debian-mirror-gitlab/doc/development/usage_ping/index.md

stage	group	info
Growth	Product Intelligence	To determine the technical writer assigned to the Stage/Group associated with this page, see https://about.gitlab.com/handbook/engineering/ux/technical-writing/#assignments

Usage Ping Guide

Introduced in GitLab Ultimate 11.2, more statistics.

This guide describes Usage Ping's purpose and how it's implemented.

For more information about Product Intelligence, see:

• Product Intelligence Guide
• Snowplow Guide

More links:

• Product Intelligence Direction
• Data Analysis Process
• Data for Product Managers
• Data Infrastructure

What is Usage Ping?

• GitLab sends a weekly payload containing usage data to GitLab Inc. Usage Ping provides high-level data to help our product, support, and sales teams. It does not send any project names, usernames, or any other specific data. The information from the usage ping is not anonymous, it is linked to the hostname of the instance. Sending usage ping is optional, and any instance can disable analytics.
• The usage data is primarily composed of row counts for different tables in the instance’s database. By comparing these counts month over month (or week over week), we can get a rough sense for how an instance is using the different features in the product. In addition to counts, other facts that help us classify and understand GitLab installations are collected.
• Usage ping is important to GitLab as we use it to calculate our Stage Monthly Active Users (SMAU) which helps us measure the success of our stages and features.
• While usage ping is enabled, GitLab gathers data from the other instances and can show usage statistics of your instance to your users.

Why should we enable Usage Ping?

• The main purpose of Usage Ping is to build a better GitLab. Data about how GitLab is used is collected to better understand feature/stage adoption and usage, which helps us understand how GitLab is adding value and helps our team better understand the reasons why people use GitLab and with this knowledge we're able to make better product decisions.
• As a benefit of having the usage ping active, GitLab lets you analyze the users’ activities over time of your GitLab installation.
• As a benefit of having the usage ping active, GitLab provides you with The DevOps Report,which gives you an overview of your entire instance’s adoption of Concurrent DevOps from planning to monitoring.
• You get better, more proactive support. (assuming that our TAMs and support organization used the data to deliver more value)
• You get insight and advice into how to get the most value out of your investment in GitLab. Wouldn't you want to know that a number of features or values are not being adopted in your organization?
• You get a report that illustrates how you compare against other similar organizations (anonymized), with specific advice and recommendations on how to improve your DevOps processes.
• Usage Ping is enabled by default. To disable it, see Disable Usage Ping.

Limitations

• Usage Ping does not track frontend events things like page views, link clicks, or user sessions, and only focuses on aggregated backend events.
• Because of these limitations we recommend instrumenting your products with Snowplow for more detailed analytics on GitLab.com and use Usage Ping to track aggregated backend events on self-managed.

Usage Ping payload

You can view the exact JSON payload sent to GitLab Inc. in the administration panel. To view the payload:

1. Navigate to Admin Area > Settings > Metrics and profiling.
2. Expand the Usage statistics section.
3. Click the Preview payload button.

For an example payload, see Example Usage Ping payload.

Disable Usage Ping

To disable Usage Ping in the GitLab UI, go to the Settings page of your administration panel and uncheck the Usage Ping checkbox.

To disable Usage Ping and prevent it from being configured in the future through the administration panel, Omnibus installs can set the following in gitlab.rb:

gitlab_rails['usage_ping_enabled'] = false

Source installations can set the following in gitlab.yml:

production: &base
  # ...
  gitlab:
    # ...
    usage_ping_enabled: false

Usage Ping request flow

The following example shows a basic request/response flow between a GitLab instance, the Versions Application, the License Application, Salesforce, the GitLab S3 Bucket, the GitLab Snowflake Data Warehouse, and Sisense:

sequenceDiagram
    participant GitLab Instance
    participant Versions Application
    participant Licenses Application
    participant Salesforce
    participant S3 Bucket
    participant Snowflake DW
    participant Sisense Dashboards
    GitLab Instance->>Versions Application: Send Usage Ping
    loop Process usage data
        Versions Application->>Versions Application: Parse usage data
        Versions Application->>Versions Application: Write to database
        Versions Application->>Versions Application: Update license ping time
    end
    loop Process data for Salesforce
        Versions Application-xLicenses Application: Request Zuora subscription id
        Licenses Application-xVersions Application: Zuora subscription id
        Versions Application-xSalesforce: Request Zuora account id  by Zuora subscription id
        Salesforce-xVersions Application: Zuora account id
        Versions Application-xSalesforce: Usage data for the Zuora account
    end
    Versions Application->>S3 Bucket: Export Versions database
    S3 Bucket->>Snowflake DW: Import data
    Snowflake DW->>Snowflake DW: Transform data using dbt
    Snowflake DW->>Sisense Dashboards: Data available for querying
    Versions Application->>GitLab Instance: DevOps Report (Conversational Development Index)

How Usage Ping works

1. The Usage Ping cron job is set in Sidekiq to run weekly.
2. When the cron job runs, it calls Gitlab::UsageData.to_json.
3. Gitlab::UsageData.to_json cascades down to ~400+ other counter method calls.
4. The response of all methods calls are merged together into a single JSON payload in Gitlab::UsageData.to_json.
5. The JSON payload is then posted to the Versions application If a firewall exception is needed, the required URL depends on several things. If the hostname is version.gitlab.com, the protocol is TCP, and the port number is 443, the required URL is https://version.gitlab.com/.

Usage Ping Metric Life cycle

1. New metrics addition

Please follow the Implementing Usage Ping guide.

2. Existing metric change

Because we do not control when customers update their self-managed instances of GitLab, we STRONGLY DISCOURAGE changes to the logic used to calculate any metric. Any such changes lead to inconsistent reports from multiple GitLab instances. If there is a problem with an existing metric, it's best to deprecate the existing metric, and use it, side by side, with the desired new metric.

