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

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

Usage Ping Guide

• Introduced in GitLab Enterprise Edition 8.10.
• More statistics were added in GitLab Enterprise Edition 8.12.
• Moved to GitLab Core in 9.1.
• More statistics were added in GitLab Ultimate 11.2.

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

For more information about Telemetry, see:

• Telemetry Guide
• Snowplow Guide

More useful links:

• Telemetry 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 within 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.
• Once usage ping is enabled, GitLab will gather data from the other instances and will be able to 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 Score,which gives you an overview of your entire instance’s adoption of Concurrent DevOps from planning to monitoring.
• You will get better, more proactive support. (assuming that our TAMs and support organization used the data to deliver more value)
• You will 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, GitLab's S3 Bucket, GitLab's 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 Score (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.

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 four 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 on given column
• Alternative Counters: Used for settings and configurations
• Redis Counters: Used for in-memory counts.

NOTE: 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

There are two batch counting methods provided, Ordinary Batch Counters and Distinct Batch Counters. Batch counting requires indexes on columns to calculate max, min, and range queries. In some cases, a specialized index may need to be added 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 in order to use batch counting
• start: custom start of the batch counting in order to avoid complex min calculations
• end: custom end of the batch counting in order 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 in order to use batch counting
• batch_size: if none set it will use default value 10000 from Gitlab::Database::BatchCounter
• start: custom start of the batch counting in order to avoid complex min calculations
• end: custom end of the batch counting in order to avoid complex min calculations

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')

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

Redis HLL Counters

With Gitlab::Redis::HLL we have available data structures used to count unique values.

Implemented using Redis methods PFADD and PFCOUNT.

Recommendations:

• Key should expire in 29 days.
• If possible, data granularity should be a week. For example a key could be composed from the metric's name and week of the year, 2020-33-{metric_name}.
• Use a feature flag in order to have a control over the impact when adding new metrics.
• If possible, data granularity should be week, for example a key could be composed from metric name and week of the year, 2020-33-{metric_name}
• Use a feature flag in order to have a control over the impact when adding new metrics

Examples of implementation:

• Gitlab::UsageDataCounters::TrackUniqueActions
• Gitlab::Analytics::UniqueVisits

Example of usage:

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

# Redis HLL counter
counter = Gitlab::UsageDataCounters::TrackUniqueActions
redis_usage_data do
          counter.count_unique_events(
            event_action: Gitlab::UsageDataCounters::TrackUniqueActions::PUSH_ACTION,
            date_from: time_period[:created_at].first,
            date_to: time_period[:created_at].last
          )

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.

Example of usage:

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

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, since most of GitLab's architectural components publish metrics to it that can be queried back, aggregated, and included as usage data.

NOTE: 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.

In order to query Prometheus for metrics, a helper method is available that will 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. 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))

2. Generate the SQL query

Your Rails console will return 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

3. 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.

In order 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, because 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.

4. Add the metric definition

When adding, changing, or updating metrics, please update the Event Dictionary's Usage Ping table.

5. 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 counts column.

For further details, see the Process to add additional instrumentation to the Usage Ping.

6. 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.

7. Add a changelog file

Ensure you comply with the Changelog entries guide.

8. Ask for a Telemetry Review

On GitLab.com, we have DangerBot setup to monitor Telemetry related files and DangerBot will recommend a Telemetry review. Mention @gitlab-org/growth/telemetry/engineers in your MR for a review.

9. Verify your metric

On GitLab.com, the Product Analytics 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, since Usage Ping should degrade gracefully in the absence of a running Prometheus server.

There are currently 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 via 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 will trigger 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 will take the following form: registry.gitlab.com/gitlab-org/build/omnibus-gitlab-mirror/gitlab-ee:<VERSION>.
4. On your local machine, make sure you are logged in to the GitLab Docker registry. You can find the instructions for this in Authenticating to the GitLab Container Registry.
5. Once logged in, download the new image via 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, since it comes with a number of difficulties when emulating a real GitLab deployment.

The GDK is not currently 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 currently 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, since 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 will appear in Usage Ping.

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"
  },
  "avg_cycle_analytics": {
    "issue": {
      "average": 999,
      "sd": 999,
      "missing": 999
    },
    "plan": {
      "average": null,
      "sd": 999,
      "missing": 999
    },
    "code": {
      "average": null,
      "sd": 999,
      "missing": 999
    },
    "test": {
      "average": null,
      "sd": 999,
      "missing": 999
    },
    "review": {
      "average": null,
      "sd": 999,
      "missing": 999
    },
    "staging": {
      "average": null,
      "sd": 999,
      "missing": 999
    },
    "production": {
      "average": null,
      "sd": 999,
      "missing": 999
    },
    "total": 999
  },
  "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,
    "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
          },
          ...
        ],
        ...
      },
      ...
    ]
  }
}