From 8194cb5a1d90d734ae52375033648383ba345b7d Mon Sep 17 00:00:00 2001 From: Alex Garcia Date: Thu, 30 Jul 2026 17:30:52 -0700 Subject: [PATCH 01/10] Add opentelemetry-api dependency and datasette/telemetry.py scaffolding Datasette core is gaining OpenTelemetry spans alongside the existing hand-rolled tracer. This commit only lays the groundwork - no span is emitted yet. Core takes a runtime dependency on opentelemetry-api and nothing more. It deliberately never creates a TracerProvider, configures an exporter, or touches sampling: that belongs to whoever runs Datasette, normally via an opentelemetry-instrument agent. Owning a provider in core was tried in an earlier design and produced a cross-request span leak, a process-global provider that tests could not tear down, and a sampling env var that silently blanked output. With no provider installed every span is a NonRecordingSpan and costs approximately nothing. datasette/telemetry.py exposes the module-level tracer plus sql_attribute(), which truncates SQL to 2048 characters. On a public instance the SQL is attacker-controlled and unbounded - someone can paste a 10MB query into ?sql= - so it must never reach a telemetry pipeline verbatim. opentelemetry-sdk goes in the dev dependency group only, because the test suite needs it to assert on spans while the package itself must not import it. tests/test_telemetry.py enforces that by importing datasette in a fresh interpreter and inspecting sys.modules, which catches a lazy import inside a function body that a grep would miss. conftest.py gains a session-scoped autouse fixture installing an SDK provider with an InMemorySpanExporter. It has to be session-scoped because set_tracer_provider() is effectively once-per-process - a second call logs a warning and is ignored. SimpleSpanProcessor rather than BatchSpanProcessor, so assertions made right after a request never race a background export thread. The otel_spans fixture that later tickets assert against is added here too. test_datasette_package_never_imports_the_sdk is moved to the front of the run. Late in a serial run the pytest process holds enough threads that the fork half of subprocess' fork+exec segfaults the interpreter on macOS/CPython 3.13. That reproduces with any subprocess call in that position on an unmodified tree, so it is a pre-existing hazard rather than something this commit introduces; the repo already moves its other subprocess-spawning tests to the front for related reasons. Co-Authored-By: Claude Opus 5 --- datasette/telemetry.py | 24 ++++++++++++++++ pyproject.toml | 2 ++ tests/conftest.py | 62 +++++++++++++++++++++++++++++++++++++++++ tests/test_telemetry.py | 25 +++++++++++++++++ 4 files changed, 113 insertions(+) create mode 100644 datasette/telemetry.py create mode 100644 tests/test_telemetry.py diff --git a/datasette/telemetry.py b/datasette/telemetry.py new file mode 100644 index 00000000..6f5ed093 --- /dev/null +++ b/datasette/telemetry.py @@ -0,0 +1,24 @@ +""" +OpenTelemetry integration for Datasette core. + +Core depends on `opentelemetry-api` only. It never creates a +`TracerProvider`, never configures an exporter, and never touches +sampling - that is the responsibility of whoever is running Datasette +(an `opentelemetry-instrument` agent, a future plugin, or a test +harness). With no provider installed every span produced here is a +`NonRecordingSpan` and costs approximately nothing. +""" + +from opentelemetry import trace as otel_trace + +tracer = otel_trace.get_tracer("datasette") + +MAX_SQL_LENGTH = 2048 + + +def sql_attribute(sql: str) -> str: + "Truncate SQL text so it is safe to attach to a span as an attribute." + sql = sql.strip() + if len(sql) <= MAX_SQL_LENGTH: + return sql + return sql[:MAX_SQL_LENGTH] + "…[truncated]" diff --git a/pyproject.toml b/pyproject.toml index e658955f..9dd231d7 100644 --- a/pyproject.toml +++ b/pyproject.toml @@ -40,6 +40,7 @@ dependencies = [ "setuptools", "pip", "pydantic>=2", + "opentelemetry-api>=1.37", ] [project.urls] @@ -70,6 +71,7 @@ dev = [ "cogapp>=3.3.0", "multipart-form-data-conformance==0.1a0", "ruff>=0.16.0", + "opentelemetry-sdk>=1.37", # docs "Sphinx==7.4.7", "furo==2025.9.25", diff --git a/tests/conftest.py b/tests/conftest.py index 12dce417..b01f111d 100644 --- a/tests/conftest.py +++ b/tests/conftest.py @@ -58,6 +58,62 @@ def find_free_port(): return sock.getsockname()[1] +_otel_span_exporter = None + + +@pytest.fixture(scope="session", autouse=True) +def _otel_provider(): + """ + Install a real OTel SDK TracerProvider + InMemorySpanExporter exactly + once, before any span is ever created in this process. + + This has to be session-scoped and autouse because + `opentelemetry.trace.set_tracer_provider()` is effectively + once-per-process: a second call logs a warning and is ignored. So the + install must happen exactly once, before anything asserts on spans. + + `datasette.telemetry.tracer` is a module-level `ProxyTracer`. Once a + provider exists, the first span it starts resolves a concrete tracer + and caches it permanently. It does *not* cache the no-op tracer, so + any span started before this fixture runs is merely lost rather than + poisoning the tracer for the rest of the process. If the SDK isn't + installed, do nothing: core spans stay no-op `NonRecordingSpan`s and + the rest of the suite is unaffected. + """ + global _otel_span_exporter + try: + from opentelemetry import trace as otel_trace + from opentelemetry.sdk.trace import TracerProvider + from opentelemetry.sdk.trace.export import SimpleSpanProcessor + from opentelemetry.sdk.trace.export.in_memory_span_exporter import ( + InMemorySpanExporter, + ) + except ImportError: + return + exporter = InMemorySpanExporter() + provider = TracerProvider() + # SimpleSpanProcessor exports synchronously on span end - no background + # batching thread, so assertions immediately after a request never race. + provider.add_span_processor(SimpleSpanProcessor(exporter)) + otel_trace.set_tracer_provider(provider) + _otel_span_exporter = exporter + + +@pytest.fixture +def otel_spans(): + """ + Function-scoped access to the finished-spans exporter: clears any spans + left over from previous tests, then yields the exporter so a test can + call `.get_finished_spans()` after making requests. Skips (rather than + fails) if the OTel SDK is not installed. + """ + pytest.importorskip("opentelemetry.sdk") + if _otel_span_exporter is None: + pytest.skip("OpenTelemetry SDK provider was not installed") + _otel_span_exporter.clear() + yield _otel_span_exporter + + @pytest.fixture def bare_ds(): """ @@ -168,6 +224,12 @@ def pytest_collection_modifyitems(config, items): move_to_front(items, "test_spatialite_error_if_attempt_to_open_spatialite") move_to_front(items, "test_package") move_to_front(items, "test_package_with_port") + # Same reason: this one shells out to a fresh interpreter. Late in a serial + # run the pytest process holds enough threads that the fork half of + # subprocess' fork+exec crashes the interpreter on macOS/CPython 3.13 + # (SIGSEGV/SIGBUS inside _execute_child). Reproduces with any subprocess + # call placed there, on an unmodified tree - running it first avoids it. + move_to_front(items, "test_datasette_package_never_imports_the_sdk") def move_to_front(items, test_name): diff --git a/tests/test_telemetry.py b/tests/test_telemetry.py new file mode 100644 index 00000000..b71be504 --- /dev/null +++ b/tests/test_telemetry.py @@ -0,0 +1,25 @@ +import subprocess +import sys + + +def test_datasette_package_never_imports_the_sdk(): + """ + Core depends on opentelemetry-api only. The SDK is a test dependency. + + Checked by importing datasette in a fresh process and inspecting + sys.modules, rather than by grepping, so a lazy `import + opentelemetry.sdk` inside a function body cannot slip past. + + conftest.py's pytest_collection_modifyitems() moves this test to the + front of the run by name - if you rename it, rename it there too. + """ + code = ( + "import datasette.app, datasette.database, datasette.telemetry, sys; " + "print([m for m in sys.modules if m.startswith('opentelemetry.sdk')])" + ) + result = subprocess.run( + [sys.executable, "-c", code], capture_output=True, text=True, check=True + ) + assert ( + result.stdout.strip() == "[]" + ), f"datasette imported the OpenTelemetry SDK: {result.stdout.strip()}" From b40b06f1cbd87115336b2aa3f3eda7c083b6bbf2 Mon Sep 17 00:00:00 2001 From: Alex Garcia Date: Thu, 30 Jul 2026 17:45:32 -0700 Subject: [PATCH 02/10] Emit a db.query span around Database.execute() Datasette's existing tracer times a "sql" block that wraps a good deal more than the query itself - queueing onto the thread pool, the pool wait, and result marshalling all disappear into one number. That is simonw/datasette#1730, "SQL tracing should much more closely track the SQL query execution", open since 2022. A db.query span here is the outer half of the answer; a later change adds the inner span drawn around the sqlite3 call itself, and the gap between the two is exactly the thread pool wait the current tracer folds away. The span carries OTel semantic-convention attributes (db.system, db.namespace, db.query.text) plus a few datasette.* ones. db.query.text goes through sql_attribute(), which caps it at 2048 characters, because on a public instance the SQL is attacker-supplied and unbounded. Only len(params) is recorded, never a parameter value. The existing `with trace(...)` wrapper stays exactly where it is and the new span nests inside it. This change removes nothing: ?_trace=1 and the trace_debug setting keep working unchanged. The two systems are independent code paths. Exception handling on the span is explicit rather than inherited from start_as_current_span's defaults, which would record the exception and set StatusCode.ERROR on anything passing through. That is wrong here because some SQL failures are the expected answer. ArrayFacet.suggest() runs json_type() against every column precisely to discover which ones raise "malformed JSON", and passes log_sql_errors=False to say so. Left to the defaults, a table with N text columns marks N queries per page as failed - burying genuine failures and tripping any alerting keyed on span status. Measured on a plain table page before this: 4 error spans out of 225, all expected. Suppressed errors now leave the status UNSET and set datasette.sql_error_suppressed instead, so they stay discoverable without reading as failures. QueryInterrupted still sets ERROR unconditionally. That is not quite right either - facet suggestion is designed to time out - but the fix needs its own reasoning and lands separately. Behaviour change worth calling out: time_limit_ms is hoisted out of sql_operation_in_thread so the span can record it on the