Class: StandardHealth::Notifiers::Metrics

Inherits:
Object
  • Object
show all
Defined in:
lib/standard_health/notifiers/metrics.rb

Overview

Emits per-check and per-evaluation metrics.

Unlike the Logger and Sentry notifiers, this one fires on EVERY poll on purpose. Counters and distributions are cheap, they aggregate, and the poll rate is exactly what makes them useful: it is what lets you chart p95 check latency and see "database check went from 3ms to 400ms an hour before the outage".

latency_ms has been computed on every check since v0.1.0 and thrown away into the response body. This is where it finally goes somewhere.

Sentry::Metrics is a SOFT dependency — guarded by defined?, never a gemspec entry.

Instance Method Summary collapse

Constructor Details

#initialize(metric_prefix: "health") ⇒ Metrics

Returns a new instance of Metrics.



19
20
21
 
# File 'lib/standard_health/notifiers/metrics.rb', line 19

def initialize(metric_prefix: "health")
  @prefix = metric_prefix
end

Instance Method Details

#call(event_name, payload) ⇒ Object 



23
24
25
26
27
28
29
30
31
32
33
34
 
# File 'lib/standard_health/notifiers/metrics.rb', line 23

def call(event_name, payload)
  return unless metrics_available?

  case event_name
  when "standard_health.check.completed" then record_check(payload)
  when "standard_health.ready.evaluated" then record_evaluation(payload)
  when "standard_health.check.timed_out" then record_timeout(payload)
  end
rescue StandardError
  # Observability must never break the health path.
  nil
end