Metrics and monitoring

The Metrics and Monitoring dashboard provides deep analytics on your merge queue’s performance, helping you identify bottlenecks, measure improvements, and optimize your workflow.

The Metrics and Monitoring dashboard provides deep analytics on your merge queue’s performance, helping you identify bottlenecks, measure improvements, and optimize your workflow.

Your merge experience directly impacts the velocity and productivity of your development team. Merge Queue Metrics provides observability for the health of your Trunk Merge Queue, so you can discover issues early and make informed optimizations.

Access metrics

You can access the metrics in your Trunk Merge Queue by navigating to the Merge Queue > Health tab.

CI Time and CI Jobs Triggered charts are only available for GitHub Actions.

Filter Metrics by Impacted Targets

When running in Parallel Mode, you can filter your merge queue health metrics by impacted targets to analyze performance for specific parts of your codebase.

Why Filter by Impacted Targets?

In repositories with multiple teams or distinct components (like a TypeScript/Python monorepo), different parts of your codebase may have different merge characteristics. Filtering by impacted targets helps you:

• Analyze team-specific performance - See how PRs from different teams move through the queue
• Identify bottlenecks by component - Determine if certain targets have slower merge times
• Optimize strategically - Focus queue configuration improvements on your highest-priority code paths
• Demonstrate value - Show engineering leadership how parallel mode benefits specific teams or projects
• Ensure fairness - Verify that all teams experience similar queue performance

How to Use the Filter

1. Navigate to Merge Queue > your repository > Health tab in the Trunk web app
2. Locate the Impacted Targets filter dropdown at the top of the metrics dashboard
3. Select one or more targets to filter by:
   • All Targets (default) - Shows aggregate metrics across all PRs
   • Specific target names - Shows metrics only for PRs affecting that target (e.g., frontend, backend, //services/api)
4. All charts and metrics on the page will update to reflect only PRs impacting the selected targets

Understanding the Data

Impacted targets are set when a PR enters the queue

Each PR’s impacted targets are calculated based on which files changed and which parts of your codebase are affected. For details on how impacted targets are computed, see Parallel Queues - Impacted Targets.

PRs can affect multiple targets

A PR that changes both frontend and backend code will be counted in metrics when filtering by either frontend OR backend. This means the numbers may not sum to 100% when viewing multiple target filters separately.

“All Targets” shows aggregate performance

Selecting “All Targets” displays metrics for every PR, regardless of which targets it impacts. This is the default view and shows overall queue health.

Requirements

Parallel Mode must be enabled

Impacted target filtering is only available when your merge queue is running in Parallel Mode. Repositories in Single Mode do not track impacted targets.

Impacted targets must be uploaded

Your CI workflow must calculate and upload impacted targets for each PR. See the Parallel Queues documentation for setup instructions using:

• Bazel
• Nx
• Custom build systems

Time buckets

The date ranges selector at the top left of the dashboard allows you to filter the data displayed by date and time. You can display time buckets by the day or hour in the day/hour dropdown.

The metrics displayed only include data that have completed within the time range, jobs started but not completed during the selected time will not be displayed.

When working across multiple time zones, enable Time in UTC to ensure everyone sees the same data.

Conclusion count

Conclusion count displays the number of pull requests that exited the merge queue during each time bucket. This includes passes, failures, and cancellations. Passes and failures signal a PR that was tested in the queue to completion, while canceled signals that the request to merge terminated before testing finished or before testing began.

Conclusion counts are an important signal to potential bottlenecks or underlying issues with your merging process, as a failure or cancellation in the merge queue can force other PRs to restart their testing. A spike in the number of failures or passes can indicate a potential problem to investigate.

Conclusions are tagged with a reason to give further insights into how merges pass or fail in the queue. You can show or hide conclusions of a particular reason by using the + Add button.

Category	Reason	Description
✅ Pass	Merged by Trunk	Passed all tests in Merge Queue and merged by Trunk
✅ Pass	Merged manually	User manually merged the PR in Git
❌ Failure	Test run timeout	User-defined timeout for tests exceeded
❌ Failure	Failed Tests	Required test failed while testing the PR in the merge queue
❌ Failure	Merge conflict	A (git) merge conflict encountered
❌ Failure	Config parsing failure	Malformed trunk.yaml that couldn’t be parsed
❌ Failure	Config bad version	Invalid version field in trunk.yaml
❌ Failure	Config bad required statuses	Failed to parse required statuses in trunk.yaml
❌ Failure	No required statuses	No source for required tests was found in trunk.yaml or branch protection settings
❌ Failure	GitHub API Failed	GitHub returned an error to us that could not be resolved while processing the PR
❌ Failure	PR updated at merge time	PR updated as Trunk was attempting to merge it
🚫 Cancel	Canceled by user	PR explicitly canceled by user
🚫 Cancel	PR closed	PR closed (not merged)
🚫 Cancel	PR pushed to	New commits pushed to the PR branch while in the merge queue
🚫 Cancel	PR draft	PR was converted to a draft, which cannot be merged
🚫 Cancel	PR base branch changed	Base branch of PR in the merge queue changed
🚫 Cancel	Admin requested	Trunk employee canceled PR during a support session (extreme cases)
🚫 Cancel	A PR in the stack had its base branch changed	A member of the PR stack had its base branch changed while in the queue (stacked PRs only)
🚫 Cancel	A PR in a PR stack was closed	A member of the PR stack was closed while in the queue (stacked PRs only)
🚫 Cancel	PR was merged as part of a different stack	The PR was already merged through a different stack (stacked PRs only)
🚫 Cancel	Part of this PR’s stack was pushed to	New commits were pushed to a PR in the stack while in the queue (stacked PRs only)

