Merge branch 'record-sidekiq-queuing-latency' into 'master'

Record sidekiq queuing latency

See merge request gitlab-org/gitlab!19005

changelogs/unreleased/record-sidekiq-queuing-latency.yml

+---
+title: Adds a Sidekiq queue duration metric
+merge_request: 19005
+author:
+type: other

lib/gitlab/instrumentation_helper.rb

@@ -21,5 +21,49 @@ module Gitlab
         payload[:rugged_duration_ms] = Gitlab::RuggedInstrumentation.query_time_ms
       end
     end
+
+    # Returns the queuing duration for a Sidekiq job in seconds, as a float, if the
+    # `enqueued_at` field or `created_at` field is available.
+    #
+    # * If the job doesn't contain sufficient information, returns nil
+    # * If the job has a start time in the future, returns 0
+    # * If the job contains an invalid start time value, returns nil
+    # @param [Hash] job a Sidekiq job, represented as a hash
+    def self.queue_duration_for_job(job)
+      # Old gitlab-shell messages don't provide enqueued_at/created_at attributes
+      enqueued_at = job['enqueued_at'] || job['created_at']
+      return unless enqueued_at
+
+      enqueued_at_time = convert_to_time(enqueued_at)
+      return unless enqueued_at_time
+
+      # Its possible that if theres clock-skew between two nodes
+      # this value may be less than zero. In that event, we record the value
+      # as zero.
+      [elapsed_by_absolute_time(enqueued_at_time), 0].max
+    end
+
+    # Calculates the time in seconds, as a float, from
+    # the provided start time until now
+    #
+    # @param [Time] start
+    def self.elapsed_by_absolute_time(start)
+      (Time.now - start).to_f.round(6)
+    end
+    private_class_method :elapsed_by_absolute_time
+
+    # Convert a representation of a time into a `Time` value
+    #
+    # @param time_value String, Float time representation, or nil
+    def self.convert_to_time(time_value)
+      return time_value if time_value.is_a?(Time)
+      return Time.iso8601(time_value) if time_value.is_a?(String)
+      return Time.at(time_value) if time_value.is_a?(Numeric) && time_value > 0
+    rescue ArgumentError
+      # Swallow invalid dates. Better to loose some observability
+      # than bring all background processing down because of a date
+      # formatting bug in a client
+    end
+    private_class_method :convert_to_time
   end
 end

lib/gitlab/sidekiq_logging/structured_logger.rb

@@ -36,11 +36,8 @@ module Gitlab
         payload['message'] = "#{base_message(payload)}: start"
         payload['job_status'] = 'start'
 
-        # Old gitlab-shell messages don't provide enqueued_at/created_at attributes
-        enqueued_at = payload['enqueued_at'] || payload['created_at']
-        if enqueued_at
-          payload['scheduling_latency_s'] = elapsed_by_absolute_time(Time.iso8601(enqueued_at))
-        end
+        scheduling_latency_s = ::Gitlab::InstrumentationHelper.queue_duration_for_job(payload)
+        payload['scheduling_latency_s'] = scheduling_latency_s if scheduling_latency_s
 
         payload
       end
@@ -98,10 +95,6 @@ module Gitlab
         end
       end
 
-      def elapsed_by_absolute_time(start)
-        (Time.now.utc - start).to_f.round(6)
-      end
-
       def elapsed(t0)
         t1 = get_time
         {

lib/gitlab/sidekiq_middleware/metrics.rb

@@ -15,6 +15,9 @@ module Gitlab
 
       def call(_worker, job, queue)
         labels = create_labels(queue)
+        queue_duration = ::Gitlab::InstrumentationHelper.queue_duration_for_job(job)
+
+        @metrics[:sidekiq_jobs_queue_duration_seconds].observe(labels, queue_duration) if queue_duration
         @metrics[:sidekiq_running_jobs].increment(labels, 1)
 
