Skip to content
Prometheus instrumentation for Sidekiq
Branch: master
Clone or download
Latest commit fe4c994 Mar 18, 2019
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
.chglog update git-chglog configuration Mar 18, 2019
bin Initial public release of sidekiq_prometheus Feb 8, 2019
lib chore: version to 0.9.0 for release Mar 18, 2019
spec refactor: don't raise if missing keys. let the client handle that Mar 18, 2019
.gitignore Initial public release of sidekiq_prometheus Feb 8, 2019
.rubocop.yml
.travis.yml Initial public release of sidekiq_prometheus Feb 8, 2019
CHANGELOG.md chore: update CHANGELOG Mar 18, 2019
CODE_OF_CONDUCT.md Initial public release of sidekiq_prometheus Feb 8, 2019
Gemfile Initial public release of sidekiq_prometheus Feb 8, 2019
LICENSE.txt
README.md doc: add badges to README Mar 18, 2019
Rakefile Initial public release of sidekiq_prometheus Feb 8, 2019
sidekiq_prometheus.gemspec loosen bundler pin for development dependency Feb 8, 2019

README.md

Sidekiq Prometheus

Status Gem MIT license

Prometheus Instrumentation for Sidekiq. This gem provides:

  • Sidekiq server middleware for reporting job metrics
  • Global metrics reporter that uses the Sidekiq API for reporting Sidekiq cluster stats (requires Sidekiq::Enterprise)
  • Sidecar Rack server to provide scrape-able endpoint for Prometheus. This allows for metrics to be reported without having to run a separate prometheus exporter process.

Installation

Add this line to your application's Gemfile:

gem 'sidekiq_prometheus'

And then execute:

$ bundle

Or install it yourself as:

$ gem install sidekiq_prometheus

Usage

To run with the defaults add this to your Sidekiq initializer

SidekiqPrometheus.setup

This will register metrics, start the global reporter (if available), and start the Rack server for scraping. The default port is 9359 but this is easily configurable.

Once Sidekiq server is running you can see your metrics or scrape them with Prometheus:

$curl http://localhost:9359/metrics

# TYPE sidekiq_job_count counter
# HELP sidekiq_job_count Count of Sidekiq jobs
# TYPE sidekiq_job_duration histogram
# HELP sidekiq_job_duration Sidekiq job processing duration

[etc]

Full documentation

Configuration

You can configure the gem by calling configure:

SidekiqPrometheus.configure do |config|
  config.base_labels = { service: 'kubernandos_api' }
end

configure will automatically call setup so

If you are running multiple services that will be reporting Sidekiq metrics you will want to take advantage of the base_labels configuration option. For example:

SidekiqPrometheus.configure do |config|
  config.base_labels  = { service: 'image_api' }
  config.metrics_port = 9090
end

Configuration options

  • base_labels: Hash of labels that will be included with every metric when they are registered.
  • gc_metrics_enabled: Boolean that determines whether to record object allocation metrics per job. The default is true. Setting this to false if you don't need this metric.
  • global_metrics_enabled: Boolean that determines whether to report global metrics from the PeriodicMetrics reporter. When true this will report on a number of stats from the Sidekiq API for the cluster. This requires Sidekiq::Enterprise as the reporter uses the leader election functionality to ensure that only one worker per cluster is reporting metrics.
  • periodic_metrics_enabled: Boolean that determines whether to run the periodic metrics reporter. PeriodicMetrics runs a separate thread that reports on global metrics (if enabled) as well worker GC stats (if enabled). It reports metrics on the interval defined by periodic_reporting_interval. Defaults to true.
  • periodic_reporting_interval: interval in seconds for reporting periodic metrics. Default: 30
  • metrics_port: Port on which the rack server will listen. Defaults to 9359
  • registry: An instance of Prometheus::Client::Registry. If you have a registry with defined metrics you can use this option to pass in your registry.
SidekiqPrometheus.configure do |config|
  config.base_labels                   = { service: 'myapp' }
  config.gc_metrics_enabled            = false
  config.global_metrics_enabled        = true
  config.periodic_metrics_enabled      = true
  config.periodic_reporting_interval   = 20
  config.metrics_port                  = 8675
end

Custom labels may be added by defining the prometheus_labels method in the worker class:

class SomeWorker
  include Sidekiq::Worker

  def prometheus_labels
    { some: 'label' }
  end
end

Metrics

JobMetrics

All Sidekiq job metrics are reported with these labels:

  • class: Sidekiq worker class name
  • queue: Sidekiq queue name
Metric Type Description
sidekiq_job_count counter Count of Sidekiq jobs
sidekiq_job_duration histogram Sidekiq job processing duration
sidekiq_job_success counter Count of successful Sidekiq jobs
sidekiq_job_allocated_objects histogram Count of ruby objects allocated by a Sidekiq job
sidekiq_job_failed counter Count of failed Sidekiq jobs