Time in queue

Time in queue shows how long each PR spends in the Merge Queue from the moment the PR enters the queue to the moment when it exits the queue, either from merging, failing, or being canceled.

Understanding the amount of time a pull request spends in the queue is important for ensuring your merge process continues to ship code quickly. A spike in the time to merge indicates a slowdown somewhere that’s impacting all developers. For example, it’s taking longer to run tests on PRs, PRs are waiting too long to start testing, or constant failures in the queue are causing PRs to take longer to merge

The time in queue can be displayed as different statistical measures. You can show or hide them by using the + Add button.

Measure	Explanation
Average	Average of all time in queue during the time bucket
Minimum	The shortest time in queue in the time bucket.
Maximum	The longest time in queue in the time bucket.
Sum	The total of all time in queue added together.
P50	The value below 50% of the time in queue falls.
P95	The value below 95% of the time in queue falls.
P99	The value below 99% of the time in queue falls.

Drill down into metrics

From the Conclusion count and Time in queue charts, you can drill into any point or window on the graph to see the exact pull requests that made up those numbers.

Why Drill Down?

Aggregated charts tell you that something happened — drilling down tells you which PRs caused it. This makes it easy to:

• Track down outliers — if the P99 on Time in queue spikes, drill into that bucket to find the specific PR that dragged the tail out.
• Investigate failure spikes — click a bar on Conclusion count where failures jumped and see exactly which PRs failed and why.
• Audit a time window — pull the full list of PRs merged, failed, or canceled during an incident window or release cut.
• Answer one-off questions — “which PRs merged between 2pm and 4pm yesterday?” without writing a query against the Prometheus endpoint.

Select Data Points

You have two ways to select:

• Click a single data point to see the PRs in that time bucket.
• Click and drag across the chart to select a range of data points spanning multiple time buckets.

Once a selection is made, a View PRs button appears. Click it to open the list of PRs that make up the selection.

Review the PR List

The PR list page shows every PR included in your selection, along with:

• Conclusion — whether the PR merged, failed, or was cancelled.
• Reason — the specific cause behind the conclusion (for example, Merged by Trunk, Required status failed, PR closed). See the Conclusion count table for the full list.
• Time in queue — how long the PR spent in the merge queue from entry to exit.

Both columns are sortable, so you can quickly surface the longest-running PRs in a window or group all failures of the same type together.

Drill down is currently available on the Conclusion count and Time in queue charts. Additional Health charts will support the same interaction as they land in the UI.

---

Prometheus metrics endpoint

Trunk exposes merge queue metrics in Prometheus text exposition format via a scrapable API endpoint. Use this to build custom Grafana dashboards, set up alerts, or integrate merge queue health into your existing observability stack.

The Prometheus metrics endpoint is available to all Merge Queue users.

Endpoint

GET https://api.trunk.io/v1/getMergeQueueMetrics

Authenticate with your Trunk API token using the x-api-token header.

Query parameters:

Parameter	Required	Description
repo	No	Repository in owner/name format (e.g., my-org/my-repo). If omitted, returns metrics for all repositories in the organization. Must be provided together with repoHost.
repoHost	Conditional	Repository host (e.g., github.com). Required if repo is specified.

The response uses content type text/plain; version=0.0.4; charset=utf-8 (standard Prometheus format).

Available metrics

All metrics include these labels:

Label	Description	Example values
repo	Repository name	my-org/my-repo
branch	Base branch name	main, develop
queue_type	Queue type	main or bisection

Point-in-time gauges

These metrics reflect the current state of your merge queue.

Metric	Type	Description
mq_depth_current	Gauge	Number of PRs currently in the queue (excludes PRs that are waiting to be mergeable before being admitted to the queue)
mq_awaiting_mergeability	Gauge	Number of PRs waiting for prerequisites like required reviews or status checks
mq_testing_slots_active	Gauge	Number of PRs currently in TESTING state (active CI slots in use)

Rolling 1-hour window metrics

These metrics summarize activity over a sliding 1-hour window. They update continuously as the window advances.

