prometheus_exporter 2.1.1 → 2.3.0

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: e5381fb78e6440d77a61269f3bf7c40a84b44a53594d687db4a683be3cb82725
-  data.tar.gz: 18216fbeae6978d5427fa9f97a67bbf3f0ee36e92091a2606b2dadcaa6497d83
+  metadata.gz: be510b790b625379993092d05de6b8c7f02e30e74ca0ca630ba5c65ca99385d4
+  data.tar.gz: 43d78c0a9ab134b504cc6e4a6f86d799f9650c5b5e7d7b0d62a33afe3e97374f
 SHA512:
-  metadata.gz: 626d1f7c05e5bf21bc499e69fd1ab0d73c94c5a565ac7eae81ec9d43fc93f6cf96b83a6fd72b25846e49371079a9c87a497e719d74c20dfa57c0fb5cbe4507e5
-  data.tar.gz: b43a57442d6326648cbc3539c50400556575e425756358d45ad39832e099671ca617ea66b6754d16d66693b9fbdf9045b96a95713b36f1f135ab8ab0e38a6d0b
+  metadata.gz: d71f41661e4af76837dcb7df1cba50b3a6150ffce3dddebb1dc1710bf9a69dafc6a10750955fda8e870ae62a835e2bbcac417c00256285e3fc4266f3e2194e8d
+  data.tar.gz: 27cf01bfec436b1d6ece387460beefe35d8cf9146ab1d24d27ffebaeeb306df34d7432a31c307aef26e8b59994796593910f90cfaa9982c00a6344a2601a486d

data/CHANGELOG

@@ -1,227 +1,404 @@
-2.1.1 - 2024-06-19
+# Changelog
+
+All notable changes to this project will be documented in this file.
+
+The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.1.0/),
+and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
+
+## [2.3.0] - 2025-08-01
+
+- FEATURE: Added puma_busy_threads metric that provides a holistic view of server workload by calculating (active threads - idle threads) + queued requests
+- FEATURE: adding marking_time and sweeping_time to GC metrics
+- DEV: bumped up minimum Ruby version to 3.2
+- FIX: better handling of malformed metrics
+- FIX: get controller labels from controller, not params
+- DEV: expose close on runner instance
+- FEATURE: improved puma metrics
+
+## [2.2.0] - 2024-12-05
+
+### Added
+
+- Feature: Add Dalli::Client memcache metrics for web_collector
+
+### Fixed
+
+- FIX: Ensure socket is closed when error is raised while opening socket
+
+## [2.1.1] - 2024-06-19
+
+### Added
 
 - FEATURE: improve good_job instrumentation
-- FIX: improve Ruby 3.X support
 - FEATURE: imstrumentation for malloc / oldmalloc increace in GC stats
 
-2.1.0 - 2024-01-08
+### Fixed
+
+- FIX: improve Ruby 3.X support
+
+## [2.1.0] - 2024-01-08
+
+### Added
 
 - FEATURE: good_job instrumentation
+
+### Changed
+
 - PERF: improve performance of histogram
 - DEV: use new metric collector pattern so we reuse code between collectors
 
-2.0.8 - 2023-01-20
+## [2.0.8] - 2023-01-20
+
+### Added
 
 - FEATURE: attempting to make our first docker release
 
-2.0.7 - 2023-01-13
+## [2.0.7] - 2023-01-13
 
+### Added
 - FEATURE: allow binding server to both ipv4 and v6
+
+### Fixed
+
 - FIX: expire stale sidekiq metrics
 
 
-2.0.6 - 2022-11-22
+## [2.0.6] - 2022-11-22
+
+### Fixed
 
 - FIX: use user specified labels over default in merge conflict
 - FIX: sidekiq stats collector memory leak
 