         if job['retry_count'].present?
@@ -49,8 +52,9 @@ module Gitlab
 
       def init_metrics
         {
-          sidekiq_jobs_cpu_seconds:        ::Gitlab::Metrics.histogram(:sidekiq_jobs_cpu_seconds, 'Seconds of cpu time to run sidekiq job', {}, SIDEKIQ_LATENCY_BUCKETS),
-          sidekiq_jobs_completion_seconds: ::Gitlab::Metrics.histogram(:sidekiq_jobs_completion_seconds, 'Seconds to complete sidekiq job', {}, SIDEKIQ_LATENCY_BUCKETS),
+          sidekiq_jobs_cpu_seconds:            ::Gitlab::Metrics.histogram(:sidekiq_jobs_cpu_seconds, 'Seconds of cpu time to run Sidekiq job', {}, SIDEKIQ_LATENCY_BUCKETS),
+          sidekiq_jobs_completion_seconds:     ::Gitlab::Metrics.histogram(:sidekiq_jobs_completion_seconds, 'Seconds to complete Sidekiq job', {}, SIDEKIQ_LATENCY_BUCKETS),
+          sidekiq_jobs_queue_duration_seconds: ::Gitlab::Metrics.histogram(:sidekiq_jobs_queue_duration_seconds, 'Duration in seconds that a Sidekiq job was queued before being executed', {}, SIDEKIQ_LATENCY_BUCKETS),
           sidekiq_jobs_failed_total:           ::Gitlab::Metrics.counter(:sidekiq_jobs_failed_total, 'Sidekiq jobs failed'),
           sidekiq_jobs_retried_total:          ::Gitlab::Metrics.counter(:sidekiq_jobs_retried_total, 'Sidekiq jobs retried'),
           sidekiq_running_jobs:                ::Gitlab::Metrics.gauge(:sidekiq_running_jobs, 'Number of Sidekiq jobs running', {}, :all),

spec/lib/gitlab/instrumentation_helper_spec.rb

+# frozen_string_literal: true
+
+require 'fast_spec_helper'
+require 'rspec-parameterized'
+
+describe Gitlab::InstrumentationHelper do
+  using RSpec::Parameterized::TableSyntax
+
+  describe '.queue_duration_for_job' do
+    where(:enqueued_at, :created_at, :time_now, :expected_duration) do
+      "2019-06-01T00:00:00.000+0000" | nil                            | "2019-06-01T02:00:00.000+0000" | 2.hours.to_f
+      "2019-06-01T02:00:00.000+0000" | nil                            | "2019-06-01T02:00:00.001+0000" | 0.001
+      "2019-06-01T02:00:00.000+0000" | "2019-05-01T02:00:00.000+0000" | "2019-06-01T02:00:01.000+0000" | 1
+      nil                            | "2019-06-01T02:00:00.000+0000" | "2019-06-01T02:00:00.001+0000" | 0.001
+      nil                            | nil                            | "2019-06-01T02:00:00.001+0000" | nil
+      "2019-06-01T02:00:00.000+0200" | nil                            | "2019-06-01T02:00:00.000-0200" | 4.hours.to_f
+      1571825569.998168              | nil                            | "2019-10-23T12:13:16.000+0200" | 26.001832
+      1571825569                     | nil                            | "2019-10-23T12:13:16.000+0200" | 27
+      "invalid_date"                 | nil                            | "2019-10-23T12:13:16.000+0200" | nil
+      ""                             | nil                            | "2019-10-23T12:13:16.000+0200" | nil
+      0                              | nil                            | "2019-10-23T12:13:16.000+0200" | nil
+      -1                             | nil                            | "2019-10-23T12:13:16.000+0200" | nil
+      "2019-06-01T02:00:00.000+0000" | nil                            | "2019-06-01T00:00:00.000+0000" | 0
+      Time.at(1571999233)            | nil                            | "2019-10-25T12:29:16.000+0200" | 123
+    end
+
+    with_them do
+      let(:job) { { 'enqueued_at' => enqueued_at, 'created_at' => created_at } }
+
+      it "returns the correct duration" do
+        Timecop.freeze(Time.iso8601(time_now)) do
+          expect(described_class.queue_duration_for_job(job)).to eq(expected_duration)
+        end
+      end
+    end
+  end
+end

