# encoding: utf-8 require "logstash/namespace" require "logstash/outputs/base" require "logstash/errors" require "zlib" # This output writes events to files on disk. You can use fields # from the event as parts of the filename and/or path. # # By default, this output writes one event per line in **json** format. # You can customise the line format using the `line` codec like # [source,ruby] # output { # file { # path => ... # codec => line { format => "custom format: %{message}"} # } # } class LogStash::Outputs::File < LogStash::Outputs::Base concurrency :shared FIELD_REF = /%\{[^}]+\}/ config_name "file" attr_reader :failure_path # The path to the file to write. Event fields can be used here, # like `/var/log/logstash/%{host}/%{application}` # One may also utilize the path option for date-based log # rotation via the joda time format. This will use the event # timestamp. # E.g.: `path => "./test-%{+YYYY-MM-dd}.txt"` to create # `./test-2013-05-29.txt` # # If you use an absolute path you cannot start with a dynamic string. # E.g: `/%{myfield}/`, `/test-%{myfield}/` are not valid paths config :path, :validate => :string, :required => true # Flush interval (in seconds) for flushing writes to log files. # 0 will flush on every message. config :flush_interval, :validate => :number, :default => 2 # Gzip the output stream before writing to disk. config :gzip, :validate => :boolean, :default => false # If the generated path is invalid, the events will be saved # into this file and inside the defined path. config :filename_failure, :validate => :string, :default => '_filepath_failures' # If the configured file is deleted, but an event is handled by the plugin, # the plugin will recreate the file. Default => true config :create_if_deleted, :validate => :boolean, :default => true # Dir access mode to use. Note that due to the bug in jruby system umask # is ignored on linux: https://github.com/jruby/jruby/issues/3426 # Setting it to -1 uses default OS value. # Example: `"dir_mode" => 0750` config :dir_mode, :validate => :number, :default => -1 # File access mode to use. Note that due to the bug in jruby system umask # is ignored on linux: https://github.com/jruby/jruby/issues/3426 # Setting it to -1 uses default OS value. # Example: `"file_mode" => 0640` config :file_mode, :validate => :number, :default => -1 # How should the file be written? # # If `append`, the file will be opened for appending and each new event will # be written at the end of the file. # # If `overwrite`, the file will be truncated before writing and only the most # recent event will appear in the file. config :write_behavior, :validate => [ "overwrite", "append" ], :default => "append" default :codec, "json_lines" public def register require "fileutils" # For mkdir_p @files = {} @io_mutex = Mutex.new @path = File.expand_path(path) validate_path if path_with_field_ref? @file_root = extract_file_root else @file_root = File.dirname(path) end @failure_path = File.join(@file_root, @filename_failure) @flush_interval = @flush_interval.to_i if @flush_interval > 0 @flusher = Interval.start(@flush_interval, -> { flush_pending_files }) end @last_stale_cleanup_cycle = Time.now @stale_cleanup_interval = 10 end # def register private def validate_path if (root_directory =~ FIELD_REF) != nil @logger.error("File: The starting part of the path should not be dynamic.", :path => @path) raise LogStash::ConfigurationError.new("The starting part of the path should not be dynamic.") end end private def root_directory parts = @path.split(File::SEPARATOR).select { |item| !item.empty? } if Gem.win_platform? # First part is the drive letter parts[1] else parts.first end end public def multi_receive_encoded(events_and_encoded) encoded_by_path = Hash.new {|h,k| h[k] = []} events_and_encoded.each do |event,encoded| file_output_path = event_path(event) encoded_by_path[file_output_path] << encoded end @io_mutex.synchronize do encoded_by_path.each do |path,chunks| fd = open(path) if @write_behavior == "overwrite" fd.truncate(0) fd.seek(0, IO::SEEK_SET) fd.write(chunks.last) else # append to the file chunks.each {|chunk| fd.write(chunk) } end fd.flush unless @flusher && @flusher.alive? end close_stale_files end end # def receive public def close @flusher.stop unless @flusher.nil? @io_mutex.synchronize do @logger.debug("Close: closing files") @files.each do |path, fd| begin fd.close @logger.debug("Closed file #{path}", :fd => fd) rescue Exception => e @logger.error("Exception while flushing and closing files.", :exception => e) end end end end private def inside_file_root?(log_path) target_file = File.expand_path(log_path) return target_file.start_with?