-2.0.5 - 2022-11-15
+## [2.0.5] - 2022-11-15
+
+### Fixed
 
 - FIX: regression :prepend style instrumentation not working correctly
 
-2.0.4 - 2022-11-10
+## [2.0.4] - 2022-11-10
+
+### Fixed
 
-- FIX/FEATURE: support for Redis 5 gem instrumentation
+- FIX support for Redis 5 gem instrumentation
 
-2.0.3 - 2022-05-23
+## [2.0.3] - 2022-05-23
+
+### Added
 
 - FEATURE: new ping endpoint for keepalive checks
+
+### Fixed
+
 - FIX: order histogram correctly for GCP support
 - FIX: improve sidekiq instrumentation
 
-2.0.2 - 2022-02-25
+## [2.0.2] - 2022-02-25
+
+### Fixed
 
 - FIX: runner was not requiring unicorn integration correctly leading to a crash
 
-2.0.1 - 2022-02-24
+## [2.0.1] - 2022-02-24
+
+### Fixed
 
 - FIX: ensure threads do not leak when calling #start repeatedly on instrumentation classes, this is an urgent patch for Puma integration
 
-2.0.0 - 2022-02-18
+## [2.0.0] - 2022-02-18
+
+### Added
 
 - FEATURE: Add per worker custom labels
 - FEATURE: support custom histogram buckets
+
+### Fixed
+
 - FIX: all metrics are exposing status label, and not only `http_requests_total`
+
+### Changed
+
 - BREAKING: rename all `http_duration` metrics to `http_request_duration` to match prometheus official naming conventions (See https://prometheus.io/docs/practices/naming/#metric-names).
 
-1.0.1 - 2021-12-22
+## [1.0.1] - 2021-12-22
+
+### Added
 
 - FEATURE: add labels to preflight requests
 - FEATURE: SidekiqStats metrics
+
+### Fixed
+
 - FIX: mintor refactors to Sidekiq metrics
 
-1.0.0 - 2021-11-23
+## [1.0.0] - 2021-11-23
+
+### Added
 
-- BREAKING: rename metrics to match prometheus official naming conventions (See https://prometheus.io/docs/practices/naming/#metric-names)
 - FEATURE: Sidekiq process metrics
 - FEATURE: Allow collecting web metrics as histograms
+
+### Fixed
+
 - FIX: logger improved for web server
 - FIX: Remove job labels from DelayedJob queues
 
-0.8.1 - 2021-08-04
+### Changed
+
+- BREAKING: rename metrics to match prometheus official naming conventions (See https://prometheus.io/docs/practices/naming/#metric-names)
+
+
+## [0.8.1] - 2021-08-04
+
+### Added
 
 - FEATURE: swap from hardcoded STDERR to logger pattern (see README for details)
 
-0.8.0 - 2021-07-05
+## [0.8.0] - 2021-07-05
+
+### Added
 
-- FIX: handle ThreadError more gracefully in cases where process shuts down
 - FEATURE: add job_name and queue_name labels to delayed job metrics
 - FEATURE: always scope puma metrics on hostname in collector
 - FEATURE: add customizable labels option to puma collector
 - FEATURE: support for Resque
-- DEV: Remove support for EOL ruby 2.5
-- FIX: Add source location to MethodProfiler patches
 - FEATURE: Improve Active Record instrumentation
 - FEATURE: Support HTTP_X_AMZN_TRACE_ID when supplied
 
-0.7.0 - 2020-12-29
+### Fixed
+
+- FIX: handle ThreadError more gracefully in cases where process shuts down
+- FIX: Add source location to MethodProfiler patches
+
+### Removed
+
+- DEV: Remove support for EOL ruby 2.5
+
+## [0.7.0] - 2020-12-29
+
+### Added
+
+- FEATURE: clean pattern for overriding middleware labels was introduced (in README)
+
+### Fixed
+
+- Fix: Better support for forking
+
+### Changed
 
