Home

Awesome

librato-rack

Gem Version Build Status Code Climate


Important note on breaking change

NOTE: Starting with version 2.0.0 librato-rack requires a Librato account that supports tagged metrics.

If your Librato account doesn't yet support tagged metrics or you are using a heroku addon, please use the 1.x.x version.


librato-rack provides rack middleware which will report key statistics for your rack applications to Librato Metrics. It will also allow you to easily track your own custom metrics. Metrics are delivered asynchronously behind the scenes so they won't affect performance of your requests.

Currently Ruby 1.9.2+ is required.

Upgrading

Upgrading from version 1.x to 2.x introduces breaking changes for legacy sources. Please contact support@librato.com to migrate an existing Librato account.

Quick Start

Install librato-rack as middleware in your application:

use Librato::Rack

Configuring and relaunching your application will start the reporting of performance and request metrics. You can also track custom metrics by adding simple one-liners to your code:

# keep counts of key events
Librato.increment 'user.signup'

# benchmark sections of code to verify production performance
Librato.timing 'my.complicated.work' do
  # do work
end

# track averages across requests
Librato.measure 'user.social_graph.nodes', user.social_graph.size

Installation & Configuration

Install the gem:

$ gem install librato-rack

Or add to your Gemfile if using bundler:

gem "librato-rack"

In your rackup file or equivalent, require and add the middleware:

require 'librato-rack'
use Librato::Rack

In order to get the most accurate measurements, it is recommended that the librato-rack middleware be the first middleware in your stack. This will ensure that timing measurements like the rack.request.time metric will include all of the time spent in the Rack middleware stack.

If you don't have a Metrics account already, sign up. In order to send measurements to Metrics you need to provide your account credentials to librato-rack. You can provide these one of two ways:

Use environment variables

By default you can use LIBRATO_USER and LIBRATO_TOKEN to pass your account data to the middleware. While these are the only required variables, there are a few more optional environment variables you may find useful.

Use a configuration object

If you want to do more complex configuration, use your own environment variables, or control your configuration in code, you can use a configuration object:

config = Librato::Rack::Configuration.new
config.user = 'myuser@mysite.com'
config.token = 'mytoken'
# …more configuration

use Librato::Rack, :config => config

See the configuration class for all available options.

Running on Heroku

If you are using the Librato Metrics Heroku addon, your LIBRATO_USER and LIBRATO_TOKEN environment variables will already be set in your Heroku environment. If you are running without the addon you will need to provide them yourself.

NOTE: if Heroku idles your application no measurements will be sent until it receives another request and is restarted. If you see intermittent gaps in your measurements during periods of low traffic this is the most likely cause.

Default Tags

Librato Metrics supports tagged measurements that are associated with a metric, one or more tag pairs, and a point in time. For more information on tagged measurements, visit our API documentation.

Detected Tags

By default, host is detected and applied as a default tag for submitted measurements. Optionally, you can override the detected values, e.g. LIBRATO_TAGS=host=myapp-prod-1

Custom Tags

In addition to the default tags, you can also provide custom tags:

config = Librato::Rack::Configuration.new
config.user = 'myuser@mysite.com'
config.token = 'mytoken'
config.tags = { service: 'myapp', environment: 'production', host: 'myapp-prod-1' }

use Librato::Rack, :config => config
Metric Suites

The metrics recorded by librato-rack are organized into named metric suites that can be selectively enabled/disabled:

All three of the metric suites listed above are enabled by default.

The metric suites can be configured via either the LIBRATO_SUITES environment variable or the suites attributes on the Librato::Rack::Configuration object.

LIBRATO_SUITES="rack,rack_method"  # use ONLY the rack and rack_method suites
LIBRATO_SUITES="+foo,+bar"         # + prefix indicates that you want the default suites plus foo and bar
LIBRATO_SUITES="-rack_status"      # - prefix indicates that you want the default suites removing rack_status
LIBRATO_SUITES="+foo,-rack_status" # Use all default suites except for rack_status while also adding foo
LIBRATO_SUITES="all"               # Enable all suites
LIBRATO_SUITES="none"              # Disable all suites
LIBRATO_SUITES=""                  # Use only the default suites (same as if env var is absent)

Note that you should EITHER specify an explict list of suites to enable OR add/subtract individual suites from the default list (using the +/- prefixes). If you try to mix these two forms a Librato::Rack::InvalidSuiteConfiguration error will be raised.

Use with EventMachine and EM Synchrony

librato-rack has experimental support for EventMachine and EM Synchrony apps.

