Librato Metrics

A convenient Ruby wrapper for the Librato Metrics API.

NOTE: Starting with version 2.1.0 librato-metrics requires a Librato account that supports tagged metrics. If your Librato account doesn't yet support tagged metrics please use the 1.6.1 version.

This gem provides granular control for scripting interactions with the Metrics core API. It is well suited for integrations, scripts, workers & background jobs. If you want to submit from a web app, take at look at librato-rails and/or librato-rack.

Installation

In your shell:

gem install librato-metrics

Then, in your application or script:

require 'librato/metrics'

Optional steps

For best performance we recommend installing yajl-ruby:

gem install yajl-ruby

If you are using jruby, you need to ensure jruby-openssl is available:

gem install jruby-openssl

Quick Start

If you are looking for the quickest possible route to getting a data into Metrics, you only need two lines:

Librato::Metrics.authenticate 'email', 'api_key'
Librato::Metrics.submit my_metric: 42, my_other_metric: 1002

Unspecified metrics will send a gauge, but if you need to send a different metric type or include additional properties, simply use a hash:

Librato::Metrics.submit my_metric: {type: :counter, value: 1002, source: 'myapp'}

While this is all you need to get started, if you are sending a number of metrics regularly a queue may be easier/more performant so read on...

Authentication

Make sure you have an account for Metrics and then authenticate with your email and API key (on your account page):

Librato::Metrics.authenticate 'email', 'api_key'

Sending Measurements

If you are sending very many measurements or sending them very often, it will be much higher performance to bundle them up together to reduce your request volume. Use Queue for this.

Queue up a simple gauge metric named temperature:

queue = Librato::Metrics::Queue.new
queue.add temperature: 32.2

While symbols are used by convention for metric names, strings will work just as well:

queue.add 'myapp.request_time' => 86.7

If you are tracking measurements over several seconds/minutes, the queue will handle storing measurement time for you (otherwise all metrics will be recorded as measured when they are submitted).

If you want to specify a time other than queuing time for the measurement:

# use a epoch integer
queue.add humidity: {measure_time: 1336508422, value: 48.2}

# use a Time object to correct for a 5 second delay
queue.add humidity: {measure_time: Time.now-5, value: 37.6}

You can queue multiple metrics at once. Here's a gauge (load) and a counter (visits):

queue.add load: 2.2, visits: {type: :counter, value: 400}

Queue up a metric with a specified source:

queue.add cpu: {source: 'app1', value: 92.6}

A complete list of metric attributes is available in the API documentation.

Send currently queued measurements to Metrics:

queue.submit

Aggregate Measurements

If you are measuring something very frequently e.g. per-request in a web application (order mS) you may not want to send each individual measurement, but rather periodically send a single aggregate measurement, spanning multiple seconds or even minutes. Use an Aggregator for this.

Aggregate a simple gauge metric named response_latency:

aggregator = Librato::Metrics::Aggregator.new
aggregator.add response_latency: 85.0
aggregator.add response_latency: 100.5
aggregator.add response_latency: 150.2
aggregator.add response_latency: 90.1
aggregator.add response_latency: 92.0

Which would result in a gauge measurement like:

{name: "response_latency", count: 5, sum: 517.8, min: 85.0, max: 150.2}

You can specify a source during aggregate construction:

aggregator = Librato::Metrics::Aggregator.new(source: 'foobar')

You can aggregate multiple metrics at once:

aggregator.add app_latency: 35.2, db_latency: 120.7

Send the currently aggregated metrics to Metrics:

aggregator.submit

Benchmarking

If you have operations in your application you want to record execution time for, both Queue and Aggregator support the #time method:

aggregator.time :my_measurement do
  # do work...
end

The difference between the two is that Queue submits each timing measurement individually, while Aggregator submits a single timing measurement spanning all executions.

If you need extra attributes for a Queue timing measurement, simply add them on:

queue.time :my_measurement, source: 'app1' do
  # do work...
end

Annotations

Annotation streams are a great way to track events like deploys, backups or anything else that might affect your system. They can be overlaid on any other metric stream so you can easily see the impact of changes.

At a minimum each annotation needs to be assigned to a stream and to have a title. Let's add an annotation for deploying v45 of our app to the deployments stream:

Librato::Metrics.annotate :deployments, 'deployed v45'

There are a number of optional fields which can make annotations even more powerful:

Librato::Metrics.annotate :deployments, 'deployed v46', source: 'frontend',
    start_time: 1354662596, end_time: 1354662608,
    description: 'Deployed 6f3bc6e67682: fix lotsa bugs…'

You can also automatically annotate the start and end time of an action by using annotate's block form:

Librato::Metrics.annotate :deployments, 'deployed v46' do
  # do work..
end

More fine-grained control of annotations is available via the Annotator object:

annotator = Librato::Metrics::Annotator.new

# list annotation streams
streams = annotator.list

# fetch a list of events in the last hour from a stream
annotator.fetch :deployments, start_time: (Time.now.to_i-3600)

# delete an event
annotator.delete_event 'deployments', 23

See the documentation of Annotator for more details and examples of use.