 - Dev: Removed support from EOL rubies, only 2.5, 2.6, 2.7 and 3.0 are supported now.
 - Dev: Better support for Ruby 3.0, explicitly depending on webrick
 - Dev: Rails 6.1 instrumentation support
-- FEATURE: clean pattern for overriding middleware labels was introduced (in README)
-- Fix: Better support for forking
 
-0.6.0 - 2020-11-17
+## [0.6.0] - 2020-11-17
+
+### Added
 
 - FEATURE: add support for basic-auth in the prometheus_exporter web server
 
-0.5.3 - 2020-07-29
+## [0.5.3] - 2020-07-29
+
+### Added
 
 - FEATURE: added #remove to all metric types so users can remove specific labels if needed
 
-0.5.2 - 2020-07-01
+## [0.5.2] - 2020-07-01
+
+### Added
 
 - FEATURE: expanded instrumentation for sidekiq
 - FEATURE: configurable default labels
 
-0.5.1 - 2020-02-25
+## [0.5.1] - 2020-02-25
+
+### Added
 
 - FEATURE: Allow configuring the default client's host and port via environment variables
 
-0.5.0 - 2020-02-14
+## [0.5.0] - 2020-02-14
+
+### Fixed
 
-- Breaking change: listen only to localhost by default to prevent unintended insecure configuration
 - FIX: Avoid calling `hostname` aggressively, instead cache it on the exporter instance
 
-0.4.17 - 2020-01-13
+### Changed
+
+- Breaking change: listen only to localhost by default to prevent unintended insecure configuration
+
+## [0.4.17] - 2020-01-13
+
+### Added
 
 - FEATURE: add support for `to_h` on all metrics which can be used to query existing key/values
 
-0.4.16 - 2019-11-04
+## [0.4.16] - 2019-11-04
+
+### Added
 
 - FEATURE: Support #reset! on all metric types to reset a metric to default
 
-0.4.15 - 2019-11-04
+## [0.4.15] - 2019-11-04
+
+### Added
 
 - FEATURE: Improve delayed job collector, add pending counts
 - FEATURE: New ActiveRecord collector (documented in readme)
 - FEATURE: Allow passing in histogram and summary options
 - FEATURE: Allow custom labels for unicorn collector
 
-0.4.14 - 2019-09-10
+## [0.4.14] - 2019-09-10
+
+### Added
 
 - FEATURE: allow finding metrics by name RemoteMetric #find_registered_metric
+
+### Fixed
+
 - FIX: guard socket closing
 
-0.4.13 - 2019-07-09
+## [0.4.13] - 2019-07-09
+
+### Fixed
 
 - Fix: Memory leak in unicorn and puma collectors
 
-0.4.12 - 2019-05-30
+## [0.4.12] - 2019-05-30
+
+### Fixed
 
 - Fix: unicorn collector reporting incorrect number of unicorn workers
 
-0.4.11 - 2019-05-15
+## [0.4.11] - 2019-05-15
+
+### Fixed
 
 - Fix: Handle stopping nil worker_threads in Client
+
+### Changed
+
 - Dev: add frozen string literals
 
-0.4.10 - 2019-04-29
+## [0.4.10] - 2019-04-29
+
+### Fixed
 
 - Fix: Custom label support for puma collector
 - Fix: Raindrops socket collector not working correctly
 
-0.4.9 - 2019-04-11
+## [0.4.9] - 2019-04-11
+
+### Fixed
 
 - Fix: Gem was not working correctly in Ruby 2.4 and below due to a syntax error
 
-0.4.8 - 2019-04-10
+## [0.4.8] - 2019-04-10
+
+### Added
 
 - Feature: added helpers for instrumenting unicorn using raindrops
 
-0.4.7 - 2019-04-08
+## [0.4.7] - 2019-04-08
+
+### Fixed
 
 - Fix: collector was not escaping " \ and \n correctly. This could lead
   to a corrupt payload in some cases.
 
