[coverage] Conformance findings: TELEMETRY-001,TELEMETRY-002,TELEMETRY-003,TELEMETRY-005 #927

Summary

Surfaced by the multi-language coverage fan-out while conformance-testing these SPEC-IDs against databricks/databricks-sql-python. Each finding is committed as an expected-failure (xfail) test in the coverage PR — the test asserts the CORRECT (post-fix) behavior and stays red until THIS driver (databricks/databricks-sql-python) is fixed, then flips green as a tripwire.

Findings

• TELEMETRY-001 [thrift]: terminal statement event never populates sql_operation.operation_detail (no operation_type, no is_internal_call), so the sync-execute path is indistinguishable from the async submit path and internal-driver statements cannot be excluded from aggregates
  • failing test: test_successful_statement_execution_emits_terminal_operation_event (see the coverage PR diff under tests/)
• TELEMETRY-001 [sea]: on the kernel/SEA backend the terminal event reports execution_result=FORMAT_UNSPECIFIED for an inline result and is_compressed from the connection's requested LZ4 setting rather than the result's declared codec, plus operation_detail is never populated (no operation_type / is_internal_call)
  • failing test: test_successful_statement_execution_emits_terminal_operation_event (see the coverage PR diff under tests/)
• TELEMETRY-002 [thrift]: failure telemetry is exported (and correctly leaks no SQL text) but the error descriptor carries no sql_statement_id — export_failure_log reports only session_id — so it cannot be joined to the statement that failed
  • failing test: test_failed_statement_emits_sanitized_error_telemetry_without_sql_text (see the coverage PR diff under tests/)
• TELEMETRY-002 [sea]: on the kernel/SEA backend NO error telemetry is exported for a failed statement: kernel/_errors.wrap_kernel_exception constructs PEP-249 exceptions without the host_url that exc.Error.init needs in order to reach export_failure_log, so every kernel-path failure is invisible to telemetry
  • failing test: test_failed_statement_emits_sanitized_error_telemetry_without_sql_text (see the coverage PR diff under tests/)
• TELEMETRY-003 [thrift]: the async statement's terminal event reports no n_operation_status_calls and no operation_status_latency_millis, and is not tagged as the async issue path (operation_detail is never populated); the single-emission guarantee itself holds
  • failing test: test_async_statement_emits_one_terminal_event_with_status_poll_counters (see the coverage PR diff under tests/)
• TELEMETRY-003 [sea]: on the kernel/SEA backend the async statement's terminal event likewise reports no n_operation_status_calls / operation_status_latency_millis and no async operation_type, so status-poll work is unmeasurable; single emission holds
  • failing test: test_async_statement_emits_one_terminal_event_with_status_poll_counters (see the coverage PR diff under tests/)
• TELEMETRY-005 [sea]: on the kernel/SEA backend a CloudFetch/external-links result reports execution_result=FORMAT_UNSPECIFIED instead of EXTERNAL_LINKS (KernelResultSet exposes no results queue for latency_logger to classify) and is_compressed reflects the requested LZ4 setting rather than the codec the result declared
  • failing test: test_telemetry_reports_external_links_delivery_and_compression (see the coverage PR diff under tests/)
• TELEMETRY-001: terminal statement telemetry never populates sql_operation.operation_detail (no operation_type, no is_internal_call), so the sync-execute and async-submit paths are indistinguishable and driver-internal statements cannot be excluded from aggregates; on the kernel/SEA backend execution_result is additionally always FORMAT_UNSPECIFIED and is_compressed reports the connection's requested LZ4 setting instead of the result's declared codec
• TELEMETRY-002: on the kernel/SEA backend NO error telemetry is exported for a failed statement (kernel/_errors.wrap_kernel_exception builds exceptions without the host_url that exc.Error.init needs to reach export_failure_log), so failures are invisible to telemetry; on Thrift the descriptor is exported but carries no sql_statement_id, so it cannot be joined to the statement it describes
• TELEMETRY-003: an async-submitted statement's terminal event reports no status-poll bookkeeping — n_operation_status_calls and operation_status_latency_millis are never populated — and is not tagged as the async issue path, so poll work performed on the application's behalf is unmeasurable (single-emission itself is correct)
• TELEMETRY-005: on the kernel/SEA backend a CloudFetch/external-links result reports execution_result=FORMAT_UNSPECIFIED instead of EXTERNAL_LINKS (KernelResultSet exposes no results queue for latency_logger to classify) and is_compressed reflects the requested LZ4 setting rather than the codec the result declared, so result-delivery telemetry does not describe how the result actually arrived

Context

• The behavior was first fixed in a DIFFERENT driver — reference PR: https://github.com/databricks/databricks-sql-kernel/pull/272 — which seeded the shared language-neutral spec. This issue tracks the same conformance gap in databricks/databricks-sql-python; the reference PR is for cross-referencing the intended behavior, NOT a change to this repo.
• Coverage PR carrying the reproducing xfail test(s): https://github.com/databricks/databricks-driver-test/pull/1389