gitlab-exporter 10.4.0 → 10.5.0

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: ace0ad673d8d36d1a6dbbacefb7f6aefde7d795e76eda8d3788c4cb4f5eb5247
-  data.tar.gz: ac42d2e6deaaa3fee45ed74ec1cec81f402abba0c4f14798878b2b006c72400b
+  metadata.gz: 4c392268186f090b025d290c818f821120f5160a9a59f2e3ba228f26da0b8505
+  data.tar.gz: 3771d5912cb2a310183a1f78a8e24d191eb502d20c2beab3c327994243ae7077
 SHA512:
-  metadata.gz: d3ac36c2b2ec584cb8ac60355649eb79d5a3d96e85a266f14794627fc3a9fa61a5cfb231f6c33e044cbb2e71d7965c50463b6673f80c69de4ebca58d658280e6
-  data.tar.gz: 43323ec5663db9dd865e84b3c7907ba43f805a958904cfef57d7566b6259a31cc1a7aa96c494ffb976b4b04148c2fbfe63bd688844672ad1938c41bf05181a4b
+  metadata.gz: c9ee25b554e51484e4cba96070677c90e3cbad17da49eef2e2f649ad4d4ff125e8e2982cfb6e952444436d7ead6f603f6173ba1270b7112b3de75126cb57b318
+  data.tar.gz: e6038eff74531eb7c0561437cb676336853d9a597b588d07a9561a4561ef165a3ad2f528711b3a2f5d61ebfc8df341ef93ec4d15274fb7efcc88859abc1bc6c3

data/.gitlab-ci.yml

@@ -6,6 +6,10 @@ include:
   - template: Security/SAST.gitlab-ci.yml  # https://gitlab.com/gitlab-org/gitlab-foss/blob/master/lib/gitlab/ci/templates/Security/SAST.gitlab-ci.yml
   - template: Security/Secret-Detection.gitlab-ci.yml # https://gitlab.com/gitlab-org/gitlab-foss/-/blob/master/lib/gitlab/ci/templates/Security/Secret-Detection.gitlab-ci.yml
 
+stages:
+  - test
+  - dast
+
 default:
   image: ruby:2.7
   cache:

data/Gemfile.lock

@@ -1,7 +1,7 @@
 PATH
   remote: .
   specs:
-    gitlab-exporter (10.4.0)
+    gitlab-exporter (10.5.0)
       connection_pool (= 2.2.5)
       pg (= 1.2.3)
       puma (= 5.3.2)

data/config/gitlab-exporter.yml.example

@@ -78,8 +78,6 @@ probes:
     opts:
       - pid_or_pattern: "sidekiq .* \\[.*?\\]"
         name: sidekiq
-      - pid_or_pattern: "unicorn.* worker\\[.*?\\]"
-        name: unicorn
       - pid_or_pattern: "git-upload-pack --stateless-rpc"
         name: git_upload_pack
         quantiles: true

data/lib/gitlab_exporter/cli.rb

@@ -217,7 +217,7 @@ module GitLab
             opts.on("--pid=123", "Process ID") do |val|
               @pid = val
             end
-            opts.on("--pattern=unicorn", "Process command pattern") do |val|
+            opts.on("--pattern=worker", "Process command pattern") do |val|
               @pattern = val
             end
             opts.on("--name=NAME", "Process name to be used in metrics") do |val|

data/lib/gitlab_exporter/prometheus.rb

@@ -15,6 +15,21 @@ module GitLab
         @include_timestamp = include_timestamp
       end
 
+      class << self
+        def describe(name, description)
+          @metric_descriptions ||= {}
+          @metric_descriptions[name] = description
+        end
+
+        def description(name)
+          @metric_descriptions && @metric_descriptions[name]
+        end
+
+        def clear_descriptions
+          @metric_descriptions = {}
+        end
+      end
+
       def add(name, value, quantile = false, **labels)
         fail "value '#{value}' must be a number" unless value.is_a?(Numeric)
 
@@ -32,6 +47,8 @@ module GitLab
 
         buffer = ""
         @metrics.each do |name, measurements|
+          buffer << "# HELP #{name} #{self.class.description(name)}\n" if self.class.description(name)
+
           measurements.each do |measurement|
             buffer << name.to_s
             labels = (measurement[:labels] || {}).map { |label, value| "#{label}=\"#{value}\"" }.join(",")

data/lib/gitlab_exporter/sidekiq.rb

@@ -10,6 +10,13 @@ module GitLab
       QUEUE_JOB_STATS_SCRIPT = File.read(File.expand_path("#{__FILE__}/../sidekiq_queue_job_stats.lua")).freeze
       QUEUE_JOB_STATS_SHA    = Digest::SHA1.hexdigest(QUEUE_JOB_STATS_SCRIPT).freeze
 
