Metrics instrumentation guide

This guide describes how to develop Service Ping metrics using metrics instrumentation.

For a video tutorial, see the Adding Service Ping metric via instrumentation class.

Nomenclature

  • Instrumentation class:
    • Inherits one of the metric classes: DatabaseMetric, NumbersMetric or GenericMetric.
    • Implements the logic that calculates the value for a Service Ping metric.
  • Metric definition The Service Data metric YAML definition.

  • Hardening: Hardening a method is the process that ensures the method fails safe, returning a fallback value like -1.

How it works

A metric definition has the instrumentation_class field, which can be set to a class.

The defined instrumentation class should inherit one of the existing metric classes: DatabaseMetric, NumbersMetric or GenericMetric.

The current convention is that a single instrumentation class corresponds to a single metric.

Using an instrumentation class ensures that metrics can fail safe individually, without breaking the entire process of Service Ping generation.

Database metrics

note
Whenever possible we recommend using internal event tracking instead of database metrics. Database metrics can create unnecessary load on the database of bigger GitLab instances and potential optimisations can affect instance performance.

You can use database metrics to track data kept in the database, for example, a count of issues that exist on a given instance.

  • operation: Operations for the given relation, one of count, distinct_count, sum, and average.
  • relation: Assigns lambda that returns the ActiveRecord::Relation for the objects we want to perform the operation. The assigned lambda can accept up to one parameter. The parameter is hashed and stored under the options key in the metric definition.
  • start: Specifies the start value of the batch counting, by default is relation.minimum(:id).
  • finish: Specifies the end value of the batch counting, by default is relation.maximum(:id).
  • cache_start_and_finish_as: Specifies the cache key for start and finish values and sets up caching them. Use this call when start and finish are expensive queries that should be reused between different metric calculations.
  • available?: Specifies whether the metric should be reported. The default is true.
  • timestamp_column: Optionally specifies timestamp column for metric used to filter records for time constrained metrics. The default is created_at.

Example of a merge request that adds a database metric.

Optimization recommendations and examples

Any single query for a Service Ping metric must stay below the 1 second execution time with cold caches.

Database metric Examples

Count Example

module Gitlab
  module Usage
    module Metrics
      module Instrumentations
        class CountIssuesMetric < DatabaseMetric
          operation :count

          relation ->(options) { Issue.where(confidential: options[:confidential]) }
        end
      end
    end
  end
end

Batch counters Example

module Gitlab
  module Usage
    module Metrics
      module Instrumentations
        class CountIssuesMetric < DatabaseMetric
          operation :count

          start { Issue.minimum(:id) }
          finish { Issue.maximum(:id) }

          relation { Issue }
        end
      end
    end
  end
end

Distinct batch counters Example

# frozen_string_literal: true

module Gitlab
  module Usage
    module Metrics
      module Instrumentations
        class CountUsersAssociatingMilestonesToReleasesMetric < DatabaseMetric
          operation :distinct_count, column: :author_id

          relation { Release.with_milestones }

          start { Release.minimum(:author_id) }
          finish { Release.maximum(:author_id) }
        end
      end
    end
  end
end

Sum Example

# frozen_string_literal: true

module Gitlab
  module Usage
    module Metrics
      module Instrumentations
        class JiraImportsTotalImportedIssuesCountMetric < DatabaseMetric
          operation :sum, column: :imported_issues_count

          relation { JiraImportState.finished }
        end
      end
    end
  end
end

Average Example

# frozen_string_literal: true

module Gitlab
  module Usage
    module Metrics
      module Instrumentations
        class CountIssuesWeightAverageMetric < DatabaseMetric
          operation :average, column: :weight

          relation { Issue }
        end
      end
    end
  end
end

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.

caution
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 cannot 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:

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

Aggregated metrics

See the video from: Product Intelligence Office Hours Oct 6th for an aggregated metrics walk-through.

The aggregated metrics feature provides insight into the number of data attributes, for example pseudonymized_user_ids, that occurred in a collection of events. For example, you can aggregate the number of users who perform multiple actions such as creating a new issue and opening a new merge request.

You can use a YAML file to define your aggregated metrics. The following arguments are required:

  • options.events: List of event names to aggregate into metric data. All events in this list must use the same data source. Additional data source requirements are described in Database sourced aggregated metrics and Event sourced aggregated metrics.
  • options.aggregate.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 the listed events.
    • AND: Removes duplicates and counts all elements that were observed triggering all of the following events.
  • options.aggregate.attribute: Information pointing to the attribute that is being aggregated across events.
  • time_frame: One or more valid time frames. Use these to limit the data included in aggregated metrics to events within a specific date-range. Valid time frames are:
    • 7d: The last 7 days of data.
    • 28d: The last 28 days of data.
    • all: All historical data, only available for database sourced aggregated metrics.
  • data_source: Data source used to collect all events data included in the aggregated metrics. Valid data sources are:

Refer to merge request 98206 for an example of a merge request that adds an AggregatedMetric metric.

Count unique user_ids that occurred in at least one of the events: incident_management_alert_status_changed, incident_management_alert_assigned, incident_management_alert_todo, incident_management_alert_create_incident.

time_frame: 28d
instrumentation_class: AggregatedMetric
data_source: internal_events
options:
    aggregate:
        operator: OR
        attribute: user_id
    events:
        - `incident_management_alert_status_changed`
        - `incident_management_alert_assigned`
        - `incident_management_alert_todo`
        - `incident_management_alert_create_incident`

Event sourced aggregated metrics

Introduced in GitLab 13.6.

