CoreWeaveData Console

OpenTelemetry (Preview)

The OpenTelemetry view explores telemetry that is already stored in ClickHouse. It correlates a selected trace with its spans, logs, and nearby metrics. It is not an OTLP receiver or telemetry ingestion pipeline.

Before you start

The selected node must expose these tables:

  • otel.logs
  • otel.traces
  • otel.metrics

The tables must use the OpenTelemetry columns expected by the preview. Create and populate them through ClickHouse. To query the tables through ElasticQuery, create them with table_disk = true so they can be synchronized as read-only tables.

Open the view

  1. Open a ClickHouse service.
  2. Select Monitoring and then OpenTelemetry (Preview).
  3. Choose the time window.
  4. Choose ClickHouse or ElasticQuery, then select a node that exposes the otel database.
  5. Optionally filter by service.
  6. Select a trace.

The trace selector shows the most recent traces in the selected time window.

Explore a trace

After selecting a trace, the view shows:

  • the root span, duration, span count, and error count;
  • a waterfall of the trace’s spans and their parent-child relationships;
  • logs correlated by trace ID;
  • metrics correlated by exemplar trace ID or by service and time.

The overview charts summarize log volume, trace latency, and metric volume for the selected time window and service. Chart definitions are SQL-backed and can be inspected or edited.

ClickHouse and ElasticQuery

Choose ClickHouse to inspect telemetry on a ClickHouse node. ClickHouse is also where telemetry is inserted and where materialized views or other write processing run.

Choose ElasticQuery to inspect synchronized telemetry from a read-only query node. ElasticQuery does not accept inserts, mutations, or DDL, and querying mounted tables does not execute materialized views.

Preview limitations

  • The database and table names are currently fixed to otel.logs, otel.traces, and otel.metrics.
  • The tables must contain the columns expected by the current preview.
  • Correlation depends on trace IDs, exemplar trace IDs, service names, and timestamps present in the data.
  • Use the SQL Console for arbitrary queries and deeper investigation.