-0.4.6 - 2019-04-02
+## [0.4.6] - 2019-04-02
+
+### Added
 
 - Feature: Allow resetting a counter
 - Feature: Add sidekiq metrics: restarted, dead jobs counters
+
+### Fixed
+
 - Fix: Client shutting down before sending metrics to collector
 
-0.4.5 - 2019-02-14
+## [0.4.5] - 2019-02-14
+
+### Added
 
 - Feature: Allow process collector to ship custom labels for all process metrics
+
+### Fixed
+
 - Fix: Always scope process metrics on hostname in collector
 
-0.4.4 - 2019-02-13
+## [0.4.4] - 2019-02-13
+
+### Added
 
 - Feature: add support for local metric collection without using HTTP
 
-0.4.3 - 2019-02-11
+## [0.4.3] - 2019-02-11
+
+### Added
 
 - Feature: Add alias for Gauge #observe called #set, this makes it a bit easier to migrate from prom
 - Feature: Add increment and decrement to Counter
 
-0.4.2 - 2018-11-30
+## [0.4.2] - 2018-11-30
+
+### Fixed
 
-- Fix/Feature: setting a Gauge to nil will remove Gauge (setting to non numeric will raise)
+- Fix: setting a Gauge to nil will remove Gauge (setting to non numeric will raise)
 
-0.4.0 - 2018-10-23
+## [0.4.0] - 2018-10-23
+
+### Added
 
 - Feature: histogram support
 - Feature: custom quantile support for summary
 - Feature: Puma metrics
+
+### Fixed
+
 - Fix: delayed job metrics
 
-0.3.4 - 2018-10-02
+## [0.3.4] - 2018-10-02
+
+### Fixed
 
 - Fix: custom collector via CLI was not working correctly
 
-0.3.3
+## [0.3.3]
+
+### Added
 
 - Feature: Add more metrics to delayed job collector
 
-0.3.2
+## [0.3.2]
+
+### Added
 
 - Feature: Add posibility to set custom_labels on multi process mode
 
-0.3.1
+## [0.3.1]
+
+### Changed
 
 - Allow runner to accept a --timeout var
 - Allow runner to accept a blank prefix
 
-0.3.0
+## [0.3.0]
+
+### Changed
 
 - Breaking change: Follow Prometheus metric [naming conventions](https://prometheus.io/docs/practices/naming/#metric-names)
 
-0.1.15 - 2018-02-19
+## [0.1.15] - 2018-02-19
+
+### Added
 
 - Feature: Prefer to use oj if it is loadable
 
-0.1.14 - 2018-02-17
+## [0.1.14] - 2018-02-17
+
+### Added
 
-- Fix: error when shipping summary metric with no labels
 - Feature: runner was extracted so it can be reused @304
+
+### Fixed
+
+- Fix: error when shipping summary metric with no labels

data/README.md

@@ -98,8 +98,8 @@ server.collector.register_metric(counter)
 server.collector.register_metric(summary)
 server.collector.register_metric(histogram)
 
-gauge.observe(get_rss)
-gauge.observe(get_rss)
+gauge.observe(server.get_rss)
+gauge.observe(server.get_rss)
 
 counter.observe(1, route: 'test/route')
 counter.observe(1, route: 'another/route')
@@ -213,13 +213,14 @@ Rails.application.middleware.unshift PrometheusExporter::Middleware, instrument:
 