event loop. It is therefore read at call time rather than at thread-execution time. Benign in practice, since ds.sql_time_limit_ms is set at startup, but it is a real change. Co-Authored-By: Claude Opus 5 --- datasette/database.py | 54 +++++++++++-- tests/test_telemetry.py | 169 ++++++++++++++++++++++++++++++++++++++++ 2 files changed, 217 insertions(+), 6 deletions(-) diff --git a/datasette/database.py b/datasette/database.py index e162d34e..f78eb925 100644 --- a/datasette/database.py +++ b/datasette/database.py @@ -11,8 +11,10 @@ from collections import namedtuple from pathlib import Path import sqlite_utils +from opentelemetry.trace import Status, StatusCode from .inspect import inspect_hash +from .telemetry import sql_attribute, tracer from .tracer import trace from .utils import ( call_with_supported_arguments, @@ -529,12 +531,11 @@ class Database: """Executes sql against db_name in a thread""" self._check_not_closed() page_size = page_size or self.ds.page_size + time_limit_ms = self.ds.sql_time_limit_ms + if custom_time_limit and custom_time_limit < time_limit_ms: + time_limit_ms = custom_time_limit def sql_operation_in_thread(conn): - time_limit_ms = self.ds.sql_time_limit_ms - if custom_time_limit and custom_time_limit < time_limit_ms: - time_limit_ms = custom_time_limit - with sqlite_timelimit(conn, time_limit_ms): try: cursor = conn.cursor() @@ -565,8 +566,49 @@ class Database: else: return Results(rows, False, cursor.description) - with trace("sql", database=self.name, sql=sql.strip(), params=params): - results = await self.execute_fn(sql_operation_in_thread) + # SIM117 wants these two context managers merged. They are kept nested + # deliberately: the hand-rolled tracer's wrapper is on its way out, and + # nesting makes removing it a single-line deletion. + with trace( # noqa: SIM117 + "sql", database=self.name, sql=sql.strip(), params=params + ): + # Exception handling is explicit rather than left to the context + # manager's defaults, so that callers passing log_sql_errors=False + # can be honoured - see the comment on the generic handler below. + with tracer.start_as_current_span( + "db.query", + record_exception=False, + set_status_on_exception=False, + ) as span: + span.set_attribute("db.system", "sqlite") + span.set_attribute("db.namespace", self.name) + span.set_attribute("db.query.text", sql_attribute(sql)) + span.set_attribute("datasette.time_limit_ms", time_limit_ms) + if params: + span.set_attribute("datasette.param_count", len(params)) + try: + results = await self.execute_fn(sql_operation_in_thread) + except QueryInterrupted as e: + span.set_status(Status(StatusCode.ERROR, str(e))) + span.set_attribute("datasette.interrupted", True) + span.record_exception(e) + raise + except Exception as e: + # log_sql_errors=False means the caller is probing and + # treats failure as an expected answer, not an error. + # Facet suggestion is the big one: it runs json_type() + # against every column precisely to find out which ones + # raise, so a table with N text columns would otherwise + # mark N queries per page as failed - burying real errors + # and setting off any alerting based on span status. + if log_sql_errors: + span.record_exception(e) + span.set_status(Status(StatusCode.ERROR, str(e))) + else: + span.set_attribute("datasette.sql_error_suppressed", True) + raise + span.set_attribute("datasette.truncated", results.truncated) + span.set_attribute("datasette.rows_returned", len(results.rows)) return results @property diff --git a/tests/test_telemetry.py b/tests/test_telemetry.py index b71be504..a13898ff 100644 --- a/tests/test_telemetry.py +++ b/tests/test_telemetry.py @@ -1,6 +1,31 @@ +import json +import sqlite3 import subprocess import sys +import pytest +from opentelemetry.trace import StatusCode + +from datasette.telemetry import MAX_SQL_LENGTH, sql_attribute + +SECRET_PARAM_VALUE = "SUPER_SECRET_PARAM_VALUE_XYZ_123" + +INVALID_SQL = "select this_is_not_valid_sql from nowhere" + + +def _db_query_spans(otel_spans): + return [span for span in otel_spans.get_finished_spans() if span.name == "db.query"] + + +def _all_attribute_values(otel_spans): + "Every attribute value across every finished span, for the 'no leaked param values' test." + values = [] + for span in otel_spans.get_finished_spans(): + values.extend((span.attributes or {}).values()) + for event in span.events: + values.extend((event.attributes or {}).values()) + return values + def test_datasette_package_never_imports_the_sdk(): """ @@ -23,3 +48,147 @@ def test_datasette_package_never_imports_the_sdk(): assert ( result.stdout.strip() == "[]" ), f"datasette imported the OpenTelemetry SDK: {result.stdout.strip()}" + + +@pytest.mark.asyncio +async def test_db_query_span_basic_attributes(ds_client, otel_spans): + response = await ds_client.get("/fixtures/-/query.json?sql=select+1") + assert response.status_code == 200 + + spans = _db_query_spans(otel_spans) + assert spans, "expected at least one db.query span" + span = spans[-1] + + assert span.attributes["db.system"] == "sqlite" + assert span.attributes["db.namespace"] == "fixtures" + assert span.attributes["db.query.text"] == "select 1" + assert span.attributes["datasette.rows_returned"] == 1 + assert span.attributes["datasette.truncated"] is False + assert isinstance(span.attributes["datasette.time_limit_ms"], int) + assert span.status.status_code == StatusCode.UNSET + + +@pytest.mark.asyncio +async def test_facetable_request_produces_db_query_spans(ds_client, otel_spans): + response = await ds_client.get("/fixtures/facetable.json") + assert response.status_code == 200 + + spans = _db_query_spans(otel_spans) + assert spans, "expected at least one db.query span" + assert all(span.attributes["db.system"] == "sqlite" for span in spans) + assert all(span.attributes["db.query.text"] for span in spans) + # Rendering the page also queries the internal database, so only some of + # these spans belong to "fixtures". + assert any(span.attributes["db.namespace"] == "fixtures" for span in spans) + + +def test_sql_attribute_truncates_at_2048(): + short_sql = "select 1" + assert sql_attribute(short_sql) == "select 1" + # Whitespace is stripped, so the same query logged twice with different + # surrounding whitespace produces one attribute value, not two. + assert sql_attribute(" select 1\n") == "select 1" + + long_sql = "select 1 -- " + ("x" * 3000) + truncated = sql_attribute(long_sql) + assert len(truncated) == MAX_SQL_LENGTH + len("…[truncated]") + assert truncated.startswith("select 1 -- ") + assert truncated.endswith("…[truncated]") + + +@pytest.mark.asyncio +async def test_db_query_text_is_truncated_in_real_span(ds_client, otel_spans): + # A long trailing SQL comment keeps the query valid and executable while + # pushing db.query.text well past the 2048 char cap. + long_sql = "select 1 -- " + ("x" * 3000) + response = await ds_client.get("/fixtures/-/query.json", params={"sql": long_sql}) + assert response.status_code == 200 + + spans = _db_query_spans(otel_spans) + assert spans + assert any(len(span.attributes["db.query.text"]) > 100 for span in spans), ( + "expected the long query to reach a span - otherwise this test would " + "pass even if truncation were never applied" + ) + for span in spans: + recorded = span.attributes["db.query.text"] + assert len(recorded) <= MAX_SQL_LENGTH + len("…[truncated]") + + +@pytest.mark.asyncio +async def test_no_span_attribute_ever_contains_a_parameter_value(ds_client, otel_spans): + response = await ds_client.get( + "/fixtures/-/query.json", + params={"sql": "select :secret", "secret": SECRET_PARAM_VALUE}, + ) + assert response.status_code == 200 + # Sanity check the value really did flow through as a bound parameter, + # not inlined into the SQL text, otherwise this test would be vacuous. + assert SECRET_PARAM_VALUE in json.dumps(response.json()) + + for value in _all_attribute_values(otel_spans): + if isinstance(value, str): + assert SECRET_PARAM_VALUE not in value + elif isinstance(value, (list, tuple)): + for item in value: + if isinstance(item, str): + assert SECRET_PARAM_VALUE not in item + + spans = _db_query_spans(otel_spans) + assert spans + span = spans[-1] + assert "select :secret" in span.attributes["db.query.text"] + assert span.attributes.get("datasette.param_count") == 1 + + +@pytest.mark.asyncio +async def test_query_interrupted_sets_error_status(ds_client, otel_spans): + response = await ds_client.get( + "/fixtures/-/query.json", + params={"sql": "select sleep(0.05)", "_timelimit": 5}, + ) + assert response.status_code == 400 + + spans = _db_query_spans(otel_spans) + assert spans + span = spans[-1] + assert span.status.status_code == StatusCode.ERROR + assert span.attributes["datasette.interrupted"] is True + assert span.events + assert all(event.name == "exception" for event in span.events) + + +@pytest.mark.asyncio +async def test_unsuppressed_sql_error_is_a_span_error(ds_client, otel_spans): + db = ds_client.ds.get_database("fixtures") + with pytest.raises(sqlite3.OperationalError): + await db.execute(INVALID_SQL) + + spans = _db_query_spans(otel_spans) + assert spans + span = spans[-1] + assert span.status.status_code == StatusCode.ERROR + assert any(event.name == "exception" for event in span.events) + assert "datasette.sql_error_suppressed" not in span.attributes + + +@pytest.mark.asyncio +async def test_suppressed_sql_error_is_not_a_span_error(ds_client, otel_spans): + """ + log_sql_errors=False means the caller is probing and expects failures. + + Facet suggestion runs `json_type(column)` against every column precisely + to discover which ones raise, so marking those spans as errors would put + two red spans per text column on every table page - burying real failures + and tripping any alerting keyed on span status. + """ + db = ds_client.ds.get_database("fixtures") + with pytest.raises(sqlite3.OperationalError): + await db.execute(INVALID_SQL, log_sql_errors=False) + + spans = _db_query_spans(otel_spans) + assert spans + span = spans[-1] + assert span.status.status_code == StatusCode.UNSET + assert span.attributes["datasette.sql_error_suppressed"] is True + assert not [event for event in span.events if event.name == "exception"] From 59bfa495cc9b31f1ba5920aa50f8916c6ef6d53a Mon Sep 17 00:00:00 2001 From: Alex Garcia Date: Thu, 30 Jul 2026 17:52:55 -0700 Subject: [PATCH 03/10] Emit db.query spans from the three write entry points execute_write(), execute_write_script() and execute_write_many() were the only Database methods that ran SQL without producing an OpenTelemetry span, so any instance doing writes - which is every instance, since Datasette builds its internal catalog through these methods at startup - showed reads in a trace and nothing else. The same db.system, db.namespace