Welcome to the Metrics 2.0 beta!
This experience is a significant upgrade to how Honeycomb ingests and queries time series metrics, and is built on top of the OpenTelemetry Metrics Data Model.
This guide outlines what’s new, how the beta works, and what to try out.
What’s New in Metrics 2.0
The Metrics 2.0 beta introduces a more structured, OpenTelemetry-aligned approach to querying metrics in Honeycomb.
New capabilities give you greater flexibility and accuracy when analyzing time series data.
Highlights include:
Support for OTLP Metric Types:
Native support for OpenTelemetry metric types, including gauges, delta and cumulative sums, and histograms.
Temporal Aggregation:
A new type of timeseries-aware query that normalizes and time-aligns data points within each individual timeseries.
This ensures values are comparable across timeseries and supports accurate aggregation.
The beta includes four temporal aggregation functions: RATE(), INCREASE(), SUMMARIZE(), and LAST().
To learn more about temporal aggregation and supported functions, visit Temporal Aggregation Concepts.
Automatic Aggregation Defaults:
Honeycomb automatically applies the appropriate temporal aggregation function based on a metric’s type and metadata (monotonicity and temporality).
This gives you good results out of the box without needing to configure each metric manually.
To learn more about default behavior, visit Applying Temporal Aggregation Functions: Default Behavior.
Query-Scoped Calculated Fields for Metrics:
Override the automatic aggregation defaults directly in your query using calculated fields.
This lets you:
Metrics Correlations:
Connect events and metrics data to understand how system resources impact application behavior.
When you query an events dataset, access correlated metrics, such as CPU utilization or memory usage, directly alongside your query results.
To learn more about Metrics Correlations, visit Metrics Correlations.
Trigger Support for Metrics:
Alert on latency, utilization, and more using metrics-based queries and existing integrations. To learn more about metrics-based Triggers, visit Triggers with Metrics 2.0.
Native Histogram Support:
Histograms are stored as distributed structures and support accurate percentiles and flexible merging across time and groupings.
How the Beta Works
Once you are opted in, Honeycomb handles the rest:
No Instrumentation Changes Needed:
You don’t need to modify your code—just opt in.
Dual Ingest Automatically Enabled:
All OTLP metrics are duplicated to a new dataset named Metrics.
If you already have a dataset with this name, you may see the beta dataset labeled as OTLP Metrics or Metrics Data.
Important
During the beta, the Metrics dataset is excluded from environment-wide queries (all datasets in $environment). Environment-wide queries will return results from all other datasets in the environment. This helps avoid mixing early Metrics data with production datasets during the beta period.
No Additional Cost:
Honeycomb covers all storage and query costs for the duplicated metrics data during the beta.
Your Role in the Beta:
Try out the new functionality, tell us what is working well, and share what needs improvement.
What to Try
To get started:
Explore the new Metrics dataset using the Query Builder.
Run queries against your data. You should see sensible results by default. Honeycomb automatically applies the appropriate temporal aggregation function (like LAST(), SUMMARIZE() or INCREASE()) based on the metadata from your telemetry.
Explore Metrics Correlations. Run an events query, then select Correlations > Metrics to analyze system resource metrics alongside your application data.
Add queries to Boards and validate your visualizations.
Create triggers from your metrics queries and test alerting behavior.
Share feedback on anything confusing, surprising, or cool!
Known Limitations
During the beta, the Metrics dataset is excluded from environment-wide queries (all datasets in $environment).
Environment-wide queries will return results from all other datasets in the environment.
This helps avoid mixing early Metrics data with production datasets during the beta period.
Support for time series math (for example, arithmetic across time series) and nested queries is coming in a future release.
Enabled Features in this Beta
Metrics 2.0 introduces foundational improvements to how you ingest, query, and alert on metric data in Honeycomb.
The following features are enabled as part of this beta.
Ingest & Query Enhancements
Metrics 2.0 builds on your existing OTLP metrics pipeline and expands how you can explore your data.
Metrics ingestion and querying are enabled for all data sent through our OTLP metrics pipeline.
Expanded query capabilities: The query engine now supports:
Metrics 2.0 introduces new temporal aggregation functions that make it easier to explore time series behavior directly in your queries.
LAST(metric): Returns the most recent value for each step, based on timestamp.
SUMMARIZE(metric): Sums all data points in a step, interpolating at step boundaries to avoid double-counting and gaps. A single data point may be partially represented in two steps but fully counted in both.
INCREASE(metric [, range_interval_seconds]): Calculates the difference between the first and last values within each step individually.
Automatically accounts for counter resets.
RATE(metric [, range_interval_seconds]): Calculates the rate of change by dividing INCREASE by the number of seconds in the time range, resulting in the rate of increase per second.
Metrics Correlations connects your events and metrics data, helping you understand how system-level resource performance relates to application behavior.
Access correlated metrics directly alongside event queries without switching datasets.
Honeycomb automatically triggers parallel queries against your metrics dataset using the same time window and compatible filters.
Analyze system metrics (CPU, memory, network) in context with application events to identify resource-related issues.
Metrics 2.0 stores histograms as distributed data structures, enabling more accurate and flexible analysis:
Compute accurate percentiles (for example, p99, p95) directly in queries.
Merge histograms across time windows or grouped dimensions.
Seamlessly handle histograms with cumulative sums.
What’s Not Supported
Saving temporal aggregation functions as persistent calculated fields (query-scoped use only).
Nesting temporal aggregation functions or using arithmetic across time series.
Creating SLOs using metrics.
Deleting metrics datasets while enrolled in the beta.
Including the Metrics dataset in environment-wide queries (all datasets in $environment) during the beta.
This helps avoid mixing early Metrics data with production datasets during the beta period.
Feedback Welcome
This is an active beta, so your feedback is crucial!
We’d love to hear from you—whether it’s a bug, a friction point, or something you found unexpectedly powerful.
Ready to Join or Share Feedback?
You’re welcome to request access to the beta or share feedback at any time:
