Client side library for logjam.
Hooks into Rails, collects log lines, performance metrics, error/exception infomation and Rack environment information and sends this data to Logjam.
Has experimental support for Sinatra.
Currently only one mechanism is available for data transport: ZeroMQ. Support for AMQP has been dropped.
For ZeroMQ, add
gem "logjam_agent"
gem "ffi-rzmq"
to your Gemfile.
Add an initializer config/initializers/logjam_agent.rb
to your app
and configure class LogjamAgent
.
module LogjamAgent
# Configure the application name (required). Must not contain dots or hyphens.
self.application_name = "myapp"
# Configure the environment name (optional). Defaults to Rails.env.
# self.environment_name = Rails.env
# Configure the application revision (optional). Defaults to (git rev-parse HEAD).
# self.application_revision = "f494e11afa0738b279517a2a96101a952052da5d"
# Configure request data forwarder for ZeroMQ. Default options as given below.
# The host parameter can be a comma separted list of zmq connection specifictions,
# where the protocol prefix and port suffix are optional. rcv_timeo and
# snd_timeo options only apply for sychronous messages.
add_forwarder(:zmq,
:host => "localhost",
:port => 9604,
:linger => 1000,
:snd_hwm => 100,
:rcv_timeo => 5000,
:snd_timeo => 5000)
# Configure ip obfuscation. Defaults to no obfuscation.
self.obfuscate_ips = true
# Configure cookie obfuscation. Defaults to [/_session\z/].
self.obfuscated_cookies = [/_session\z/]
# Configure asset request logging and forwarding. Defaults to ignore
# asset requests in development mode. Set this to false if you need
# to debug asset request handling.
self.ignore_asset_requests = Rails.env.development?
# Configure a list of URL patterns for which no data should be sent
# to logjam and nothing should be sent to the log device. Please not
# that the log lines will still show up in dev mode on the console.
# Defaults to the empty list.
# self.ignored_request_urls = [%r{/_system/}]
# Disable ActiveSupport::Notifications (and thereby logging) of ActionView
# render events. Defaults to false.
# self.ignore_render_events = Rails.env.production?
# Configure log level for logging on disk: only lines with a log level
# greater than or equal to the specified one will be logged to disk.
# Defaults to Logger::INFO. Note that logjam_agent extends the standard
# logger log levels by the constant NONE, which indicates no logging.
# Also, setting the level has no effect on console logging in development.
# self.log_device_log_level = Logger::WARN # log warnings, errors, fatals and unknown log messages
# self.log_device_log_level = Logger::NONE # log nothing at all
# Configure lines which will not be logged locally.
# They will still be sent to the logjam server. Defaults to nil.
self.log_device_ignored_lines = /^\s*Rendered/
# It is also possible to ovveride this on a per request basis,
# for example in a Rails before_action
# LogjamAgent.request.log_device_ignored_lines = /^\s*(?:Rendered|REDIS)/
# Configure maximum size of logged parameters and environment variables sent to
# logjam. Defaults to 1024.
# self.max_logged_param_size = 1024
# Configure maximum size of logged parameters and environment variables sent to
# logjam. Defaults to 1024 * 100.
# self.max_logged_cookie_size = 1024 * 100
# Configure maximum log line length. Defaults to 2048.
# This setting only applies to the lines sent with the request.
self.max_line_length = 2048
# Configure max bytes allowed for all log lines. Defaults to 1Mb.
# This setting only applies to the lines sent with the request.
self.max_bytes_all_lines = 1024 * 1024
# Configure compression method. Defaults to NO_COMPRESSION. Available
# compression methods are ZLIB_COMPRESSION, SNAPPY_COMPRESSION, LZ4_COMPRESSION.
# Snappy and LZ4 are faster and less CPU intensive than ZLIB, ZLIB achieves
# higher compression rates. LZ4 is faster to decompress than Snappy
# and recommended.
# self.compression_method = ZLIB_COMPRESSION
# self.compression_method = SNAPPY_COMPRESSION
# self.compression_method = LZ4_COMPRESSION
# Activate the split between hard and soft-exceptions. Soft exceptions are
# all exceptions below a log level of Logger::ERROR. Logjam itself can then
# display those soft exceptions differently. Defaults to `true`.
# self.split_hard_and_soft_exceptions = true
end
The gem ships with a railtie that sets up a logjam agent specific logger which serves the dual purpose of collecting log lines to be published on the logjam logging bus once the request is complete and sending them to the log device as they are produced.
The logger setup code in the Rails environment specific initialization files should be
removed. Setup code which can remain without affecting logjam functionality is setting the
log level or changing the log formatter. It is advisable to read the logger initialization
code contained in file lib/logjam_agent/railtie.rb
to understand what's going on behind
the scenes.
The agent generates unique request ids for all request handled using standard
SecureRandom
class shipped with Ruby.
The agent will try to use the Oj to
generate JSON. If this is not available in your application, it will
fall back to the to_json
method.
Supports both classic and modular Sinatra applications. Since Sinatra doesn't have built in action names like Rails, you'll have to declare them in your handlers, or in a before filter. Example:
require 'logjam_agent/sinatra'
use LogjamAgent::Sinatra::Middleware
class SinatraTestApp < Sinatra::Base
register LogjamAgent::Sinatra
configure do
set :loglevel, :debug
setup_logjam_logger
LogjamAgent.application_name = "myapp"
LogjamAgent.add_forwarder(:zmq, :host => "my-logjam-broker")
LogjamAgent.parameter_filters << :password
end
before '/index' do
action_name "Simple#index"
end
get '/index' do
logger.info 'Hello World!'
'Hello World!'
end
end
The environment name is picked up from either the environment variable LOGJAM_ENV
, or
Sinatra's environment setting.
Set the environment variable APP_LOG_TO_STDOUT
if you want to log to STDOUT
.
Otherwise, logs will appear in the subdirectory log
of your application's root.
The agent adds log lines to the request information sent via ZMQ on the logging bus and also writes log lines to the configured log device.
Using one of the logjam helpers LogjamAgent.logjam_only
or
LogjamAgent.logdevice_only
it is possible to send information to
only one of those log line sinks for the duration of a given block.
When LogjamAgent.log_format
is set to :json
, the logjam agent
logger will make sure that the content of line logged to the log
device is converted to JSON.
Rails.logger.info("foobar is great!")
# ----->
{ "message": "foobar is great!"}
Rails.logger.info(message: "foobar", user: 5)
# ----->
{ "message": "foobar", "user": 5 }
e = Standard::Error.new()
Rails.logger.info(e)
# ----->
{ "message": "#{e.message}", "error": <execption with formatted backtrace> }
If the agent experiences problems when sending data, it will log information to a file named
logjam_agent_error.log
which you can find under Rails.root/log
.
If you set the RAILS_LOG_TO_STDOUT
environment variable, those logs will be available through stderr
.
This behavior is customizable via a module level call back method:
LogjamAgent.error_handler = lambda {|exception| ... }