and db.query.text attributes the read path already sets now appear here, with db.query.text going through sql_attribute() so attacker-supplied SQL cannot put an unbounded string on a span. execute_write_many() records the parameter-set count as datasette.param_sets, not datasette.rows_returned. executemany() consumes parameter sets and returns no rows at all, so a rows_returned name would be describing something that does not exist - and a consumer building a "rows written" dashboard on top of it would be charting the wrong number. These spans only cover the event-loop side of a write. The time actually spent waiting on the write queue and executing on the write thread is not attributed yet; that needs context propagation across the thread boundary and lands separately. Writes with block=False are worse still - execute_write_fn returns before the write happens, so the span closes early. Span links fix that later. As with the read path, the existing `with trace(...)` wrappers stay put and the new spans nest inside them, so ?_trace=1 keeps working unchanged - including execute_write_many's `count`, which the old tracer stashes through the context manager's return value. Co-Authored-By: Claude Opus 5 --- datasette/database.py | 50 +++++++++++++++++++++------ tests/test_telemetry.py | 75 +++++++++++++++++++++++++++++++++++++++++ 2 files changed, 114 insertions(+), 11 deletions(-) diff --git a/datasette/database.py b/datasette/database.py index f78eb925..cc35c038 100644 --- a/datasette/database.py +++ b/datasette/database.py @@ -259,10 +259,21 @@ class Database: cursor, return_all=return_all, returning_limit=returning_limit ) - with trace("sql", database=self.name, sql=sql.strip(), params=params): - results = await self.execute_write_fn( - _inner, block=block, request=request, transaction=transaction - ) + # SIM117 wants these two context managers merged. They are kept nested + # deliberately: the hand-rolled tracer's wrapper is on its way out, and + # nesting makes removing it a single-line deletion. + with trace( # noqa: SIM117 + "sql", database=self.name, sql=sql.strip(), params=params + ): + with tracer.start_as_current_span("db.query") as span: + span.set_attribute("db.system", "sqlite") + span.set_attribute("db.namespace", self.name) + span.set_attribute("db.query.text", sql_attribute(sql)) + if params: + span.set_attribute("datasette.param_count", len(params)) + results = await self.execute_write_fn( + _inner, block=block, request=request, transaction=transaction + ) return results async def execute_write_script(self, sql, block=True, request=None): @@ -271,10 +282,18 @@ class Database: def _inner(conn): return conn.executescript(sql) - with trace("sql", database=self.name, sql=sql.strip(), executescript=True): - results = await self.execute_write_fn( - _inner, block=block, transaction=False, request=request - ) + # Nested on purpose - see the note in execute_write(). + with trace( # noqa: SIM117 + "sql", database=self.name, sql=sql.strip(), executescript=True + ): + with tracer.start_as_current_span("db.query") as span: + span.set_attribute("db.system", "sqlite") + span.set_attribute("db.namespace", self.name) + span.set_attribute("db.query.text", sql_attribute(sql)) + span.set_attribute("datasette.executescript", True) + results = await self.execute_write_fn( + _inner, block=block, transaction=False, request=request + ) return results async def execute_write_many(self, sql, params_seq, block=True, request=None): @@ -291,12 +310,21 @@ class Database: return conn.executemany(sql, count_params(params_seq)), count + # Nested on purpose - see the note in execute_write(). with trace( "sql", database=self.name, sql=sql.strip(), executemany=True ) as kwargs: - results, count = await self.execute_write_fn( - _inner, block=block, request=request - ) + with tracer.start_as_current_span("db.query") as span: + span.set_attribute("db.system", "sqlite") + span.set_attribute("db.namespace", self.name) + span.set_attribute("db.query.text", sql_attribute(sql)) + span.set_attribute("datasette.executemany", True) + results, count = await self.execute_write_fn( + _inner, block=block, request=request + ) + # count is the number of parameter *sets* consumed by + # executemany(), not a row count - executemany returns no rows. + span.set_attribute("datasette.param_sets", count) kwargs["count"] = count return results diff --git a/tests/test_telemetry.py b/tests/test_telemetry.py index a13898ff..d8e89446 100644 --- a/tests/test_telemetry.py +++ b/tests/test_telemetry.py @@ -6,6 +6,7 @@ import sys import pytest from opentelemetry.trace import StatusCode +from datasette.app import Datasette from datasette.telemetry import MAX_SQL_LENGTH, sql_attribute SECRET_PARAM_VALUE = "SUPER_SECRET_PARAM_VALUE_XYZ_123" @@ -17,6 +18,21 @@ def _db_query_spans(otel_spans): return [span for span in otel_spans.get_finished_spans() if span.name == "db.query"] +def _spans_for_namespace(otel_spans, namespace): + """ + db.query spans belonging to one database. + + Datasette queries its internal catalog constantly - including while a + Datasette instance is being constructed - so a test that just grabbed + every db.query span would be reading someone else's traffic. + """ + return [ + span + for span in _db_query_spans(otel_spans) + if span.attributes["db.namespace"] == namespace + ] + + def _all_attribute_values(otel_spans): "Every