Example: Consider following change. Before GitLab 12.6, the example_metric was implemented as:

{
  ...
  example_metric: distinct_count(Project, :creator_id)
}

For GitLab 12.6, the metric was changed to filter out archived projects:

{
  ...
  example_metric: distinct_count(Project.non_archived, :creator_id)
}

In this scenario all instances running up to GitLab 12.5 continue to report example_metric, including all archived projects, while all instances running GitLab 12.6 and higher filters out such projects. As Usage Ping data is collected from all reporting instances, the resulting dataset includes mixed data, which distorts any following business analysis.

The correct approach is to add a new metric for GitLab 12.6 release with updated logic:

{
  ...
  example_metric_without_archived: distinct_count(Project.non_archived, :creator_id)
}

and update existing business analysis artefacts to use example_metric_without_archived instead of example_metric

3. Metrics deprecation and removal

The process for deprecating and removing metrics is under development. For more information, see the following issue.

Implementing Usage Ping

Usage Ping consists of two kinds of data, counters and observations. Counters track how often a certain event happened over time, such as how many CI pipelines have run. They are monotonic and always trend up. Observations are facts collected from one or more GitLab instances and can carry arbitrary data. There are no general guidelines around how to collect those, due to the individual nature of that data.

There are several types of counters which are all found in usage_data.rb:

• Ordinary Batch Counters: Simple count of a given ActiveRecord_Relation
• Distinct Batch Counters: Distinct count of a given ActiveRecord_Relation in a given column
• Sum Batch Counters: Sum the values of a given ActiveRecord_Relation in a given column
• Alternative Counters: Used for settings and configurations
• Redis Counters: Used for in-memory counts.

NOTE: Only use the provided counter methods. Each counter method contains a built in fail safe to isolate each counter to avoid breaking the entire Usage Ping.

Why batch counting

For large tables, PostgreSQL can take a long time to count rows due to MVCC (Multi-version Concurrency Control). Batch counting is a counting method where a single large query is broken into multiple smaller queries. For example, instead of a single query querying 1,000,000 records, with batch counting, you can execute 100 queries of 10,000 records each. Batch counting is useful for avoiding database timeouts as each batch query is significantly shorter than one single long running query.

For GitLab.com, there are extremely large tables with 15 second query timeouts, so we use batch counting to avoid encountering timeouts. Here are the sizes of some GitLab.com tables:

Table	Row counts in millions
merge_request_diff_commits	2280
ci_build_trace_sections	1764
merge_request_diff_files	1082
events	514

We have several batch counting methods available:

• Ordinary Batch Counters
• Distinct Batch Counters
• Sum Batch Counters
• Estimated Batch Counters

Batch counting requires indexes on columns to calculate max, min, and range queries. In some cases, you may need to add a specialized index on the columns involved in a counter.

Ordinary Batch Counters

Handles ActiveRecord::StatementInvalid error

Simple count of a given ActiveRecord_Relation, does a non-distinct batch count, smartly reduces batch_size, and handles errors.

Method: count(relation, column = nil, batch: true, start: nil, finish: nil)

Arguments:

• relation the ActiveRecord_Relation to perform the count
• column the column to perform the count on, by default is the primary key
• batch: default true to use batch counting
• start: custom start of the batch counting to avoid complex min calculations
• end: custom end of the batch counting to avoid complex min calculations

Examples:

count(User.active)
count(::Clusters::Cluster.aws_installed.enabled, :cluster_id)
count(::Clusters::Cluster.aws_installed.enabled, :cluster_id, start: ::Clusters::Cluster.minimum(:id), finish: ::Clusters::Cluster.maximum(:id))

Distinct Batch Counters

Handles ActiveRecord::StatementInvalid error

Distinct count of a given ActiveRecord_Relation on given column, a distinct batch count, smartly reduces batch_size, and handles errors.

Method: distinct_count(relation, column = nil, batch: true, batch_size: nil, start: nil, finish: nil)

Arguments:

• relation the ActiveRecord_Relation to perform the count
• column the column to perform the distinct count, by default is the primary key
• batch: default true to use batch counting
• batch_size: if none set it uses default value 10000 from Gitlab::Database::BatchCounter
• start: custom start of the batch counting to avoid complex min calculations
• end: custom end of the batch counting to avoid complex min calculations

WARNING: Counting over non-unique columns can lead to performance issues. Take a look at the iterating tables in batches guide for more details.

Examples:

distinct_count(::Project, :creator_id)
distinct_count(::Note.with_suggestions.where(time_period), :author_id, start: ::User.minimum(:id), finish: ::User.maximum(:id))
distinct_count(::Clusters::Applications::CertManager.where(time_period).available.joins(:cluster), 'clusters.user_id')

Sum Batch Counters

Handles ActiveRecord::StatementInvalid error

Sum the values of a given ActiveRecord_Relation on given column and handles errors.

Method: sum(relation, column, batch_size: nil, start: nil, finish: nil)

Arguments:

• relation the ActiveRecord_Relation to perform the operation
• column the column to sum on
• batch_size: if none set it uses default value 1000 from Gitlab::Database::BatchCounter
• start: custom start of the batch counting to avoid complex min calculations
• end: custom end of the batch counting to avoid complex min calculations

Examples:

sum(JiraImportState.finished, :imported_issues_count)

Grouping & Batch Operations

The count, distinct_count, and sum batch counters can accept an ActiveRecord::Relation object, which groups by a specified column. With a grouped relation, the methods do batch counting, handle errors, and returns a hash table of key-value pairs.