To declare the aggregate of events collected with Internal Events, make sure time_frame does not include the all value, which is unavailable for Redis-sourced aggregated metrics.

While it is possible to aggregate EE-only events together with events that occur in all GitLab editions, it’s important to remember that doing so may produce high variance between data collected from EE and CE GitLab instances.

Database sourced aggregated metrics

Introduced in GitLab 13.9.

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 Service Ping.
  • recorded_at_timestamp: The timestamp representing the moment when a given Service 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.by_scan_types(scan_type).where(time_period)

    pipelines_with_secure_jobs['dependency_scanning_pipeline'] = estimate_batch_distinct_count(relation, :pipeline_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. 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.

Availability-restrained Aggregated metrics

If the Aggregated metric should only be available in the report under specific conditions, then you must specify these conditions in a new class that is a child of the AggregatedMetric class.

# frozen_string_literal: true

module Gitlab
  module Usage
    module Metrics
      module Instrumentations
        class MergeUsageCountAggregatedMetric < AggregatedMetric
          available? { Feature.enabled?(:merge_usage_data_missing_key_paths) }
        end
      end
    end
  end
end

You must also use the class’s name in the YAML setup.

time_frame: 28d
instrumentation_class: MergeUsageCountAggregatedMetric
data_source: redis_hll
options:
    aggregate:
        operator: OR
        attribute: user_id
    events:
        - `incident_management_alert_status_changed`
        - `incident_management_alert_assigned`
        - `incident_management_alert_todo`
        - `incident_management_alert_create_incident`

Numbers metrics

  • operation: Operations for the given data block. Currently we only support add operation.
  • data: a block which contains an array of numbers.
  • available?: Specifies whether the metric should be reported. The default is true.
# frozen_string_literal: true

module Gitlab
  module Usage
    module Metrics
      module Instrumentations
          class IssuesBoardsCountMetric < NumbersMetric
            operation :add

            data do |time_frame|
              [
                 CountIssuesMetric.new(time_frame: time_frame).value,
                 CountBoardsMetric.new(time_frame: time_frame).value
              ]
            end
          end
        end
      end
    end
  end
end

You must also include the instrumentation class name in the YAML setup.

time_frame: 28d
instrumentation_class: IssuesBoardsCountMetric

Generic metrics

You can use generic metrics for other metrics, for example, an instance’s database version.

  • value: Specifies the value of the metric.
  • available?: Specifies whether the metric should be reported. The default is true.

Example of a merge request that adds a generic metric.

module Gitlab
  module Usage
    module Metrics
      module Instrumentations
        class UuidMetric < GenericMetric
          value do
            Gitlab::CurrentSettings.uuid
          end
        end
      end
    end
  end
end

Prometheus metrics

This instrumentation class lets you handle Prometheus queries by passing a Prometheus client object as an argument to the value block. Any Prometheus error handling should be done in the block itself.

  • value: Specifies the value of the metric. A Prometheus client object is passed as the first argument.
  • available?: Specifies whether the metric should be reported. The default is true.

Example of a merge request that adds a Prometheus metric.

module Gitlab
  module Usage
    module Metrics
      module Instrumentations
        class GitalyApdexMetric < PrometheusMetric
          value do |client|
            result = client.query('avg_over_time(gitlab_usage_ping:gitaly_apdex:ratio_avg_over_time_5m[1w])').first

            break FALLBACK unless result

            result['value'].last.to_f
          end
        end
      end
    end
  end
end

Create a new metric instrumentation class

The generator takes the class name as an argument and the following options:

  • --type=TYPE Required. Indicates the metric type. It must be one of: database, generic, redis, numbers.
  • --operation Required for database & numbers type.
    • For database it must be one of: count, distinct_count, estimate_batch_distinct_count, sum, average.
    • For numbers it must be: add.
  • --ee Indicates if the metric is for EE.
rails generate gitlab:usage_metric CountIssues --type database --operation distinct_count
        create lib/gitlab/usage/metrics/instrumentations/count_issues_metric.rb
        create spec/lib/gitlab/usage/metrics/instrumentations/count_issues_metric_spec.rb

Migrate Service Ping metrics to instrumentation classes

This guide describes how to migrate a Service Ping metric from lib/gitlab/usage_data.rb or ee/lib/ee/gitlab/usage_data.rb to instrumentation classes.

  1. Choose the metric type:
  1. Determine the location of instrumentation class: either under ee or outside ee.

  2. Generate the instrumentation class file.

  3. Fill the instrumentation class body:

  4. Generate the metric definition file.

  5. Remove the code from lib/gitlab/usage_data.rb or ee/lib/ee/gitlab/usage_data.rb.

  6. Remove the tests from spec/lib/gitlab/usage_data.rb or ee/spec/lib/ee/gitlab/usage_data.rb.

Troubleshoot metrics

Sometimes metrics fail for reasons that are not immediately clear. The failures can be related to performance issues or other problems. The following pairing session video gives you an example of an investigation in to a real-world failing metric.

See the video from: Product Intelligence Office Hours Oct 27th to learn more about the metrics troubleshooting process.