attribute value across every finished span, for the 'no leaked param values' test." values = [] @@ -192,3 +208,62 @@ async def test_suppressed_sql_error_is_not_a_span_error(ds_client, otel_spans): assert span.status.status_code == StatusCode.UNSET assert span.attributes["datasette.sql_error_suppressed"] is True assert not [event for event in span.events if event.name == "exception"] + + +@pytest.mark.asyncio +async def test_execute_write_produces_db_query_span(otel_spans): + # Named in-memory databases are shared-cache, so every test in this file + # needs its own name or the second `create table` hits an existing table. + db = Datasette(memory=True).add_memory_database("t03_write_span") + await db.execute_write("create table docs (id integer primary key, name text)") + await db.execute_write("insert into docs (id, name) values (?, ?)", [1, "one"]) + + spans = _spans_for_namespace(otel_spans, "t03_write_span") + assert spans, "expected db.query spans from execute_write()" + span = spans[-1] + + assert span.attributes["db.system"] == "sqlite" + assert span.attributes["db.namespace"] == "t03_write_span" + assert span.attributes["db.query.text"] == ( + "insert into docs (id, name) values (?, ?)" + ) + assert span.attributes["datasette.param_count"] == 2 + + +@pytest.mark.asyncio +async def test_execute_write_script_sets_executescript_attribute(otel_spans): + db = Datasette(memory=True).add_memory_database("t03_write_script_span") + await db.execute_write_script( + "create table docs (id integer primary key);\n" + "insert into docs (id) values (1);" + ) + + spans = _spans_for_namespace(otel_spans, "t03_write_script_span") + assert spans, "expected a db.query span from execute_write_script()" + span = spans[-1] + + assert span.attributes["db.system"] == "sqlite" + assert span.attributes["datasette.executescript"] is True + assert "insert into docs" in span.attributes["db.query.text"] + + +@pytest.mark.asyncio +async def test_execute_write_many_records_param_sets_not_rows_returned(otel_spans): + db = Datasette(memory=True).add_memory_database("t03_write_many_span") + await db.execute_write("create table docs (id integer primary key)") + await db.execute_write_many( + "insert into docs (id) values (?)", [[i] for i in range(1, 6)] + ) + + spans = _spans_for_namespace(otel_spans, "t03_write_many_span") + many_spans = [ + span for span in spans if span.attributes.get("datasette.executemany") is True + ] + assert len(many_spans) == 1 + span = many_spans[0] + + assert span.attributes["datasette.param_sets"] == 5 + # executemany() consumes parameter sets and returns no rows at all, so + # calling this a row count would be a lie. Asserted explicitly because the + # attribute really was named datasette.rows_returned at one point. + assert "datasette.rows_returned" not in span.attributes From 582d79a148f65e4b6e1295719dbbe3886536612d Mon Sep 17 00:00:00 2001 From: Alex Garcia Date: Thu, 30 Jul 2026 18:05:34 -0700 Subject: [PATCH 04/10] Propagate otel context across the thread boundaries Spans created on a worker thread resolve their parent from that thread's ambient context, so without this every span produced below Database came back as an unparented root, disconnected from the request that caused it. Carrying the caller's context across each boundary is also what makes the thread-pool wait visible: db.query covers the full round trip, the new db.query.execute covers only the work inside the worker, and the gap between them is the queueing the old tracer folds invisibly into one number. - execute_fn()'s executor.submit() and execute_isolated_fn()'s run_in_executor() (immutable databases) now run the callable inside a contextvars.copy_context(). A *fresh* copy per submit is required: concurrently entering one shared Context raises "RuntimeError: cannot enter context ... already entered". - WriteTask carries the otel Context captured on the event loop at enqueue time plus an enqueued_at_ns timestamp (both need __slots__ entries, or they fail with AttributeError at runtime). _execute_writes attaches that context right after the _SHUTDOWN check and detaches it in a finally spanning all three execution branches - the write thread is persistent and shared, so a leaked token would grow its context stack for every write processed afterwards, and a wrong-token detach only logs rather than raising. - New spans: db.query.execute (read worker thread), db.write.queue_wait (explicit start/end timestamps, so its duration is the real enqueue -> dequeue wait rather than the microseconds spent building the span) and db.write.execute (skipped in the conn_exception branch, where fn never runs). db.query.execute honours log_sql_errors for the same reason db.query does: facet suggestion probes with log_sql_errors=False and would otherwise paint two red spans per text column on every table page. - The write-thread warm-up prepare_connection is left as a documented orphan root - no caller context exists that early. Tests assert actual parent/child span-id relationships in a shared trace, not just that spans exist, since an unparented root looks identical to a correct span if you only check presence. Note that copy_context() copies every ContextVar, not just OTel's, so Datasette's own context vars (_skip_permission_checks, _permission_check_cache, _in_datasette_client) now flow into worker threads where they previously