 #### Metrics collected by Rails integration middleware
 
-| Type    | Name                                   | Description                                                 |
-| ---     | ---                                    | ---                                                         |
-| Counter | `http_requests_total`                  | Total HTTP requests from web app                            |
-| Summary | `http_request_duration_seconds`        | Time spent in HTTP reqs in seconds                          |
-| Summary | `http_request_redis_duration_seconds`¹ | Time spent in HTTP reqs in Redis, in seconds                |
-| Summary | `http_request_sql_duration_seconds`²   | Time spent in HTTP reqs in SQL in seconds                   |
-| Summary | `http_request_queue_duration_seconds`³ | Time spent queueing the request in load balancer in seconds |
+| Type    | Name                                      | Description                                                 |
+| ---     | ---                                       | ---                                                         |
+| Counter | `http_requests_total`                     | Total HTTP requests from web app                            |
+| Summary | `http_request_duration_seconds`           | Time spent in HTTP reqs in seconds                          |
+| Summary | `http_request_redis_duration_seconds`¹    | Time spent in HTTP reqs in Redis, in seconds                |
+| Summary | `http_request_sql_duration_seconds`²      | Time spent in HTTP reqs in SQL in seconds                   |
+| Summary | `http_request_queue_duration_seconds`³    | Time spent queueing the request in load balancer in seconds |
+| Summary | `http_request_memcache_duration_seconds`⁴ | Time spent in HTTP reqs in Memcache in seconds              |
 
 All metrics have a `controller` and an `action` label.
 `http_requests_total` additionally has a (HTTP response) `status` label.
@@ -268,6 +269,7 @@ ruby_http_request_duration_seconds{path="/api/v1/teams/:id",method="GET",status=
 ¹) Only available when Redis is used.
 ²) Only available when Mysql or PostgreSQL are used.
 ³) Only available when [Instrumenting Request Queueing Time](#instrumenting-request-queueing-time) is set up.
+⁴) Only available when Dalli is used.
 
 #### Activerecord Connection Pool Metrics
 
@@ -371,6 +373,8 @@ end
 | Counter | `major_gc_ops_total`      | Major GC operations by process               |
 | Counter | `minor_gc_ops_total`      | Minor GC operations by process               |
 | Counter | `allocated_objects_total` | Total number of allocated objects by process |
+| Gauge   | `marking_time`            | Marking time spent (Ruby 3.3 minimum)        |
+| Gauge   | `sweeping_time`           | Sweeping time spent (Ruby 3.3 minimum)       |
 
 _Metrics marked with * are only collected when `MiniRacer` is defined._
 
@@ -607,15 +611,16 @@ end
 
 #### Metrics collected by Puma Instrumentation
 
-| Type  | Name                        | Description                                                 |
-| ---   | ---                         | ---                                                         |
-| Gauge | `puma_workers`              | Number of puma workers                                      |
-| Gauge | `puma_booted_workers`       | Number of puma workers booted                               |
-| Gauge | `puma_old_workers`          | Number of old puma workers                                  |
-| Gauge | `puma_running_threads`      | Number of puma threads currently running                    |
-| Gauge | `puma_request_backlog`      | Number of requests waiting to be processed by a puma thread |
-| Gauge | `puma_thread_pool_capacity` | Number of puma threads available at current scale           |
-| Gauge | `puma_max_threads`          | Number of puma threads at available at max scale            |
+| Type  | Name                        | Description                                                                                                         |
+| ---   | ---                         | ---                                                                                                                 |
+| Gauge | `puma_workers`              | Number of puma workers                                                                                              |
+| Gauge | `puma_booted_workers`       | Number of puma workers booted                                                                                       |
+| Gauge | `puma_old_workers`          | Number of old puma workers                                                                                          |
+| Gauge | `puma_running_threads`      | How many threads are spawned. A spawned thread may be busy processing a request or waiting for a new request        |
+| Gauge | `puma_request_backlog`      | Number of requests waiting to be processed by a puma thread                                                         |
+| Gauge | `puma_thread_pool_capacity` | Number of puma threads available at current scale                                                                   |
+| Gauge | `puma_max_threads`          | Number of puma threads at available at max scale                                                                    |
+| Gauge | `puma_busy_threads`         | Running - how many threads are waiting to receive work + how many requests are waiting for a thread to pick them up |
 
 All metrics may have a `phase` label and all custom labels provided with the `labels` option.
 