When using in an evented context set LIBRATO_EVENT_MODE to 'eventmachine' if using EventMachine or 'synchrony' if using EM Synchrony and/or Rack::FiberPool. We're interested in maturing this support, so please let us know if you have any issues.

Custom Measurements

Tracking anything that interests you is easy with Metrics. There are four primary helpers available:

increment

Use for tracking a running total of something across requests, examples:

# increment the 'sales_completed' metric by one
Librato.increment 'sales.completed'
# => {:host=>"myapp-prod-1"}

# increment by five
Librato.increment 'items.purchased', by: 5
# => {:host=>"myapp-prod-1"}

# increment with custom per-measurement tags
Librato.increment 'user.purchases', tags: { user_id: user.id, currency: 'USD' }
# => {:user_id=>43, :currency=>"USD"}

# increment with custom per-measurement tags and inherited default tags
Librato.increment 'user.purchases', tags: { user_id: user.id, currency: 'USD' }, inherit_tags: true
# => {:host=>"myapp-prod-1", :user_id=>43, :currency=>"USD"}

Other things you might track this way: user signups, requests of a certain type or to a certain route, total jobs queued or processed, emails sent or received

Sporadic Increment Reporting

Note that increment is primarily used for tracking the rate of occurrence of some event. Given this increment metrics are continuous by default: after being called on a metric once they will report on every interval, reporting zeros for any interval when increment was not called on the metric.

Especially with custom per-measurement tags you may want the opposite behavior - reporting a measurement only during intervals where increment was called on the metric:

# report a value for 'user.uploaded_file' only during non-zero intervals
Librato.increment 'user.uploaded_file', tags: { user_id: user.id, bucket: bucket.name }, sporadic: true

measure

Use when you want to track an average value per-request. Examples:

Librato.measure 'user.social_graph.nodes', 212

# report from custom per-measurement tags
Librato.measure 'jobs.queued', 3, tags: { priority: 'high', worker: 'worker.12' }

timing

Like Librato.measure this is per-request, but specialized for timing information:

Librato.timing 'twitter.lookup.time', 21.2

The block form auto-submits the time it took for its contents to execute as the measurement value:

Librato.timing 'twitter.lookup.time' do
  @twitter = Twitter.lookup(user)
end

percentiles

By defaults timings will send the average, sum, max and min for every minute. If you want to send percentiles as well you can specify them inline while instrumenting:

# track a single percentile
Librato.timing 'api.request.time', time, percentile: 95

# track multiple percentiles
Librato.timing 'api.request.time', time, percentile: [95, 99]

You can also use percentiles with the block form of timings:

Librato.timing 'my.important.event', percentile: 95 do
  # do work
end

group

There is also a grouping helper, to make managing nested metrics easier. So this:

Librato.measure 'memcached.gets', 20
Librato.measure 'memcached.sets', 2
Librato.measure 'memcached.hits', 18

Can also be written as:

Librato.group 'memcached' do |g|
  g.measure 'gets', 20
  g.measure 'sets', 2
  g.measure 'hits', 18
end

Symbols can be used interchangeably with strings for metric names.

Use with Background Workers / Cron Jobs

librato-rack is designed to run within a long-running process and report periodically. Intermittently running rake tasks and most background job tools (delayed job, resque, queue_classic) don't run long enough for this to work.

Never fear, we have some guidelines for how to instrument your workers properly.

If you are using librato-rack with sidekiq, see these notes about setup.

Cross-Process Aggregation

librato-rack submits measurements back to the Librato platform on a per-process basis. By default these measurements are then combined into a single measurement per default tags (detects host) before persisting the data.

For example if you have 4 hosts with 8 unicorn instances each (i.e. 32 processes total), on the Metrics site you'll find 4 data streams (1 per host) instead of 32. Current pricing applies after aggregation, so in this case you will be charged for 4 streams instead of 32.

Troubleshooting

Note that it may take 2-3 minutes for the first results to show up in your Metrics account after you have started your servers with librato-rack enabled and the first request has been received.

For more information about startup and submissions to the Metrics service you can set your log_level to debug. If you are having an issue with a specific metric, using trace will add the exact measurements being sent to your logs along with other details about librato-rack execution. Neither of these modes are recommended long-term in production as they will add significant volume to your log file and may slow operation somewhat.

Submission times are total time but submission I/O is non-blocking - your process will continue to handle requests during submissions.

If you are debugging setup locally you can set flush_interval to something shorter (say 10s) to force submission more frequently. Don't change your flush_interval in production as it will not result in measurements showing up more quickly, but may affect performance.

Contribution

Copyright

Copyright (c) 2013-2017 Librato Inc. See LICENSE for details.