did not. Co-Authored-By: Claude Opus 5 --- datasette/database.py | 228 ++++++++++++++++++++++--------- tests/test_internals_database.py | 109 +++++++++++++++ tests/test_telemetry.py | 192 +++++++++++++++++++++++++- 3 files changed, 463 insertions(+), 66 deletions(-) diff --git a/datasette/database.py b/datasette/database.py index cc35c038..d7cbc99b 100644 --- a/datasette/database.py +++ b/datasette/database.py @@ -1,16 +1,19 @@ import asyncio import atexit +import contextvars import inspect import os import queue import sys import tempfile import threading +import time import uuid from collections import namedtuple from pathlib import Path import sqlite_utils +from opentelemetry import context as otel_context_api from opentelemetry.trace import Status, StatusCode from .inspect import inspect_hash @@ -351,9 +354,15 @@ class Database: return _run() if not write: # Immutable database - no writes can ever occur, so there is no - # write queue to block; run against a fresh read-only connection + # write queue to block; run against a fresh read-only connection. + # A fresh copy_context() is required per submit (not one shared + # copy reused across calls): concurrent execution of the same + # Context raises "RuntimeError: cannot enter context ... already + # entered". This propagates the caller's otel context (e.g. the + # enclosing db.query span) onto the worker thread. + ctx = contextvars.copy_context() return await asyncio.get_running_loop().run_in_executor( - self.ds.executor, _run + self.ds.executor, ctx.run, _run ) # Threaded mode - send to write thread return await self._send_to_write_thread(fn, isolated_connection=True) @@ -458,8 +467,21 @@ class Database: task_id = uuid.uuid5(uuid.NAMESPACE_DNS, "datasette.io") loop = asyncio.get_running_loop() reply_future = loop.create_future() + # Captured here, on the event loop, at enqueue time: the otel + # Context (carrying the enclosing db.query span, if any) and the + # timestamp used to build the db.write.queue_wait span once this + # task is dequeued on the write thread. self._write_queue.put( - WriteTask(fn, task_id, loop, reply_future, isolated_connection, transaction) + WriteTask( + fn, + task_id, + loop, + reply_future, + isolated_connection, + transaction, + otel_context_api.get_current(), + time.time_ns(), + ) ) if block: return await reply_future @@ -473,6 +495,9 @@ class Database: conn = None try: conn = self.connect(write=True) + # This warm-up runs before any write has ever been queued, so + # there is no caller otel context yet to attach - any spans + # created by plugin hooks here are orphans (roots). self.ds._prepare_connection(conn, self.name) except Exception as e: # noqa: BLE001 # Stored and re-raised to whoever queues the next write @@ -487,40 +512,79 @@ class Database: # Best-effort close as the write thread exits pass return - exception = None - result = None - if conn_exception is not None: - exception = conn_exception - elif task.isolated_connection: - try: - isolated_connection = self.connect(write=True) + # Restore the caller's otel context (captured on the event loop + # at enqueue time) so spans created while processing this task + # parent correctly to the request that queued it. Must be + # detached below in `finally` - a leaked token silently poisons + # this thread's ambient context for every write processed after + # it, and a *wrong*-token detach only logs a warning rather than + # raising, so this pairing is load-bearing and easy to get wrong + # silently. + token = otel_context_api.attach(task.otel_context) + try: + exception = None + result = None + # Explicit start_time/end_time rather than a `with` block: + # this span's duration is the time the task actually spent + # waiting in the queue (enqueue -> dequeue), not the near- + # zero time spent constructing/ending the span object here. + tracer.start_span( + "db.write.queue_wait", start_time=task.enqueued_at_ns + ).end(end_time=time.time_ns()) + if conn_exception is not None: + # fn never runs in this branch, so there is nothing to + # wrap in a db.write.execute span. + exception = conn_exception + elif task.isolated_connection: try: - result = task.fn(isolated_connection) - finally: - isolated_connection.close() - try: - self._all_file_connections.remove(isolated_connection) - except ValueError: - # Was probably a memory connection - pass - except Exception as e: # noqa: BLE001 - # Write thread must survive any task failure or the database wedges - sys.stderr.write(f"{e}\n") - sys.stderr.flush() - exception = e - else: - try: - if task.transaction: - with conn: - conn.execute("BEGIN IMMEDIATE") - result = task.fn(conn) - else: - result = task.fn(conn) - except Exception as e: # noqa: BLE001 - sys.stderr.write(f"{e}\n") - sys.stderr.flush() - exception = e - _deliver_write_result(task, result, exception) + with tracer.start_as_current_span("db.write.execute") as span: + span.set_attribute( + "datasette.isolated_connection", + task.isolated_connection, + ) + span.set_attribute( + "datasette.transaction", task.transaction + ) + isolated_connection = self.connect(write=True) + try: + result = task.fn(isolated_connection) + finally: + isolated_connection.close() + try: + self._all_file_connections.remove( + isolated_connection + ) + except ValueError: + # Was probably a memory connection + pass + except Exception as e: # noqa: BLE001 + # Write thread must