Examples:

count(Namespace.group(:type))
# returns => {nil=>179, "Group"=>54}

distinct_count(Project.group(:visibility_level), :creator_id)
# returns => {0=>1, 10=>1, 20=>11}

sum(Issue.group(:state_id), :weight))
# returns => {1=>3542, 2=>6820}

Estimated Batch Counters

• Introduced in GitLab 13.7.

Estimated batch counter functionality handles ActiveRecord::StatementInvalid errors when used through the provided estimate_batch_distinct_count method. Errors return a value of -1.

WARNING: This functionality estimates a distinct count of a specific ActiveRecord_Relation in a given column, which uses the HyperLogLog algorithm. As the HyperLogLog algorithm is probabilistic, the results always include error. The highest encountered error rate is 4.9%.

When correctly used, the estimate_batch_distinct_count method enables efficient counting over columns that contain non-unique values, which can not be assured by other counters.

estimate_batch_distinct_count method

Method: estimate_batch_distinct_count(relation, column = nil, batch_size: nil, start: nil, finish: nil)

The method includes the following arguments:

• relation: The ActiveRecord_Relation to perform the count.
• column: The column to perform the distinct count. The default is the primary key.
• batch_size: From Gitlab::Database::PostgresHll::BatchDistinctCounter::DEFAULT_BATCH_SIZE. Default value: 10,000.
• start: The custom start of the batch count, to avoid complex minimum calculations.
• finish: The custom end of the batch count to avoid complex maximum calculations.

The method includes the following prerequisites:

1. The supplied relation must include the primary key defined as the numeric column. For example: id bigint NOT NULL.

2. The estimate_batch_distinct_count can handle a joined relation. To use its ability to count non-unique columns, the joined relation must NOT have a one-to-many relationship, such as has_many :boards.

3. Both start and finish arguments should always represent primary key relationship values, even if the estimated count refers to another column, for example:

     estimate_batch_distinct_count(::Note, :author_id, start: ::Note.minimum(:id), finish: ::Note.maximum(:id))
   

Examples:

1. Simple execution of estimated batch counter, with only relation provided, returned value represents estimated number of unique values in id column (which is the primary key) of Project relation:

     estimate_batch_distinct_count(::Project)
   

2. Execution of estimated batch counter, where provided relation has applied additional filter (.where(time_period)), number of unique values estimated in custom column (:author_id), and parameters: start and finish together apply boundaries that defines range of provided relation to analyze:

     estimate_batch_distinct_count(::Note.with_suggestions.where(time_period), :author_id, start: ::Note.minimum(:id), finish: ::Note.maximum(:id))
   

3. Execution of estimated batch counter with joined relation (joins(:cluster)), for a custom column ('clusters.user_id'):

     estimate_batch_distinct_count(::Clusters::Applications::CertManager.where(time_period).available.joins(:cluster), 'clusters.user_id')
   

When instrumenting metric with usage of estimated batch counter please add _estimated suffix to its name, for example:

  "counts": {
    "ci_builds_estimated": estimate_batch_distinct_count(Ci::Build),
    ...

Redis Counters

Handles ::Redis::CommandError and Gitlab::UsageDataCounters::BaseCounter::UnknownEvent returns -1 when a block is sent or hash with all values -1 when a counter(Gitlab::UsageDataCounters) is sent different behavior due to 2 different implementations of Redis counter

Method: redis_usage_data(counter, &block)

Arguments:

• counter: a counter from Gitlab::UsageDataCounters, that has fallback_totals method implemented
• or a block: which is evaluated

Ordinary Redis Counters

Examples of implementation:

• Using Redis methods INCR, GET, and Gitlab::UsageDataCounters::WikiPageCounter
• Using Redis methods HINCRBY, HGETALL, and Gitlab::UsageCounters::PodLogs

UsageData API Tracking

1. Track event using UsageData API

   Increment event count using ordinary Redis counter, for given event name.

   Tracking events using the UsageData API requires the usage_data_api feature flag to be enabled, which is enabled by default.

   API requests are protected by checking for a valid CSRF token.

   To be able to increment the values, the related feature usage_data_<event_name> should be enabled.

   POST /usage_data/increment_counter
   

   Attribute	Type	Required	Description
   event	string	yes	The event name it should be tracked

   Response

   • 200 if event was tracked
   • 400 Bad request if event parameter is missing
   • 401 Unauthorized if user is not authenticated
   • 403 Forbidden for invalid CSRF token provided

2. Track events using JavaScript/Vue API helper which calls the API above

   Note that usage_data_api and usage_data_#{event_name} should be enabled to be able to track events

   import api from '~/api';
   
   api.trackRedisCounterEvent('my_already_defined_event_name'),
   

Redis HLL Counters

WARNING: HyperLogLog (HLL) is a probabilistic algorithm and its results always includes some small error. According to Redis documentation, data from used HLL implementation is "approximated with a standard error of 0.81%".

With Gitlab::UsageDataCounters::HLLRedisCounter we have available data structures used to count unique values.

Implemented using Redis methods PFADD and PFCOUNT.

Adding new events

1. Define events in known_events.

   Example event:

   - name: i_compliance_credential_inventory
     category: compliance
     redis_slot: compliance
     expiry: 42  # 6 weeks
     aggregation: weekly
   

   Keys:

   • name: unique event name.

     Name format <prefix>_<redis_slot>_name.

     Use one of the following prefixes for the event's name:

     • g_ for group, as an event which is tracked for group.
     • p_ for project, as an event which is tracked for project.
     • i_ for instance, as an event which is tracked for instance.
     • a_ for events encompassing all g_, p_, i_.
     • o_ for other.

     Consider including in the event's name the Redis slot to be able to count totals for a specific category.

     Example names: i_compliance_credential_inventory, g_analytics_contribution.