("#{@file_root.to_s}/") end private def event_path(event) file_output_path = generate_filepath(event) if path_with_field_ref? && !inside_file_root?(file_output_path) @logger.warn("File: the event tried to write outside the files root, writing the event to the failure file", :event => event, :filename => @failure_path) file_output_path = @failure_path elsif !@create_if_deleted && deleted?(file_output_path) file_output_path = @failure_path end @logger.debug("File, writing event to file.", :filename => file_output_path) file_output_path end private def generate_filepath(event) event.sprintf(@path) end private def path_with_field_ref? path =~ FIELD_REF end private def extract_file_root parts = File.expand_path(path).split(File::SEPARATOR) parts.take_while { |part| part !~ FIELD_REF }.join(File::SEPARATOR) end # the back-bone of @flusher, our periodic-flushing interval. private def flush_pending_files @io_mutex.synchronize do @logger.debug("Starting flush cycle") @files.each do |path, fd| @logger.debug("Flushing file", :path => path, :fd => fd) fd.flush end end rescue => e # squash exceptions caught while flushing after logging them @logger.error("Exception flushing files", :exception => e.message, :backtrace => e.backtrace) end # every 10 seconds or so (triggered by events, but if there are no events there's no point closing files anyway) private def close_stale_files now = Time.now return unless now - @last_stale_cleanup_cycle >= @stale_cleanup_interval @logger.debug("Starting stale files cleanup cycle", :files => @files) inactive_files = @files.select { |path, fd| not fd.active } @logger.debug("%d stale files found" % inactive_files.count, :inactive_files => inactive_files) inactive_files.each do |path, fd| @logger.info("Closing file %s" % path) fd.close @files.delete(path) end # mark all files as inactive, a call to write will mark them as active again @files.each { |path, fd| fd.active = false } @last_stale_cleanup_cycle = now end private def cached?(path) @files.include?(path) && !@files[path].nil? end private def deleted?(path) !File.exist?(path) end private def open(path) if !deleted?(path) && cached?(path) return @files[path] end if deleted?(path) if @create_if_deleted @logger.debug("Required path was deleted, creating the file again", :path => path) @files.delete(path) else return @files[path] if cached?(path) end end @logger.info("Opening file", :path => path) dir = File.dirname(path) if !Dir.exist?(dir) @logger.info("Creating directory", :directory => dir) if @dir_mode != -1 FileUtils.mkdir_p(dir, :mode => @dir_mode) else FileUtils.mkdir_p(dir) end end # work around a bug opening fifos (bug JRUBY-6280) stat = File.stat(path) rescue nil if stat && stat.ftype == "fifo" fd = java.io.FileWriter.new(java.io.File.new(path)) else if @file_mode != -1 fd = File.new(path, "a+", @file_mode) else fd = File.new(path, "a+") end end if gzip fd = Zlib::GzipWriter.new(fd) end @files[path] = IOWriter.new(fd) end ## # Bare-bones utility for running a block of code at an interval. # class Interval ## # Initializes a new Interval with the given arguments and starts it before returning it. # # @param interval [Integer] (see: Interval#initialize) # @param procsy [#call] (see: Interval#initialize) # # @return [Interval] # def self.start(interval, procsy) self.new(interval, procsy).tap(&:start) end ## # @param interval [Integer]: time in seconds to wait between calling the given proc # @param procsy [#call]: proc or lambda to call periodically; must not raise exceptions. def initialize(interval, procsy) @interval = interval @procsy = procsy require 'thread' # Mutex, ConditionVariable, etc. @mutex = Mutex.new @sleeper = ConditionVariable.new end ## # Starts the interval, or returns if it has already been started. # # @return [void] def start @mutex.synchronize do return if @thread && @thread.alive? @thread = Thread.new { run } end end ## # Stop the interval. # Does not interrupt if execution is in-progress. def stop @mutex.synchronize do @stopped = true end @thread && @thread.join end ## # @return [Boolean] def alive? @thread && @thread.alive? end private def run @mutex.synchronize do loop do @sleeper.wait(@mutex, @interval) break if @stopped @procsy.call end end ensure @sleeper.broadcast end end # class LogStash::Outputs::File::Interval end # class LogStash::Outputs::File # wrapper class class IOWriter def initialize(io) @io = io end def write(*args) @io.write(*args) @active = true end def flush @io.flush if @io.class == Zlib::GzipWriter @io.to_io.flush end end def method_missing(method_name, *args, &block) if @io.respond_to?(method_name) @io.send(method_name, *args, &block) else super end end attr_accessor :active end