spec/lib/gitlab/sidekiq_logging/structured_logger_spec.rb

 # frozen_string_literal: true
 
-require 'spec_helper'
+require 'fast_spec_helper'
 
 describe Gitlab::SidekiqLogging::StructuredLogger do
   describe '#call' do

spec/lib/gitlab/sidekiq_middleware/metrics_spec.rb

 # frozen_string_literal: true
 
-require 'spec_helper'
+require 'fast_spec_helper'
 
 describe Gitlab::SidekiqMiddleware::Metrics do
   let(:middleware) { described_class.new }
-
   let(:concurrency_metric) { double('concurrency metric') }
+
+  let(:queue_duration_seconds) { double('queue duration seconds metric') }
   let(:completion_seconds_metric) { double('completion seconds metric') }
   let(:user_execution_seconds_metric) { double('user execution seconds metric') }
   let(:failed_total_metric) { double('failed total metric') }
@@ -13,6 +14,7 @@ describe Gitlab::SidekiqMiddleware::Metrics do
   let(:running_jobs_metric) { double('running jobs metric') }
 
   before do
+    allow(Gitlab::Metrics).to receive(:histogram).with(:sidekiq_jobs_queue_duration_seconds, anything, anything, anything).and_return(queue_duration_seconds)
     allow(Gitlab::Metrics).to receive(:histogram).with(:sidekiq_jobs_completion_seconds, anything, anything, anything).and_return(completion_seconds_metric)
     allow(Gitlab::Metrics).to receive(:histogram).with(:sidekiq_jobs_cpu_seconds, anything, anything, anything).and_return(user_execution_seconds_metric)
     allow(Gitlab::Metrics).to receive(:counter).with(:sidekiq_jobs_failed_total, anything).and_return(failed_total_metric)
@@ -20,7 +22,6 @@ describe Gitlab::SidekiqMiddleware::Metrics do
     allow(Gitlab::Metrics).to receive(:gauge).with(:sidekiq_running_jobs, anything, {}, :all).and_return(running_jobs_metric)
     allow(Gitlab::Metrics).to receive(:gauge).with(:sidekiq_concurrency, anything, {}, :all).and_return(concurrency_metric)
 
-    allow(running_jobs_metric).to receive(:increment)
     allow(concurrency_metric).to receive(:set)
   end
 
@@ -32,62 +33,76 @@ describe Gitlab::SidekiqMiddleware::Metrics do
     end
   end
 
+  it 'ignore user execution when measured 0' do
+    allow(completion_seconds_metric).to receive(:observe)
+
+    expect(user_execution_seconds_metric).not_to receive(:observe)
+  end
+
   describe '#call' do
     let(:worker) { double(:worker) }
 
-    it 'yields block' do
-      allow(completion_seconds_metric).to receive(:observe)
-      allow(user_execution_seconds_metric).to receive(:observe)
+    let(:job) { {} }
+    let(:job_status) { :done }
+    let(:labels) { { queue: :test } }
+    let(:labels_with_job_status) { { queue: :test, job_status: job_status } }
 
-      expect { |b| middleware.call(worker, {}, :test, &b) }.to yield_control.once
-    end
+    let(:thread_cputime_before) { 1 }
+    let(:thread_cputime_after) { 2 }
+    let(:thread_cputime_duration) { thread_cputime_after - thread_cputime_before }
 