   • category: event category. Used for getting total counts for events in a category, for easier access to a group of events.

   • redis_slot: optional Redis slot; default value: event name. Used if needed to calculate totals for a group of metrics. Ensure keys are in the same slot. For example: i_compliance_credential_inventory with redis_slot: 'compliance' builds Redis key i_{compliance}_credential_inventory-2020-34. If redis_slot is not defined the Redis key will be {i_compliance_credential_inventory}-2020-34.

   • expiry: expiry time in days. Default: 29 days for daily aggregation and 6 weeks for weekly aggregation.

   • aggregation: may be set to a :daily or :weekly key. Defines how counting data is stored in Redis. Aggregation on a daily basis does not pull more fine grained data.

   • feature_flag: optional default_enabled: :yaml. If no feature flag is set then the tracking is enabled. For details, see our GitLab internal Feature flags documentation. The feature flags are owned by the group adding the event tracking.

Use one of the following methods to track events:

1. Track event in controller using RedisTracking module with track_redis_hll_event(*controller_actions, name:, if: nil).

   Arguments:

   • controller_actions: controller actions we want to track.
   • name: event name.
   • if: optional custom conditions, using the same format as with Rails callbacks.

   Example usage:

   # controller
   class ProjectsController < Projects::ApplicationController
     include RedisTracking
   
     skip_before_action :authenticate_user!, only: :show
     track_redis_hll_event :index, :show, name: 'g_compliance_example_feature_visitors'
   
     def index
       render html: 'index'
     end
   
    def new
      render html: 'new'
    end
   
    def show
      render html: 'show'
    end
   end
   

2. Track event in API using increment_unique_values(event_name, values) helper method.

   To be able to track the event, Usage Ping must be enabled and the event feature usage_data_<event_name> must be enabled.

   Arguments:

   • event_name: event name.
   • values: values counted, one value or array of values.

   Example usage:

   get ':id/registry/repositories' do
     repositories = ContainerRepositoriesFinder.new(
       user: current_user, subject: user_group
     ).execute
   
     increment_unique_values('i_list_repositories', current_user.id)
   
     present paginate(repositories), with: Entities::ContainerRegistry::Repository, tags: params[:tags], tags_count: params[:tags_count]
   end
   

3. Track event using track_usage_event(event_name, values) in services and GraphQL

   Increment unique values count using Redis HLL, for given event name.

   Example:

   Track usage event for incident created in service

   Track usage event for incident created in GraphQL

     track_usage_event(:incident_management_incident_created, current_user.id)
   

1. Track event using UsageData API

   Increment unique users count using Redis HLL, for given event name.

   Tracking events using the UsageData API requires the usage_data_api feature flag to be enabled, which is enabled by default.

   API requests are protected by checking for a valid CSRF token.

   To increment the values, the related feature usage_data_<event_name> should be set to default_enabled: true. For more information, see Feature flags in development of GitLab.

   POST /usage_data/increment_unique_users
   

   Attribute	Type	Required	Description
   event	string	yes	The event name it should be tracked

   Response

   Return 200 if tracking failed for any reason.

   • 200 if event was tracked or any errors
   • 400 Bad request if event parameter is missing
   • 401 Unauthorized if user is not authenticated
   • 403 Forbidden for invalid CSRF token provided

2. Track events using JavaScript/Vue API helper which calls the API above

   Example usage for an existing event already defined in known events:

   Usage Data API is behind usage_data_api feature flag which, as of GitLab 13.7, is now set to default_enabled: true.

   Each event tracked using Usage Data API is behind a feature flag usage_data_#{event_name} which should be default_enabled: true

   import api from '~/api';
   
   api.trackRedisHllUserEvent('my_already_defined_event_name'),
   

3. Get event data using Gitlab::UsageDataCounters::HLLRedisCounter.unique_events(event_names:, start_date:, end_date:, context: '').

   Arguments:

   • event_names: the list of event names.
   • start_date: start date of the period for which we want to get event data.
   • end_date: end date of the period for which we want to get event data.
   • context: context of the event. Allowed values are default, free, bronze, silver, gold, starter, premium, ultimate.

4. Testing tracking and getting unique events

Trigger events in rails console by using track_event method

Gitlab::UsageDataCounters::HLLRedisCounter.track_event('g_compliance_audit_events', values: 1)
Gitlab::UsageDataCounters::HLLRedisCounter.track_event('g_compliance_audit_events', values: [2, 3])

Next, get the unique events for the current week.

# Get unique events for metric for current_week
Gitlab::UsageDataCounters::HLLRedisCounter.unique_events(event_names: 'g_compliance_audit_events',
start_date: Date.current.beginning_of_week, end_date: Date.current.next_week)

Recommendations

We have the following recommendations for Adding new events:

• Event aggregation: weekly.
• Key expiry time:
  • Daily: 29 days.
  • Weekly: 42 days.
• When adding new metrics, use a feature flag to control the impact.
• For feature flags triggered by another service, set default_enabled: false,
  • Events can be triggered using the UsageData API, which helps when there are > 10 events per change

Enable/Disable Redis HLL tracking

Events are tracked behind feature flags due to concerns for Redis performance and scalability.

For a full list of events and corresponding feature flags see, known_events files.

To enable or disable tracking for specific event in https://gitlab.com or https://about.staging.gitlab.com, run commands such as the following to enable or disable the corresponding feature.

/chatops run feature set <feature_name> true
/chatops run feature set <feature_name> false

Known events are added automatically in usage data payload

All events added in known_events/common.yml are automatically added to usage data generation under the redis_hll_counters key. This column is stored in version-app as a JSON. For each event we add metrics for the weekly and monthly time frames, and totals for each where applicable:

• #{event_name}_weekly: Data for 7 days for daily aggregation events and data for the last complete week for weekly aggregation events.
• #{event_name}_monthly: Data for 28 days for daily aggregation events and data for the last 4 complete weeks for weekly aggregation events.