Metric	Type	Extra labels	Description
mq_pr_conclusions_1h_total	Gauge	conclusion (merged, failed, cancelled)	PRs that exited the queue in the last hour
mq_pr_restarts_1h_total	Gauge	—	PR restarts (TESTING to PENDING transitions) in the last hour
mq_pr_wait_duration_1h_seconds	Histogram	le (bucket boundary)	Distribution of time PRs spent waiting before testing starts
mq_pr_test_duration_1h_seconds	Histogram	le (bucket boundary)	Distribution of time PRs spent in the testing phase
Each histogram emits _bucket{le="..."}, _sum, and _count series. Bucket boundaries (in seconds): 60, 300, 600, 900, 1800, 3600, 5400, 7200, +Inf.

Rolling window metrics use gauge semantics, not true Prometheus counters. They represent a snapshot of the last hour, not cumulative totals. PromQL functions like rate() and increase() are not meaningful on these metrics. Use the values directly instead.

Scrape configuration

Configure your Prometheus instance to scrape the Trunk metrics endpoint:

scrape_configs:
  - job_name: trunk-merge-queue
    scrape_interval: 60s
    scheme: https
    static_configs:
      - targets: ['api.trunk.io']
    metrics_path: /v1/getMergeQueueMetrics
    params:
      repo: ['my-org/my-repo']
      repoHost: ['github.com']
    http_headers:
      x-api-token:
        values: ['<your-trunk-api-token>']

To scrape metrics for all repositories in your organization, omit both the repo and repoHost parameters.

Datadog Agent configuration

You can ingest Trunk merge queue metrics into Datadog using the Datadog Agent’s built-in OpenMetrics integration. This lets Datadog scrape the Prometheus endpoint directly without requiring a separate Prometheus server.

1. Enable the OpenMetrics integration

Create or edit /etc/datadog-agent/conf.d/openmetrics.d/conf.yaml:

instances:
  - openmetrics_endpoint: https://api.trunk.io/v1/getMergeQueueMetrics?repo=my-org/my-repo&repoHost=github.com
    namespace: trunk_merge_queue
    metrics:
      - mq_.*
    headers:
      x-api-token: <your-trunk-api-token>
    min_collection_interval: 60
    send_distribution_buckets: true

To collect metrics for all repositories in your organization, omit the query parameters:

    openmetrics_endpoint: https://api.trunk.io/v1/getMergeQueueMetrics

2. Restart the Datadog Agent

sudo systemctl restart datadog-agent

3. Validate

sudo -u dd-agent -- datadog-agent check openmetrics

All metrics are prefixed with your configured namespace value. For example, mq_depth_current becomes trunk_merge_queue.mq_depth_current in Datadog.

Example queries

Queue health alerts:

# Alert when queue depth exceeds threshold
mq_depth_current{branch="main"} > 10

# Max queue depth over the last hour
max_over_time(mq_depth_current{branch="main"}[1h])

# CI utilization (if you have 8 concurrency slots)
mq_testing_slots_active{branch="main",queue_type="main"} / 8

Failure analysis:

# Failure rate over the last hour
mq_pr_conclusions_1h_total{conclusion="failed"}
  /
ignoring(conclusion) sum(mq_pr_conclusions_1h_total)

# Alert on high failure count
mq_pr_conclusions_1h_total{conclusion="failed"} > 5

Duration analysis:

# P90 wait time (time before testing starts)
histogram_quantile(0.90, sum(mq_pr_wait_duration_1h_seconds_bucket) by (le))

# Average wait time
mq_pr_wait_duration_1h_seconds_sum / mq_pr_wait_duration_1h_seconds_count

# Restart ratio (restarts per merge)
mq_pr_restarts_1h_total / mq_pr_conclusions_1h_total{conclusion="merged"}

Sample output

# HELP mq_depth_current PRs currently in the queue
# TYPE mq_depth_current gauge
mq_depth_current{repo="my-org/my-repo",branch="main",queue_type="main"} 4

# HELP mq_awaiting_mergeability Number of PRs currently awaiting mergeability
# TYPE mq_awaiting_mergeability gauge
mq_awaiting_mergeability{repo="my-org/my-repo",branch="main",queue_type="main"} 1

# HELP mq_testing_slots_active PRs currently in TESTING state
# TYPE mq_testing_slots_active gauge
mq_testing_slots_active{repo="my-org/my-repo",branch="main",queue_type="main"} 3

# HELP mq_pr_conclusions_1h_total PRs exiting the queue in last hour
# TYPE mq_pr_conclusions_1h_total gauge
mq_pr_conclusions_1h_total{repo="my-org/my-repo",branch="main",queue_type="main",conclusion="merged"} 12
mq_pr_conclusions_1h_total{repo="my-org/my-repo",branch="main",queue_type="main",conclusion="failed"} 1
mq_pr_conclusions_1h_total{repo="my-org/my-repo",branch="main",queue_type="main",conclusion="cancelled"} 0

# HELP mq_pr_restarts_1h_total PR restarts in last hour
# TYPE mq_pr_restarts_1h_total gauge
mq_pr_restarts_1h_total{repo="my-org/my-repo",branch="main",queue_type="main"} 2