+      # The maximum depth (from the head) of each queue to probe. Probing the
+      # entirety of a very large queue will take longer and run the risk of
+      # timing out. But when we have a very large queue, we are most in need of
+      # reliable metrics. This trades off completeness for predictability by
+      # only taking a limited amount of items from the head of the queue.
+      PROBE_JOBS_LIMIT = 1_000
+
       POOL_SIZE = 3
 
       # This timeout is configured to higher interval than scrapping
@@ -17,6 +24,9 @@ module GitLab
       # needed to be re-initialized
       POOL_TIMEOUT = 90
 
+      PrometheusMetrics.describe("sidekiq_enqueued_jobs",
+                                 "Total number of jobs enqueued by class name. Only inspects the first #{PROBE_JOBS_LIMIT} jobs per queue.") # rubocop:disable Layout/LineLength
+
       def self.connection_pool
         @@connection_pool ||= Hash.new do |h, connection_hash| # rubocop:disable Style/ClassVars
           config = connection_hash.merge(pool_timeout: POOL_TIMEOUT, size: POOL_SIZE)
@@ -62,6 +72,13 @@ module GitLab
         self
       end
 
+      # Count worker classes present in Sidekiq queues. This uses a Lua
+      # script to find all jobs in all queues. That script will block
+      # all other Redis commands:
+      # https://redis.io/commands/eval#atomicity-of-scripts
+      #
+      # The script is generally fast, but may be slower with very large
+      # queues, which is why this is not enabled by default.
       def probe_jobs
         with_sidekiq do
           job_stats = {}
@@ -84,6 +101,38 @@ module GitLab
         self
       end
 
+      # This does the same as #probe_jobs, but only looks at the first
+      # PROBE_JOBS_LIMIT jobs in each queue. This means that we run a
+      # single LRANGE command for each queue, which does not block other
+      # commands. For queues over PROBE_JOBS_LIMIT in size, this means
+      # that we will not have completely accurate statistics, but the
+      # probe performance will also not degrade as the queue gets
+      # larger.
+      #
+      # DO NOT USE this and probe_jobs together, as they export the same
+      # metric (sidekiq_enqueued_jobs).
+      def probe_jobs_limit
+        with_sidekiq do
+          job_stats = Hash.new(0)
+
+          Sidekiq::Queue.all.each do |queue|
+            Sidekiq.redis do |conn|
+              conn.lrange("queue:#{queue.name}", 0, PROBE_JOBS_LIMIT).each do |job|
+                job_class = Sidekiq.load_json(job)["class"]
+
+                job_stats[job_class] += 1
+              end
+            end
+          end
+
+          job_stats.each do |class_name, count|
+            @metrics.add("sidekiq_enqueued_jobs", count, name: class_name)
+          end
+        end
+
+        self
+      end
+
       def probe_workers
         with_sidekiq do
           worker_stats = Hash.new(0)

data/lib/gitlab_exporter/version.rb

@@ -1,5 +1,5 @@
 module GitLab
   module Exporter
-    VERSION = "10.4.0".freeze
+    VERSION = "10.5.0".freeze
   end
 end

data/spec/prometheus_metrics_spec.rb

@@ -23,4 +23,21 @@ describe GitLab::Exporter::PrometheusMetrics do
       subject.add("mymetric", "invalid", mylabel: "x", myotherlabel: "y").to_s
     }.to raise_error(RuntimeError)
   end
+
+  it "supports described metrics" do
+    time = Time.now
+
+    allow(Time).to receive(:now).and_return(time)
+
+    described_class.describe("mymetric", "description")
+    described_class.describe("missingmetric", "otherdescription")
+    subject.add("mymetric", 1.3, mylabel: "x", myotherlabel: "y")
+
+    expect(subject.to_s).to eq(<<~METRICS)
+      # HELP mymetric description
+      mymetric{mylabel="x",myotherlabel="y"} 1.3 #{(time.to_f * 1000).to_i}
+    METRICS
+
+    described_class.clear_descriptions
+  end
 end

metadata

@@ -1,7 +1,7 @@
 --- !ruby/object:Gem::Specification
 name: gitlab-exporter
 version: !ruby/object:Gem::Version
-  version: 10.4.0
+  version: 10.5.0
 platform: ruby
 authors:
 - Pablo Carranza