Redis HLL implementation calculates automatic total metrics, if there are more than one metric for the same category, aggregation, and Redis slot.

• #{category}_total_unique_counts_weekly: Total unique counts for events in the same category for the last 7 days or the last complete week, if events are in the same Redis slot and we have more than one metric.
• #{category}_total_unique_counts_monthly: Total unique counts for events in same category for the last 28 days or the last 4 complete weeks, if events are in the same Redis slot and we have more than one metric.

Example of redis_hll_counters data:

{:redis_hll_counters=>
  {"compliance"=>
    {"g_compliance_dashboard_weekly"=>0,
     "g_compliance_dashboard_monthly"=>0,
     "g_compliance_audit_events_weekly"=>0,
     "g_compliance_audit_events_monthly"=>0,
     "compliance_total_unique_counts_weekly"=>0,
     "compliance_total_unique_counts_monthly"=>0},
   "analytics"=>
    {"g_analytics_contribution_weekly"=>0,
     "g_analytics_contribution_monthly"=>0,
     "g_analytics_insights_weekly"=>0,
     "g_analytics_insights_monthly"=>0,
     "analytics_total_unique_counts_weekly"=>0,
     "analytics_total_unique_counts_monthly"=>0},
   "ide_edit"=>
    {"g_edit_by_web_ide_weekly"=>0,
     "g_edit_by_web_ide_monthly"=>0,
     "g_edit_by_sfe_weekly"=>0,
     "g_edit_by_sfe_monthly"=>0,
     "ide_edit_total_unique_counts_weekly"=>0,
     "ide_edit_total_unique_counts_monthly"=>0},
   "search"=>
    {"i_search_total_weekly"=>0, "i_search_total_monthly"=>0, "i_search_advanced_weekly"=>0, "i_search_advanced_monthly"=>0, "i_search_paid_weekly"=>0, "i_search_paid_monthly"=>0, "search_total_unique_counts_weekly"=>0, "search_total_unique_counts_monthly"=>0},
   "source_code"=>{"wiki_action_weekly"=>0, "wiki_action_monthly"=>0}
 }

Example usage:

# Redis Counters
redis_usage_data(Gitlab::UsageDataCounters::WikiPageCounter)
redis_usage_data { ::Gitlab::UsageCounters::PodLogs.usage_totals[:total] }

# Define events in common.yml https://gitlab.com/gitlab-org/gitlab/-/blob/master/lib/gitlab/usage_data_counters/known_events/common.yml

# Tracking events
Gitlab::UsageDataCounters::HLLRedisCounter.track_event('expand_vulnerabilities', values: visitor_id)

# Get unique events for metric
redis_usage_data { Gitlab::UsageDataCounters::HLLRedisCounter.unique_events(event_names: 'expand_vulnerabilities', start_date: 28.days.ago, end_date: Date.current) }

Alternative Counters

Handles StandardError and fallbacks into -1 this way not all measures fail if we encounter one exception. Mainly used for settings and configurations.

Method: alt_usage_data(value = nil, fallback: -1, &block)

Arguments:

• value: a simple static value in which case the value is simply returned.
• or a block: which is evaluated
• fallback: -1: the common value used for any metrics that are failing.

Usage:

alt_usage_data { Gitlab::VERSION }
alt_usage_data { Gitlab::CurrentSettings.uuid }
alt_usage_data(999)

Adding counters to build new metrics

When adding the results of two counters, use the add usage data method that handles fallback values and exceptions. It also generates a valid SQL export.

Example usage:

add(User.active, User.bot)

Prometheus Queries

In those cases where operational metrics should be part of Usage Ping, a database or Redis query is unlikely to provide useful data. Instead, Prometheus might be more appropriate, because most GitLab architectural components publish metrics to it that can be queried back, aggregated, and included as usage data.

NOTE: Prometheus as a data source for Usage Ping is currently only available for single-node Omnibus installations that are running the bundled Prometheus instance.

To query Prometheus for metrics, a helper method is available to yield a fully configured PrometheusClient, given it is available as per the note above:

with_prometheus_client do |client|
  response = client.query('<your query>')
  ...
end

Please refer to the PrometheusClient definition for how to use its API to query for data.

Developing and testing Usage Ping

1. Naming and placing the metrics

Add the metric in one of the top level keys

• license: for license related metrics.
• settings: for settings related metrics.
• counts_weekly: for counters that have data for the most recent 7 days.
• counts_monthly: for counters that have data for the most recent 28 days.
• counts: for counters that have data for all time.

2. Use your Rails console to manually test counters

# count
Gitlab::UsageData.count(User.active)
Gitlab::UsageData.count(::Clusters::Cluster.aws_installed.enabled, :cluster_id)

# count distinct
Gitlab::UsageData.distinct_count(::Project, :creator_id)
Gitlab::UsageData.distinct_count(::Note.with_suggestions.where(time_period), :author_id, start: ::User.minimum(:id), finish: ::User.maximum(:id))

3. Generate the SQL query

Your Rails console returns the generated SQL queries.

Example:

pry(main)> Gitlab::UsageData.count(User.active)
   (2.6ms)  SELECT "features"."key" FROM "features"
   (15.3ms)  SELECT MIN("users"."id") FROM "users" WHERE ("users"."state" IN ('active')) AND ("users"."user_type" IS NULL OR "users"."user_type" IN (6, 4))
   (2.4ms)  SELECT MAX("users"."id") FROM "users" WHERE ("users"."state" IN ('active')) AND ("users"."user_type" IS NULL OR "users"."user_type" IN (6, 4))
   (1.9ms)  SELECT COUNT("users"."id") FROM "users" WHERE ("users"."state" IN ('active')) AND ("users"."user_type" IS NULL OR "users"."user_type" IN (6, 4)) AND "users"."id" BETWEEN 1 AND 100000

4. Optimize queries with #database-lab

Paste the SQL query into #database-lab to see how the query performs at scale.