-    it 'sets queue specific metrics' do
-      labels = { queue: :test }
-      labels_with_job_status = { queue: :test, job_status: :done }
-      allow(middleware).to receive(:get_thread_cputime).and_return(1, 3)
-      allow(Gitlab::Metrics::System).to receive(:monotonic_time).and_return(2, 3)
+    let(:monotonic_time_before) { 11 }
+    let(:monotonic_time_after) { 20 }
+    let(:monotonic_time_duration) { monotonic_time_after - monotonic_time_before }
+
+    let(:queue_duration_for_job) { 0.01 }
+
+    before do
+      allow(middleware).to receive(:get_thread_cputime).and_return(thread_cputime_before, thread_cputime_after)
+      allow(Gitlab::Metrics::System).to receive(:monotonic_time).and_return(monotonic_time_before, monotonic_time_after)
+      allow(Gitlab::InstrumentationHelper).to receive(:queue_duration_for_job).with(job).and_return(queue_duration_for_job)
 
       expect(running_jobs_metric).to receive(:increment).with(labels, 1)
       expect(running_jobs_metric).to receive(:increment).with(labels, -1)
-      expect(user_execution_seconds_metric).to receive(:observe).with(labels_with_job_status, 2)
-      expect(completion_seconds_metric).to receive(:observe).with(labels_with_job_status, 1)
 
-      middleware.call(worker, {}, :test) { nil }
+      expect(queue_duration_seconds).to receive(:observe).with(labels, queue_duration_for_job) if queue_duration_for_job
+      expect(user_execution_seconds_metric).to receive(:observe).with(labels_with_job_status, thread_cputime_duration)
+      expect(completion_seconds_metric).to receive(:observe).with(labels_with_job_status, monotonic_time_duration)
     end
 
-    it 'ignore user execution when measured 0' do
-      allow(completion_seconds_metric).to receive(:observe)
-      allow(middleware).to receive(:get_thread_cputime).and_return(0, 0)
+    it 'yields block' do
+      expect { |b| middleware.call(worker, job, :test, &b) }.to yield_control.once
+    end
 
-      expect(user_execution_seconds_metric).not_to receive(:observe)
+    it 'sets queue specific metrics' do
+      middleware.call(worker, job, :test) { nil }
+    end
+
+    context 'when job_duration is not available' do
+      let(:queue_duration_for_job) { nil }
+
+      it 'does not set the queue_duration_seconds histogram' do
+        middleware.call(worker, job, :test) { nil }
+      end
     end
 
     context 'when job is retried' do
-      it 'sets sidekiq_jobs_retried_total metric' do
-        allow(completion_seconds_metric).to receive(:observe)
-        expect(user_execution_seconds_metric).to receive(:observe)
+      let(:job) { { 'retry_count' => 1 } }
 
+      it 'sets sidekiq_jobs_retried_total metric' do
         expect(retried_total_metric).to receive(:increment)
 
-        middleware.call(worker, { 'retry_count' => 1 }, :test) { nil }
+        middleware.call(worker, job, :test) { nil }
       end
     end
 
     context 'when error is raised' do
-      it 'sets sidekiq_jobs_failed_total and reraises' do
-        labels = { queue: :test }
-        labels_with_job_status = { queue: :test, job_status: :fail }
-        allow(middleware).to receive(:get_thread_cputime).and_return(1, 4)
-        allow(Gitlab::Metrics::System).to receive(:monotonic_time).and_return(2, 6)
+      let(:job_status) { :fail }
 
-        expect(running_jobs_metric).to receive(:increment).with(labels, 1)
-        expect(running_jobs_metric).to receive(:increment).with(labels, -1)
+      it 'sets sidekiq_jobs_failed_total and reraises' do
         expect(failed_total_metric).to receive(:increment).with(labels, 1)
-        expect(user_execution_seconds_metric).to receive(:observe).with(labels_with_job_status, 3)
-        expect(completion_seconds_metric).to receive(:observe).with(labels_with_job_status, 4)
 
-        expect { middleware.call(worker, {}, :test) { raise StandardError, "Failed" } }.to raise_error(StandardError, "Failed")
+        expect { middleware.call(worker, job, :test) { raise StandardError, "Failed" } }.to raise_error(StandardError, "Failed")
       end
     end
   end