Merge branch '207095-monitoring-for-bulk-es-queue' into 'master'

Monitoring for Elasticsearch incremental updates buffer queue

See merge request gitlab-org/gitlab!27384

config/gitlab.yml.example

@@ -459,6 +459,11 @@ production: &base
     elastic_index_bulk_cron_worker:
       cron: "*/1 * * * *"
 
+    # Elasticsearch metrics
+    # NOTE: This will only take effect if Elasticsearch is enabled.
+    elastic_metrics_update_worker:
+      cron: "*/1 * * * *"
+
   registry:
     # enabled: true
     # host: registry.example.com

config/initializers/1_settings.rb

@@ -546,6 +546,9 @@ Gitlab.ee do
   Settings.cron_jobs['elastic_index_bulk_cron_worker'] ||= Settingslogic.new({})
   Settings.cron_jobs['elastic_index_bulk_cron_worker']['cron'] ||= '*/1 * * * *'
   Settings.cron_jobs['elastic_index_bulk_cron_worker']['job_class'] ||= 'ElasticIndexBulkCronWorker'
+  Settings.cron_jobs['elastic_metrics_update_worker'] ||= Settingslogic.new({})
+  Settings.cron_jobs['elastic_metrics_update_worker']['cron'] ||= '*/1 * * * *'
+  Settings.cron_jobs['elastic_metrics_update_worker']['job_class'] ||= 'ElasticMetricsUpdateWorker'
   Settings.cron_jobs['sync_seat_link_worker'] ||= Settingslogic.new({})
   Settings.cron_jobs['sync_seat_link_worker']['cron'] ||= "#{rand(60)} 0 * * *"
   Settings.cron_jobs['sync_seat_link_worker']['job_class'] = 'SyncSeatLinkWorker'

doc/administration/monitoring/prometheus/gitlab_metrics.md

@@ -17,7 +17,13 @@ GitLab monitors its own internal service metrics, and makes them available at th
 `/-/metrics` endpoint. Unlike other [Prometheus](https://prometheus.io) exporters, in order to access
 it, the client IP needs to be [included in a whitelist](../ip_whitelist.md).
 
-For Omnibus and Chart installations, these metrics are automatically enabled and collected as of [GitLab 9.4](https://gitlab.com/gitlab-org/omnibus-gitlab/-/merge_requests/1702). For source installations or earlier versions, these metrics will need to be enabled manually and collected by a Prometheus server.
+For Omnibus and Chart installations, these metrics are automatically enabled
+and collected as of [GitLab
+9.4](https://gitlab.com/gitlab-org/omnibus-gitlab/-/merge_requests/1702). For
+source installations or earlier versions, these metrics will need to be enabled
+manually and collected by a Prometheus server.
+
+See also [Sidekiq metrics](#sidekiq-metrics) for how to enable and view metrics from Sidekiq nodes.
 
 ## Metrics available
 
@@ -105,10 +111,12 @@ The following metrics can be controlled by feature flags:
 | `gitlab_method_call_duration_seconds`                          | `prometheus_metrics_method_instrumentation`                        |
 | `gitlab_view_rendering_duration_seconds`                       | `prometheus_metrics_view_instrumentation`                          |
 
-## Sidekiq Metrics available for Geo **(PREMIUM)**
+## Sidekiq metrics
 
-Sidekiq jobs may also gather metrics, and these metrics can be accessed if the Sidekiq exporter is enabled (e.g. via
-the `monitoring.sidekiq_exporter` configuration option in `gitlab.yml`.
+Sidekiq jobs may also gather metrics, and these metrics can be accessed if the
+Sidekiq exporter is enabled (for example, using the `monitoring.sidekiq_exporter`
+configuration option in `gitlab.yml`. These metrics are served from the
+`/metrics` path on the configured port.
 
 | Metric                                         | Type    | Since | Description | Labels |
 |:---------------------------------------------- |:------- |:----- |:----------- |:------ |
@@ -145,6 +153,7 @@ the `monitoring.sidekiq_exporter` configuration option in `gitlab.yml`.
 | `geo_repositories_checked_failed_count`        | Gauge   | 11.1  | Number of repositories that have a failure from `git fsck` | url |
 | `geo_repositories_retrying_verification_count` | Gauge   | 11.2  | Number of repositories verification failures that Geo is actively trying to correct on secondary  | url |
 | `geo_wikis_retrying_verification_count`        | Gauge   | 11.2  | Number of wikis verification failures that Geo is actively trying to correct on secondary | url |
+| `global_search_bulk_cron_queue_size`           | Gauge   | 12.10 | Number of database records waiting to be synchronized to Elasticsearch | |
 
