exploring-endpoint-execution-logs

Query the endpoint's execution logs filtered to ERROR level. Each failed run logs one entry with an error token (e.g. error=ResolutionError) and the execution path, telling you what broke and where.

Search the endpoint's execution logs for the cache token, such as cache=miss or cache=hit. Each run's log entry records whether the query result cache was used, except on ducklake paths where the token is omitted.

Each run emits exactly one entry with key=value tokens: path (materialized, inline, ducklake), cache (hit/miss), duration_ms, rows, version, and error on failures. Level is INFO on success and ERROR on failure.

Yes, query the log_entries table with log_source='endpoints' and log_source_id set to the endpoint UUID from endpoint-get. Group by day and level to compute run counts and error rates over a period.

Execution logs are retained for approximately 90 days due to the log_entries TTL. Runs older than that will not appear in the log viewer or SQL queries.

Use execution logs to answer what happened during a run and why it failed. Use performance diagnosis when deciding whether to materialize the endpoint or change cache TTL, since that reasons over config and query_log cost metrics.