Notes:

  • when a job fails only sidekiq_job_count and sidekiq_job_failed will be reported.
  • sidekiq_job_allocated_objects will only be reported if SidekiqPrometheus.gc_metrics_enabled? == true

Periodic GC Metrics

These require SidekiqPrometheus.gc_metrics_enabled? == true and SidekiqPrometheus.periodic_metrics_enabled? == true

Metric Type Description
sidekiq_allocated_objects counter Count of allocated objects by the worker
sidekiq_heap_free_slots gauge Number of free heap slots as reported by GC.stat
sidekiq_heap_live_slots gauge Number of live heap slots as reported by GC.stat
sidekiq_major_gc_count counter Count of major GC runs
sidekiq_minor_gc_count counter Count of minor GC runs
sidekiq_rss gauge RSS memory usage for worker process

Periodic Global Metrics

These require SidekiqPrometheus.global_metrics_enabled? == true and SidekiqPrometheus.periodic_metrics_enabled? == true

Periodic metric reporting relies on Sidekiq Enterprise's leader election functionality (Ent Leader Election ) which ensures that metrics are only reported once per cluster.

Metric Type Description
sidekiq_workers_size gauge Total number of workers processing jobs
sidekiq_dead_size gauge Total Dead Size
sidekiq_enqueued gauge Total Size of all known queues
sidekiq_queue_latency summary Latency (in seconds) of all queues
sidekiq_failed gauge Number of job executions which raised an error
sidekiq_processed gauge Number of job executions completed (success or failure)
sidekiq_retry_size gauge Total Retries Size
sidekiq_scheduled_size gauge Total Scheduled Size
sidekiq_redis_connected_clients gauge Number of clients connected to Redis instance for Sidekiq
sidekiq_redis_used_memory gauge Used memory from Redis.info
sidekiq_redis_used_memory_peak gauge Used memory peak from Redis.info
sidekiq_redis_keys gauge Number of redis keys
sidekiq_redis_expires gauge Number of redis keys with expiry set

The global metrics are reported with the only the base_labels with the exception of sidekiq_enqueued which will add a queue label and record a metric per Sidekiq queue.

Custom Worker Metrics

There are a few different ways to register custom metrics with SidekiqPrometheus. Each custom metric should be defined as a Hash with the following form:

{
  name:        :metric_name,
  type:        :gauge,
  docstring:   'description',
  base_labels: { label_name: 'label_text' },
}
  • :name (required) - Unique name of the metric and should be a symbol.
  • :type (required) - Prometheus metric type. Supported values are: :counter, :gauge, :histogram, and :summary.
  • :docstring (required) - Human readable description of the metric.
  • :base_labels (optional) - Hash of labels that will be applied to every instance of this metric.

Registering custom metrics:

Registering a set of custom metrics is done by defining custom_metrics in the configure block:

SidekiqPrometheus.configure do |config|
  config.custom_metrics = [
    { name: :imported_records, type: :counter, :docstring: 'Count of successfully imported records' },
    { name: :failed_records,   type: counter:, :docstring: 'Count of failed records' },
  ]
end

Metrics can also be registered directly. This must done after SidekiqPrometheus.configure or setup has been run.

SidekiqPrometheus::Metrics.register(name: :logged_in_users, type: :gauge, docstring: 'Logged in users')

There is also a method to register more than one metric at a time:

customer_worker_metrics = [
  {
    name: :file_count, type: :counter, docstring: 'Number of active files',
    name: :file_size,  type: :gauge,   docstring: 'Size of files in bytes',    
  }
]

SidekiqPrometheus::Metrics.register_metrics(customer_worker_metrics)

Using custom metrics:

Once metrics are registered they can be used in your Sidekiq workers.

class ImportWorker
  include Sidekiq::Worker

  LABELS = {}

  def perform(*args)
    # worker code

    SidekiqPrometheus[:file_count].increment(LABELS, new_file_count)
    SidekiqPrometheus[:file_size].set(LABELS, total_file_size)
  end

end

See the documentation of the Prometheus::Client library for all of the available options for setting metric values.

Development

After checking out the repo, run bin/setup to install dependencies. Then, run rake test to run the tests. You can also run bin/console for an interactive prompt that will allow you to experiment.

To install this gem onto your local machine, run bundle exec rake install.

Contributing

Bug reports and pull requests are welcome on GitHub at https://github.com/fastly/sidekiq-prometheus. This project is intended to be a safe, welcoming space for collaboration, and contributors are expected to adhere to the Contributor Covenant code of conduct.

Copyright

Copyright 2019 Fastly, Inc.

License

The gem is available as open source under the terms of the MIT License.

Code of Conduct

Everyone interacting in the SidekiqPrometheus project’s codebases, issue trackers, chat rooms and mailing lists is expected to follow the code of conduct.

You can’t perform that action at this time.