• #database-lab is a Slack channel which uses a production-sized environment to test your queries.
• GitLab.com’s production database has a 15 second timeout.
• Any single query must stay below 1 second execution time with cold caches.
• Add a specialized index on columns involved to reduce the execution time.

To have an understanding of the query's execution we add in the MR description the following information:

• For counters that have a time_period test we add information for both cases:
  • time_period = {} for all time periods
  • time_period = { created_at: 28.days.ago..Time.current } for last 28 days period
• Execution plan and query time before and after optimization
• Query generated for the index and time
• Migration output for up and down execution

We also use #database-lab and explain.depesz.com. For more details, see the database review guide.

Optimization recommendations and examples

• Use specialized indexes example 1, example 2.
• Use defined start and finish, and simple queries. These values can be memoized and reused, example.
• Avoid joins and write the queries as simply as possible, example.
• Set a custom batch_size for distinct_count, example.

5. Add the metric definition

Check Metrics Dictionary Guide

When adding, updating, or removing metrics, please update the Metrics Dictionary.

6. Add new metric to Versions Application

Check if new metrics need to be added to the Versions Application. See usage_data schema and usage data parameters accepted. Any metrics added under the counts key are saved in the stats column.

7. Add the feature label

Add the feature label to the Merge Request for new Usage Ping metrics. These are user-facing changes and are part of expanding the Usage Ping feature.

8. Add a changelog file

Ensure you comply with the Changelog entries guide.

9. Ask for a Product Intelligence Review

On GitLab.com, we have DangerBot setup to monitor Product Intelligence related files and DangerBot recommends a Product Intelligence review. Mention @gitlab-org/growth/product_intelligence/engineers in your MR for a review.

10. Verify your metric

On GitLab.com, the Product Intelligence team regularly monitors Usage Ping. They may alert you that your metrics need further optimization to run quicker and with greater success. You may also use the Usage Ping QA dashboard to check how well your metric performs. The dashboard allows filtering by GitLab version, by "Self-managed" & "SaaS" and shows you how many failures have occurred for each metric. Whenever you notice a high failure rate, you may re-optimize your metric.

Optional: Test Prometheus based Usage Ping

If the data submitted includes metrics queried from Prometheus that you would like to inspect and verify, then you need to ensure that a Prometheus server is running locally, and that furthermore the respective GitLab components are exporting metrics to it. If you do not need to test data coming from Prometheus, no further action is necessary. Usage Ping should degrade gracefully in the absence of a running Prometheus server.

There are three kinds of components that may export data to Prometheus, and which are included in Usage Ping:

• node_exporter - Exports node metrics from the host machine
• gitlab-exporter - Exports process metrics from various GitLab components
• various GitLab services such as Sidekiq and the Rails server that export their own metrics

Test with an Omnibus container

This is the recommended approach to test Prometheus based Usage Ping.

The easiest way to verify your changes is to build a new Omnibus image from your code branch by using CI, then download the image and run a local container instance:

1. From your merge request, click on the qa stage, then trigger the package-and-qa job. This job triggers an Omnibus build in a downstream pipeline of the omnibus-gitlab-mirror project.
2. In the downstream pipeline, wait for the gitlab-docker job to finish.
3. Open the job logs and locate the full container name including the version. It takes the following form: registry.gitlab.com/gitlab-org/build/omnibus-gitlab-mirror/gitlab-ee:<VERSION>.
4. On your local machine, make sure you are signed in to the GitLab Docker registry. You can find the instructions for this in Authenticate to the GitLab Container Registry.
5. Once signed in, download the new image by using docker pull registry.gitlab.com/gitlab-org/build/omnibus-gitlab-mirror/gitlab-ee:<VERSION>
6. For more information about working with and running Omnibus GitLab containers in Docker, please refer to GitLab Docker images in the Omnibus documentation.

Test with GitLab development toolkits

This is the less recommended approach, because it comes with a number of difficulties when emulating a real GitLab deployment.

The GDK is not set up to run a Prometheus server or node_exporter alongside other GitLab components. If you would like to do so, Monitoring the GDK with Prometheus is a good start.

The GCK has limited support for testing Prometheus based Usage Ping. By default, it already comes with a fully configured Prometheus service that is set up to scrape a number of components, but with the following limitations:

• It does not run a gitlab-exporter instance, so several process_* metrics from services such as Gitaly may be missing.
• While it runs a node_exporter, docker-compose services emulate hosts, meaning that it would normally report itself to not be associated with any of the other services that are running. That is not how node metrics are reported in a production setup, where node_exporter always runs as a process alongside other GitLab components on any given node. From Usage Ping's perspective none of the node data would therefore appear to be associated to any of the services running, because they all appear to be running on different hosts. To alleviate this problem, the node_exporter in GCK was arbitrarily "assigned" to the web service, meaning only for this service node_* metrics appears in Usage Ping.

Aggregated metrics

• Introduced in GitLab 13.6.

WARNING: This feature is intended solely for internal GitLab use.

To add data for aggregated metrics into Usage Ping payload you should add corresponding definition at lib/gitlab/usage_data_counters/aggregated_metrics/*.yaml for metrics available at Community Edition and at ee/lib/gitlab/usage_data_counters/aggregated_metrics/*.yaml for Enterprise Edition ones.