Auto-Submitting Metrics

Both Queue and Aggregator support automatically submitting measurements on a given time interval:

# submit once per minute
timed_queue = Librato::Metrics::Queue.new(autosubmit_interval: 60)

# submit every 5 minutes
timed_aggregator = Librato::Metrics::Aggregator.new(autosubmit_interval: 300)

Queue also supports auto-submission based on measurement volume:

# submit when the 400th measurement is queued
volume_queue = Librato::Metrics::Queue.new(autosubmit_count: 400)

These options can also be combined for more flexible behavior.

Both options are driven by the addition of measurements. If you are adding measurements irregularly (less than once per second), time-based submission may lag past your specified interval until the next measurement is added.

If your goal is to collect metrics every x seconds and submit them, check out this code example.

Submitting tagged measurements

Librato Metrics supports tagged measurements that are associated with a metric, one or more tag pairs, and a point in time.

Tags are a set of name=value tag pairs that describe the particular data stream. Tags behave as extra dimensions that data streams can be filtered and aggregated along.

Top-Level Tags

You can initialize Queue and/or Aggregator with top-level tags that will be applied to every measurement:

queue = Librato::Metrics::Queue.new(tags: { service: 'auth', environment: 'prod', host: 'auth-prod-1' })
queue.add my_metric: 10

Per-Measurement Tags

Optionally, you can submit per-measurement tags by passing a tags Hash when adding measurements:

queue.add my_other_metric: { value: 25, tags: { db: 'rr1' } }

For more information, visit the API documentation.

Querying Metrics

Get name and properties for all metrics you have in the system:

metrics = Librato::Metrics.metrics

Get only metrics whose name includes time:

metrics = Librato::Metrics.metrics name: 'time'

Querying Metric Data

Get attributes for metric temperature:

data = Librato::Metrics.get_metric :temperature

Get the 20 most recent data points for temperature:

data = Librato::Metrics.get_measurements :temperature, count: 20

Get the 20 most recent data points for temperature from a specific source:

data = Librato::Metrics.get_measurements :temperature, count: 20, source: 'app1'

Get the 20 most recent 15 minute data point rollups for temperature:

data = Librato::Metrics.get_measurements :temperature, count: 20, resolution: 900

Get the 5 minute moving average for temperature for the last hour, assuming temperature is submitted once per minute:

data = Librato::Metrics.get_composite 'moving_average(mean(series("temperature", "*"), {size: "5"}))', start_time: Time.now.to_i - 60*60, resolution: 300

There are many more options supported for querying, take a look at the REST API docs or the individual method documentation for more details.

Retrieving tagged measurements (beta)

Get the series for exceptions in production grouped by sum within the last hour:

query = {
  resolution: 1,
  duration: 3600,
  group_by: "environment",
  group_by_function: "sum",
  tags_search: "environment=prod*"
}
Librato::Metrics.get_series :exceptions, query

For more information, visit the API documentation.

Setting Metric Properties

Setting custom properties on your metrics is easy:

# assign a period and default color
Librato::Metrics.update_metric :temperature, period: 15, attributes: { color: 'F00' }

It is also possible to update properties for multiple metrics at once, see the #update_metric method documentation for more information.

Deleting Metrics

If you ever need to remove a metric and all of its measurements, doing so is easy:

# delete the metrics 'temperature' and 'humidity'
Librato::Metrics.delete_metrics :temperature, :humidity

You can also delete using wildcards:

# delete metrics that start with cpu. except for cpu.free
Librato::Metrics.delete_metrics names: 'cpu.*', exclude: ['cpu.free']

Note that deleted metrics and their measurements are unrecoverable, so use with care.

Using Multiple Accounts Simultaneously

If you need to use metrics with multiple sets of authentication credentials simultaneously, you can do it with Client:

joe = Librato::Metrics::Client.new
joe.authenticate 'email1', 'api_key1'

mike = Librato::Metrics::Client.new
mike.authenticate 'email2', 'api_key2'

All of the same operations you can call directly from Librato::Metrics are available per-client:

# list Joe's metrics
joe.metrics

# fetch the last 20 data points for Mike's metric, humidity
mike.get_measurements :humidity, count: 20

There are two ways to associate a new queue with a client:

# these are functionally equivalent
joe_queue = Librato::Metrics::Queue.new(client: joe)
joe_queue = joe.new_queue

Once the queue is associated you can use it normally:

joe_queue.add temperature: {source: 'sf', value: 65.2}
joe_queue.submit

Thread Safety

The librato-metrics gem currently does not do internal locking for thread safety. When used in multi-threaded applications, please add your own mutexes for sensitive operations.

More Information

librato-metrics is sufficiently complex that not everything can be documented in the README. Additional options are documented regularly in the codebase. You are encouraged to take a quick look through the docs and source for more.

We also maintain a set of examples of common uses and appreciate contributions if you have them.

Contribution

Check out the latest master to make sure the feature hasn't been implemented or the bug hasn't been fixed yet.
Check out the issue tracker to make sure someone already hasn't requested it and/or contributed it.
Fork the project and submit a pull request from a feature or bugfix branch.
Please review our code conventions.
Please include specs. This is important so we don't break your changes unintentionally in a future version.
Please don't modify the gemspec, Rakefile, version, or changelog. If you do change these files, please isolate a separate commit so we can cherry-pick around it.