data/lib/prometheus_exporter/client.rb

@@ -17,13 +17,7 @@ module PrometheusExporter
       end
 
       def standard_values(value, keys, prometheus_exporter_action = nil)
-        values = {
-          type: @type,
-          help: @help,
-          name: @name,
-          keys: keys,
-          value: value
-        }
+        values = { type: @type, help: @help, name: @name, keys: keys, value: value }
         values[
           :prometheus_exporter_action
         ] = prometheus_exporter_action if prometheus_exporter_action
@@ -59,16 +53,14 @@ module PrometheusExporter
 
     def initialize(
       host: ENV.fetch("PROMETHEUS_EXPORTER_HOST", "localhost"),
-      port: ENV.fetch(
-        "PROMETHEUS_EXPORTER_PORT",
-        PrometheusExporter::DEFAULT_PORT
-      ),
+      port: ENV.fetch("PROMETHEUS_EXPORTER_PORT", PrometheusExporter::DEFAULT_PORT),
       max_queue_size: nil,
       thread_sleep: 0.5,
       json_serializer: nil,
       custom_labels: nil,
       logger: Logger.new(STDERR),
-      log_level: Logger::WARN
+      log_level: Logger::WARN,
+      process_queue_once_and_stop: false
     )
       @logger = logger
       @logger.level = log_level
@@ -83,9 +75,7 @@ module PrometheusExporter
       max_queue_size ||= MAX_QUEUE_SIZE
       max_queue_size = max_queue_size.to_i
 
-      if max_queue_size <= 0
-        raise ArgumentError, "max_queue_size must be larger than 0"
-      end
+      raise ArgumentError, "max_queue_size must be larger than 0" if max_queue_size <= 0
 
       @max_queue_size = max_queue_size
       @host = host
@@ -94,10 +84,10 @@ module PrometheusExporter
       @mutex = Mutex.new
       @thread_sleep = thread_sleep
 
-      @json_serializer =
-        json_serializer == :oj ? PrometheusExporter::OjCompat : JSON
+      @json_serializer = json_serializer == :oj ? PrometheusExporter::OjCompat : JSON
 
       @custom_labels = custom_labels
+      @process_queue_once_and_stop = process_queue_once_and_stop
     end
 
     def custom_labels=(custom_labels)
@@ -105,14 +95,7 @@ module PrometheusExporter
     end
 
     def register(type, name, help, opts = nil)
-      metric =
-        RemoteMetric.new(
-          type: type,
-          name: name,
-          help: help,
-          client: self,
-          opts: opts
-        )
+      metric = RemoteMetric.new(type: type, name: name, help: help, client: self, opts: opts)
       @metrics << metric
       metric
     end
@@ -163,7 +146,7 @@ module PrometheusExporter
           @socket.write("\r\n")
         rescue => e
           logger.warn "Prometheus Exporter is dropping a message: #{e}"
-          @socket = nil
+          close_socket!
           raise
         end
       end
@@ -189,6 +172,11 @@ module PrometheusExporter
     end
 
     def ensure_worker_thread!
+      if @process_queue_once_and_stop
+        worker_loop
+        return
+      end
+
       unless @worker_thread&.alive?
         @mutex.synchronize do
           return if @worker_thread&.alive?
@@ -253,8 +241,7 @@ module PrometheusExporter
 
       nil
     rescue StandardError
-      @socket = nil
-      @socket_started = nil
+      close_socket!
       @socket_pid = nil
       raise
     end
@@ -262,10 +249,7 @@ module PrometheusExporter
     def wait_for_empty_queue_with_timeout(timeout_seconds)
       start_time = ::Process.clock_gettime(::Process::CLOCK_MONOTONIC)
       while @queue.length > 0
-        if start_time + timeout_seconds <
-             ::Process.clock_gettime(::Process::CLOCK_MONOTONIC)
-          break
-        end
+        break if start_time + timeout_seconds < ::Process.clock_gettime(::Process::CLOCK_MONOTONIC)
         sleep(0.05)
       end
     end