Honeycomb

Honeycomb is an observability platform that enables engineering teams to understand the behavior of complex, distributed systems by analyzing high-cardinality, high-dimensionality event data. Founded by charity majors and Liz Fong-Jones, prominent observability thought leaders, Honeycomb was built on the insight that modern distributed systems require a fundamentally different approach to debugging than traditional metrics and logs can provide.

The central concept in Honeycomb is the wide event. Instead of recording narrow metrics (a single counter or gauge) or unstructured log lines, Honeycomb encourages instrumenting applications to emit wide, structured events that capture the full context of each request, transaction, or operation. A single event might contain dozens of fields including the user ID, request path, HTTP status code, database query count and duration, cache hit/miss, feature flags in effect, downstream service latencies, and any other relevant context. This wide event model enables slicing and dicing data along any dimension after the fact.

High cardinality is a core design principle of Honeycomb. Many observability tools struggle with high-cardinality dimensions such as user IDs, trace IDs, and request IDs, because storing and indexing millions of unique values is expensive with traditional time-series databases. Honeycomb is built from the ground up to handle high-cardinality data efficiently, enabling queries that group by user ID, trace ID, or any other high-cardinality field without performance degradation.

The query engine in Honeycomb is interactive and collaborative. Teams write queries using a visual builder that selects breakdowns (GROUP BY dimensions), calculations (COUNT, COUNT DISTINCT, P99, HEATMAP, etc.), and filters, then visualize the results as time-series charts, heatmaps, or tables. The Derived Columns feature enables creating computed fields from raw event data at query time, without requiring schema changes.

BubbleUp is Honeycomb's automated anomaly investigation feature. When an elevated error rate or latency spike is observed on a chart, BubbleUp automatically identifies which field values are over-represented in the anomalous subset compared to the baseline, pinpointing the specific combination of user attributes, service versions, feature flags, or other factors that characterize the problem. This dramatically accelerates root cause analysis.

Trace Waterfall provides distributed trace visualization where the full request path across microservices is shown as a waterfall diagram, with each span showing duration, service name, and custom attributes. Traces are correlated with the wide events in the same Honeycomb environment, enabling direct pivoting from a specific span to all events with matching context.

Honeycomb uses OpenTelemetry as its preferred instrumentation standard, supporting traces, metrics, and logs via the OpenTelemetry Protocol (OTLP).