::F_SETFD, Fcntl::FD_CLOEXEC) }
    after_fork.call(self, worker) # can drop perms
    worker.user(*user) if user.kind_of?(Array) && ! worker.switched
    self.timeout /= 2.0 # halve it for select()
    @config = nil
    build_app! unless preload_app
  end

  def reopen_worker_logs(worker_nr)
    logger.info "worker=#{worker_nr} reopening logs..."
    Unicorn::Util.reopen_logs
    logger.info "worker=#{worker_nr} done reopening logs"
    init_self_pipe!
    rescue => e
      logger.error(e) rescue nil
      exit!(77) # EX_NOPERM in sysexits.h
  end

  # runs inside each forked worker, this sits around and waits
  # for connections and doesn't die until the parent dies (or is
  # given a INT, QUIT, or TERM signal)
  def worker_loop(worker)
    ppid = master_pid
    init_worker_process(worker)
    nr = 0 # this becomes negative if we need to reopen logs
    l = LISTENERS.dup
    ready = l.dup

    # closing anything we IO.select on will raise EBADF
    trap(:USR1) { nr = -65536; SELF_PIPE[0].close rescue nil }
    trap(:QUIT) { worker = nil; LISTENERS.each { |s| s.close rescue nil }.clear }
    [:TERM, :INT].each { |sig| trap(sig) { exit!(0) } } # instant shutdown
    logger.info "worker=#{worker.nr} ready"

    begin
      nr < 0 and reopen_worker_logs(worker.nr)
      nr = 0

      worker.tick = Time.now.to_i
      while sock = ready.shift
        if client = sock.kgio_tryaccept
          process_client(client)
          nr += 1
          worker.tick = Time.now.to_i
        end
        break if nr < 0
      end

      # make the following bet: if we accepted clients this round,
      # we're probably reasonably busy, so avoid calling select()
      # and do a speculative non-blocking accept() on ready listeners
      # before we sleep again in select().
      unless nr == 0 # (nr < 0) => reopen logs (unlikely)
        ready = l.dup
        redo
      end

      ppid == Process.ppid or return

      # timeout used so we can detect parent death:
      worker.tick = Time.now.to_i
      ret = IO.select(l, nil, SELF_PIPE, @timeout) and ready = ret[0]
    rescue Errno::EBADF
      nr < 0 or return
    rescue => e
      Unicorn.log_error(@logger, "listen loop error", e) if worker
    end while worker
  end

  # delivers a signal to a worker and fails gracefully if the worker
  # is no longer running.
  def kill_worker(signal, wpid)
    Process.kill(signal, wpid)
    rescue Errno::ESRCH
      worker = WORKERS.delete(wpid) and worker.close rescue nil
  end

  # delivers a signal to each worker
  def kill_each_worker(signal)
    WORKERS.keys.each { |wpid| kill_worker(signal, wpid) }
  end

  # unlinks a PID file at given +path+ if it contains the current PID
  # still potentially racy without locking the directory (which is
  # non-portable and may interact badly with other programs), but the
  # window for hitting the race condition is small
  def unlink_pid_safe(path)
    (File.read(path).to_i == $$ and File.unlink(path)) rescue nil
  end

  # returns a PID if a given path contains a non-stale PID file,
  # nil otherwise.
  def valid_pid?(path)
    wpid = File.read(path).to_i
    wpid <= 0 and return
    Process.kill(0, wpid)
    wpid
    rescue Errno::ESRCH, Errno::ENOENT
      # don't unlink stale pid files, racy without non-portable locking...
  end

  def load_config!
    loaded_app = app
    logger.info "reloading config_file=#{config.config_file}"
    config[:listeners].replace(@init_listeners)
    config.reload
    config.commit!(self)
    kill_each_worker(:QUIT)
    Unicorn::Util.reopen_logs
    self.app = orig_app
    build_app! if preload_app
    logger.info "done reloading config_file=#{config.config_file}"
  rescue StandardError, LoadError, SyntaxError => e
    Unicorn.log_error(@logger,
        "error reloading config_file=#{config.config_file}", e)
    self.app = loaded_app
  end

  # returns an array of string names for the given listener array
  def listener_names(listeners = LISTENERS)
    listeners.map { |io| sock_name(io) }
  end

  def build_app!
    if app.respond_to?(:arity) && app.arity == 0
      if defined?(Gem) && Gem.respond_to?(:refresh)
        logger.info "Refreshing Gem list"
        Gem.refresh
      end
      self.app = app.call
    end
  end

  def proc_name(tag)
    $0 = ([ File.basename(START_CTX[0]), tag
          ]).concat(START_CTX[:argv]).join(' ')
  end

  def redirect_io(io, path)
    File.open(path, 'ab') { |fp| io.reopen(fp) } if path
    io.sync = true
  end

  def init_self_pipe!
    SELF_PIPE.each { |io| io.close rescue nil }
    SELF_PIPE.replace(Kgio::Pipe.new)
    SELF_PIPE.each { |io| io.fcntl(Fcntl::F_SETFD, Fcntl::FD_CLOEXEC) }
  end

  def inherit_listeners!
    # inherit sockets from parents, they need to be plain Socket objects
    # before they become Kgio::UNIXServer or Kgio::TCPServer
    inherited = ENV['UNICORN_FD'].to_s.split(/,/).map do |fd|
      io = Socket.for_fd(fd.to_i)
      set_server_sockopt(io, listener_opts[sock_name(io)])
      IO_PURGATORY << io
      logger.info "inherited addr=#{sock_name(io)} fd=#{fd}"
      server_cast(io)
    end

    config_listeners = config[:listeners].dup
    LISTENERS.replace(inherited)

    # we start out with generic Socket objects that get cast to either
    # Kgio::TCPServer or Kgio::UNIXServer objects; but since the Socket
    # objects share the same OS-level file descriptor as the higher-level
    # *Server objects; we need to prevent Socket objects from being
    # garbage-collected
    config_listeners -= listener_names
    if config_listeners.empty? && LISTENERS.empty?
      config_listeners << Unicorn::Const::DEFAULT_LISTEN
      @init_listeners << Unicorn::Const::DEFAULT_LISTEN
      START_CTX[:argv] << "-l#{Unicorn::Const::DEFAULT_LISTEN}"
    end
    config_listeners.each { |addr| listen(addr) }
    raise ArgumentError, "no listeners" if LISTENERS.empty?
  end
end