Each aggregate definition includes following parts:

• name: Unique name under which the aggregate metric is added to the Usage Ping payload.
• operator: Operator that defines how the aggregated metric data is counted. Available operators are:
  • OR: Removes duplicates and counts all entries that triggered any of listed events.
  • AND: Removes duplicates and counts all elements that were observed triggering all of following events.
• time_frame: One or more valid time frames. Use these to limit the data included in aggregated metric to events within a specific date-range. Valid time frames are:
  • 7d: Last seven days of data.
  • 28d: Last twenty eight days of data.
  • all: All historical data, only available for database sourced aggregated metrics.
• source: Data source used to collect all events data included in aggregated metric. Valid data sources are:
  • database
  • redis
• events: list of events names to aggregate into metric. All events in this list must relay on the same data source. Additional data source requirements are described in the Database sourced aggregated metrics and Redis sourced aggregated metrics sections.
• feature_flag: Name of development feature flag that is checked before metrics aggregation is performed. Corresponding feature flag should have default_enabled attribute set to false. The feature_flag attribute is optional and can be omitted. When feature_flag is missing, no feature flag is checked.

Example aggregated metric entries:

- name: example_metrics_union
  operator: OR
  events: ['i_search_total', 'i_search_advanced', 'i_search_paid']
  source: redis
  time_frame:
    - 7d
    - 28d
- name: example_metrics_intersection
  operator: AND
  source: database
  time_frame:
    - 28d
    - all
  events: ['dependency_scanning_pipeline_all_time', 'container_scanning_pipeline_all_time']
  feature_flag: example_aggregated_metric

Aggregated metrics collected in 7d and 28d time frames are added into Usage Ping payload under the aggregated_metrics sub-key in the counts_weekly and counts_monthly top level keys.

{
  :counts_monthly => {
    :deployments => 1003,
    :successful_deployments => 78,
    :failed_deployments => 275,
    :packages => 155,
    :personal_snippets => 2106,
    :project_snippets => 407,
    :promoted_issues => 719,
    :aggregated_metrics => {
      :example_metrics_union => 7,
      :example_metrics_intersection => 2
    },
    :snippets => 2513
  }
}

Aggregated metrics for all time frame are present in the count top level key, with the aggregate_ prefix added to their name.

For example:

example_metrics_intersection

Becomes:

counts.aggregate_example_metrics_intersection

{
  :counts => {
    :deployments => 11003,
    :successful_deployments => 178,
    :failed_deployments => 1275,
    :aggregate_example_metrics_intersection => 12
  }
}

Redis sourced aggregated metrics

• Introduced in GitLab 13.6.

To declare the aggregate of events collected with Redis HLL Counters, you must fulfill the following requirements:

1. All events listed at events attribute must come from known_events/*.yml files.
2. All events listed at events attribute must have the same redis_slot attribute.
3. All events listed at events attribute must have the same aggregation attribute.
4. time_frame does not include all value, which is unavailable for Redis sourced aggregated metrics.

Database sourced aggregated metrics

• Introduced in GitLab 13.9.
• It's deployed behind a feature flag, disabled by default.
• It's enabled on GitLab.com.

To declare an aggregate of metrics based on events collected from database, follow these steps:

1. Persist the metrics for aggregation.
2. Add new aggregated metric definition.

Persist metrics for aggregation

Only metrics calculated with Estimated Batch Counters can be persisted for database sourced aggregated metrics. To persist a metric, inject a Ruby block into the estimate_batch_distinct_count method. This block should invoke the Gitlab::Usage::Metrics::Aggregates::Sources::PostgresHll.save_aggregated_metrics method, which stores estimate_batch_distinct_count results for future use in aggregated metrics.

The Gitlab::Usage::Metrics::Aggregates::Sources::PostgresHll.save_aggregated_metrics method accepts the following arguments:

• metric_name: The name of metric to use for aggregations. Should be the same as the key under which the metric is added into Usage Ping.
• recorded_at_timestamp: The timestamp representing the moment when a given Usage Ping payload was collected. You should use the convenience method recorded_at to fill recorded_at_timestamp argument, like this: recorded_at_timestamp: recorded_at
• time_period: The time period used to build the relation argument passed into estimate_batch_distinct_count. To collect the metric with all available historical data, set a nil value as time period: time_period: nil.
• data: HyperLogLog buckets structure representing unique entries in relation. The estimate_batch_distinct_count method always passes the correct argument into the block, so data argument must always have a value equal to block argument, like this: data: result

Example metrics persistence:

class UsageData
  def count_secure_pipelines(time_period)
    ...
    relation = ::Security::Scan.latest_successful_by_build.by_scan_types(scan_type).where(security_scans: time_period)

    pipelines_with_secure_jobs['dependency_scanning_pipeline'] = estimate_batch_distinct_count(relation, :commit_id, batch_size: 1000, start: start_id, finish: finish_id) do |result|
      ::Gitlab::Usage::Metrics::Aggregates::Sources::PostgresHll
        .save_aggregated_metrics(metric_name: 'dependency_scanning_pipeline', recorded_at_timestamp: recorded_at, time_period: time_period, data: result)
    end
  end
end

Add new aggregated metric definition

After all metrics are persisted, you can add an aggregated metric definition at aggregated_metrics/.

To declare the aggregate of metrics collected with Estimated Batch Counters, you must fulfill the following requirements:

• Metrics names listed in the events: attribute, have to use the same names you passed in the metric_name argument while persisting metrics in previous step.
• Every metric listed in the events: attribute, has to be persisted for every selected time_frame: value.

Example definition:

- name: example_metrics_intersection_database_sourced
  operator: AND
  source: database
  events: ['dependency_scanning_pipeline', 'container_scanning_pipeline']
  time_frame:
    - 28d
    - all

Example Usage Ping payload

The following is example content of the Usage Ping payload.

