Limitations

Tracing Service is a managed distributed tracing solution that allows you to ingest, store, and analyze traces generated by your applications. However, there are some limitations to be aware of when using the service.

The following limits follow the Grafana Tempo defaults that back the platform.

Data retention

The Tracing Service retains trace data for a limited period. The default retention period is 30 days. After this period, trace data is automatically deleted and is no longer available for analysis.

Maximum trace size

A single trace cannot exceed 5 MB of combined span data. If the spans that share one trace ID add up to more than this limit, the trace is rejected at ingestion and does not appear in Grafana. This usually affects very long-running requests or traces with many spans or attributes. To stay within the limit, reduce the number or size of spans and attributes, or split very large operations into separate traces.

Ingestion rate limits

Trace ingestion is rate-limited to protect platform stability. The default sustained limit is 15 MB/s, with a burst allowance of 20 MB. When an exporter exceeds this rate, the platform returns a rate-limit response: HTTP 429 for OTLP/HTTP and RESOURCE_EXHAUSTED for OTLP/gRPC. Your OpenTelemetry SDK or Collector retries with backoff. Sustained traffic above the limit results in discarded spans.

Live traces

The platform holds a bounded number of concurrent in-progress traces, with a default of 10,000. If your applications produce more simultaneously open traces than this, new traces are refused until in-flight traces are completed and flushed.

Late-arriving spans

Tempo groups spans into a trace by trace ID as they arrive. There is no fixed maximum trace duration, but spans that arrive after a trace has already been flushed to storage are stored as a separate, partial trace rather than merged into the original. Keep clock skew across services small and export spans promptly to avoid fragmented traces.

Search availability delay

Newly ingested traces are not instantly searchable. Traces are briefly buffered before they are flushed and indexed, so a short delay of a few seconds can occur between sending a trace and finding it through TraceQL search in Grafana. Lookups by exact trace ID become available sooner.

Query and search limits

TraceQL searches are bounded by a maximum query time range and a maximum number of returned results. Very broad searches across large time windows may be truncated or rejected. Narrow the time range or add more selective filters to return complete results.