 ## Database load balancing metrics **(PREMIUM ONLY)**
 

ee/app/workers/all_queues.yml

@@ -31,6 +31,13 @@
   :resource_boundary: :unknown
   :weight: 1
   :idempotent: true
+- :name: cronjob:elastic_metrics_update
+  :feature_category: :global_search
+  :has_external_dependencies:
+  :urgency: :low
+  :resource_boundary: :unknown
+  :weight: 1
+  :idempotent: true
 - :name: cronjob:geo_container_repository_sync_dispatch
   :feature_category: :geo_replication
   :has_external_dependencies: 

ee/app/workers/elastic_metrics_update_worker.rb

+# frozen_string_literal: true
+
+class ElasticMetricsUpdateWorker
+  include ApplicationWorker
+  include ExclusiveLeaseGuard
+  # rubocop:disable Scalability/CronWorkerContext
+  # This worker does not perform work scoped to a context
+  include CronjobQueue
+  # rubocop:enable Scalability/CronWorkerContext
+
+  feature_category :global_search
+  idempotent!
+
+  LEASE_TIMEOUT = 5.minutes
+
+  def perform
+    try_obtain_lease { Elastic::MetricsUpdateService.new.execute }
+  end
+
+  private
+
+  def lease_timeout
+    LEASE_TIMEOUT
+  end
+end

ee/changelogs/unreleased/207095-monitoring-for-bulk-es-queue.yml

+---
+title: Monitoring for Elasticsearch incremental updates buffer queue
+merge_request: 27384
+author:
+type: other

ee/lib/elastic/metrics_update_service.rb

+# frozen_string_literal: true
+
+module Elastic
+  class MetricsUpdateService
+    def execute
+      return unless elasticsearch_enabled?
+      return unless prometheus_enabled?
+
+      gauge = Gitlab::Metrics.gauge(:global_search_bulk_cron_queue_size, 'Number of database records waiting to be synchronized to Elasticsearch', {}, :max)
+      gauge.set({}, Elastic::ProcessBookkeepingService.queue_size)
+    end
+
+    private
+
+    def elasticsearch_enabled?
+      ::Gitlab::CurrentSettings.elasticsearch_indexing?
+    end
+
+    def prometheus_enabled?
+      ::Gitlab::Metrics.prometheus_metrics_enabled?
+    end
+  end
+end

ee/spec/services/elastic/metrics_update_service_spec.rb

+# frozen_string_literal: true
+
+require 'spec_helper'
+
+describe Elastic::MetricsUpdateService, :prometheus do
+  subject { described_class.new }
+
+  before do
+    stub_ee_application_setting(elasticsearch_indexing: true)
+    allow(Gitlab::Metrics).to receive(:prometheus_metrics_enabled?).and_return(true)
+  end
+
+  describe '#execute' do
+    it 'sets a gauge for global_search_bulk_cron_queue_size' do
+      expect(Elastic::ProcessBookkeepingService).to receive(:queue_size).and_return(4)
+
+      gauge_double = instance_double(Prometheus::Client::Gauge)
+      expect(Gitlab::Metrics).to receive(:gauge)
+        .with(:global_search_bulk_cron_queue_size, anything, {}, :max)
+        .and_return(gauge_double)
+
+      expect(gauge_double).to receive(:set).with({}, 4)
+
+      subject.execute
+    end
+
+    context 'when prometheus metrics is disabled' do
+      before do
+        allow(Gitlab::Metrics).to receive(:prometheus_metrics_enabled?).and_return(false)
+      end
+
+      it 'does not set a gauge' do
+        expect(Gitlab::Metrics).not_to receive(:gauge)
+
+        subject.execute
+      end
+    end
+
+    context 'when elasticsearch indexing and search is disabled' do
+      before do
+        stub_ee_application_setting(elasticsearch_indexing: false)
+      end
+
+      it 'does not set a gauge' do
+        expect(Gitlab::Metrics).not_to receive(:gauge)
+
+        subject.execute
+      end
+    end
+  end
+end

ee/spec/workers/elastic_metrics_update_worker_spec.rb

+# frozen_string_literal: true
+
+require 'spec_helper'
+
+describe ElasticMetricsUpdateWorker do
+  include ExclusiveLeaseHelpers
+
+  describe '.perform' do
+    it 'executes the service under an exclusive lease' do
+      expect_to_obtain_exclusive_lease('elastic_metrics_update_worker')
+
+      expect_next_instance_of(::Elastic::MetricsUpdateService) do |service|
+        expect(service).to receive(:execute)
+      end
+
+      described_class.new.perform
+    end
+  end
+end