{
  "uuid": "0000000-0000-0000-0000-000000000000",
  "hostname": "example.com",
  "version": "12.10.0-pre",
  "installation_type": "omnibus-gitlab",
  "active_user_count": 999,
  "recorded_at": "2020-04-17T07:43:54.162+00:00",
  "edition": "EEU",
  "license_md5": "00000000000000000000000000000000",
  "license_id": null,
  "historical_max_users": 999,
  "licensee": {
    "Name": "ABC, Inc.",
    "Email": "email@example.com",
    "Company": "ABC, Inc."
  },
  "license_user_count": 999,
  "license_starts_at": "2020-01-01",
  "license_expires_at": "2021-01-01",
  "license_plan": "ultimate",
  "license_add_ons": {
  },
  "license_trial": false,
  "counts": {
    "assignee_lists": 999,
    "boards": 999,
    "ci_builds": 999,
    ...
  },
  "container_registry_enabled": true,
  "dependency_proxy_enabled": false,
  "gitlab_shared_runners_enabled": true,
  "gravatar_enabled": true,
  "influxdb_metrics_enabled": true,
  "ldap_enabled": false,
  "mattermost_enabled": false,
  "omniauth_enabled": true,
  "prometheus_enabled": false,
  "prometheus_metrics_enabled": false,
  "reply_by_email_enabled": "incoming+%{key}@incoming.gitlab.com",
  "signup_enabled": true,
  "web_ide_clientside_preview_enabled": true,
  "ingress_modsecurity_enabled": true,
  "projects_with_expiration_policy_disabled": 999,
  "projects_with_expiration_policy_enabled": 999,
  ...
  "elasticsearch_enabled": true,
  "license_trial_ends_on": null,
  "geo_enabled": false,
  "git": {
    "version": {
      "major": 2,
      "minor": 26,
      "patch": 1
    }
  },
  "gitaly": {
    "version": "12.10.0-rc1-93-g40980d40",
    "servers": 56,
    "clusters": 14,
    "filesystems": [
      "EXT_2_3_4"
    ]
  },
  "gitlab_pages": {
    "enabled": true,
    "version": "1.17.0"
  },
  "container_registry_server": {
    "vendor": "gitlab",
    "version": "2.9.1-gitlab"
  },
  "database": {
    "adapter": "postgresql",
    "version": "9.6.15",
    "pg_system_id": 6842684531675334351
  },
  "analytics_unique_visits": {
    "g_analytics_contribution": 999,
    ...
  },
  "usage_activity_by_stage": {
    "configure": {
      "project_clusters_enabled": 999,
      ...
    },
    "create": {
      "merge_requests": 999,
      ...
    },
    "manage": {
      "events": 999,
      ...
    },
    "monitor": {
      "clusters": 999,
      ...
    },
    "package": {
      "projects_with_packages": 999
    },
    "plan": {
      "issues": 999,
      ...
    },
    "release": {
      "deployments": 999,
      ...
    },
    "secure": {
      "user_container_scanning_jobs": 999,
      ...
    },
    "verify": {
      "ci_builds": 999,
      ...
    }
  },
  "usage_activity_by_stage_monthly": {
    "configure": {
      "project_clusters_enabled": 999,
      ...
    },
    "create": {
      "merge_requests": 999,
      ...
    },
    "manage": {
      "events": 999,
      ...
    },
    "monitor": {
      "clusters": 999,
      ...
    },
    "package": {
      "projects_with_packages": 999
    },
    "plan": {
      "issues": 999,
      ...
    },
    "release": {
      "deployments": 999,
      ...
    },
    "secure": {
      "user_container_scanning_jobs": 999,
      ...
    },
    "verify": {
      "ci_builds": 999,
      ...
    }
  },
  "topology": {
    "duration_s": 0.013836685999194742,
    "application_requests_per_hour": 4224,
    "query_apdex_weekly_average": 0.996,
    "failures": [],
    "nodes": [
      {
        "node_memory_total_bytes": 33269903360,
        "node_memory_utilization": 0.35,
        "node_cpus": 16,
        "node_cpu_utilization": 0.2,
        "node_uname_info": {
          "machine": "x86_64",
          "sysname": "Linux",
          "release": "4.19.76-linuxkit"
        },
        "node_services": [
          {
            "name": "web",
            "process_count": 16,
            "process_memory_pss": 233349888,
            "process_memory_rss": 788220927,
            "process_memory_uss": 195295487,
            "server": "puma"
          },
          {
            "name": "sidekiq",
            "process_count": 1,
            "process_memory_pss": 734080000,
            "process_memory_rss": 750051328,
            "process_memory_uss": 731533312
          },
          ...
        ],
        ...
      },
      ...
    ]
  }
}

Notable changes

In GitLab 13.5, pg_system_id was added to send the PostgreSQL system identifier.

Exporting Usage Ping SQL queries and definitions

Two Rake tasks exist to export Usage Ping definitions.

• The Rake tasks export the raw SQL queries for count, distinct_count, sum.
• The Rake tasks export the Redis counter class or the line of the Redis block for redis_usage_data.
• The Rake tasks calculate the alt_usage_data metrics.

In the home directory of your local GitLab installation run the following Rake tasks for the YAML and JSON versions respectively:

# for YAML export
bin/rake gitlab:usage_data:dump_sql_in_yaml

# for JSON export
bin/rake gitlab:usage_data:dump_sql_in_json

# You may pipe the output into a file
bin/rake gitlab:usage_data:dump_sql_in_yaml > ~/Desktop/usage-metrics-2020-09-02.yaml

Generating and troubleshooting usage ping

To get a usage ping, or to troubleshoot caching issues on your GitLab instance, please follow instructions to generate usage ping.