survive any task failure or the database wedges + sys.stderr.write(f"{e}\n") + sys.stderr.flush() + exception = e + else: + try: + with tracer.start_as_current_span("db.write.execute") as span: + span.set_attribute( + "datasette.isolated_connection", + task.isolated_connection, + ) + span.set_attribute( + "datasette.transaction", task.transaction + ) + if task.transaction: + with conn: + conn.execute("BEGIN IMMEDIATE") + result = task.fn(conn) + else: + result = task.fn(conn) + except Exception as e: # noqa: BLE001 + sys.stderr.write(f"{e}\n") + sys.stderr.flush() + exception = e + _deliver_write_result(task, result, exception) + finally: + otel_context_api.detach(token) async def execute_fn(self, fn): self._check_not_closed() @@ -542,7 +606,13 @@ class Database: with self._pending_execute_futures_lock: self._check_not_closed() - future = self.ds.executor.submit(in_thread) + # A fresh copy_context() is required per submit (not one shared + # copy reused across calls): concurrent execution of the same + # Context raises "RuntimeError: cannot enter context ... + # already entered". This propagates the caller's otel context + # (e.g. the enclosing db.query span) onto the worker thread. + ctx = contextvars.copy_context() + future = self.ds.executor.submit(ctx.run, in_thread) self._pending_execute_futures.add(future) future.add_done_callback(self._remove_pending_execute_future) return await asyncio.wrap_future(future) @@ -564,35 +634,51 @@ class Database: time_limit_ms = custom_time_limit def sql_operation_in_thread(conn): - with sqlite_timelimit(conn, time_limit_ms): - try: - cursor = conn.cursor() - cursor.execute(sql, params if params is not None else {}) - max_returned_rows = self.ds.max_returned_rows - if max_returned_rows == page_size: - max_returned_rows += 1 - if max_returned_rows and truncate: - rows = cursor.fetchmany(max_returned_rows + 1) - truncated = len(rows) > max_returned_rows - rows = rows[:max_returned_rows] - else: - rows = cursor.fetchall() - truncated = False - except (sqlite3.OperationalError, sqlite3.DatabaseError) as e: - if e.args == ("interrupted",): - raise QueryInterrupted(e, sql, params) - if log_sql_errors: - sys.stderr.write( - f"ERROR: conn={conn}, sql = {sql!r}, params = {params}: {e}\n" - ) - sys.stderr.flush() - raise + # This span is created inside the worker thread. Its parent is + # resolved from the ambient otel context, which was propagated + # onto this thread via copy_context() at the executor.submit() + # boundary in execute_fn() (or run_in_executor() for immutable + # databases) - so it parents correctly to the enclosing + # db.query span despite running on a different thread. + # + # Callers passing log_sql_errors=False are probing and treat a + # failure as an expected answer - see the matching handling on the + # db.query span in execute(). Without this, facet suggestion marks + # two spans per text column as failed on every table page. + with tracer.start_as_current_span( + "db.query.execute", + record_exception=log_sql_errors, + set_status_on_exception=log_sql_errors, + ): + with sqlite_timelimit(conn, time_limit_ms): + try: + cursor = conn.cursor() + cursor.execute(sql, params if params is not None else {}) + max_returned_rows = self.ds.max_returned_rows + if max_returned_rows == page_size: + max_returned_rows += 1 + if max_returned_rows and truncate: + rows = cursor.fetchmany(max_returned_rows + 1) + truncated = len(rows) > max_returned_rows + rows = rows[:max_returned_rows] + else: + rows = cursor.fetchall() + truncated = False + except (sqlite3.OperationalError, sqlite3.DatabaseError) as e: + if e.args == ("interrupted",): + raise QueryInterrupted(e, sql, params) + if log_sql_errors: + sys.stderr.write( + f"ERROR: conn={conn}, sql = {sql!r}, params = {params}: {e}\n" + ) + sys.stderr.flush() + raise - if truncate: - return Results(rows, truncated, cursor.description) + if truncate: + return Results(rows, truncated, cursor.description) - else: - return Results(rows, False, cursor.description) + else: + return Results(rows, False, cursor.description) # SIM117 wants these two context managers merged. They are kept nested # deliberately: the hand-rolled tracer's wrapper is on its way out, and @@ -924,16 +1010,26 @@ def _apply_write_wrapper(fn, wrapper_factory, track_event): class WriteTask: __slots__ = ( + "enqueued_at_ns", "fn", "isolated_connection", "loop", + "otel_context", "reply_future", "task_id", "transaction", ) def __init__( - self, fn, task_id, loop, reply_future, isolated_connection, transaction + self, + fn, + task_id, + loop, + reply_future, + isolated_connection, + transaction, + otel_context, + enqueued_at_ns, ): self.fn = fn self.task_id = task_id @@ -941,6 +1037,8 @@ class WriteTask: self.reply_future = reply_future self.isolated_connection = isolated_connection self.transaction = transaction + self.otel_context = otel_context + self.enqueued_at_ns = enqueued_at_ns def _deliver_write_result(task, result, exception): diff --git a/tests/test_internals_database.py b/tests/test_internals_database.py index b1093b1c..ce0d7a53 100644 --- a/tests/test_internals_database.py +++ b/tests/test_internals_database.py @@ -3,11 +3,13 @@ Tests for the datasette.database.Database class """ import asyncio +import threading import uuid from types import SimpleNamespace import pytest import sqlite_utils +from opentelemetry import context as otel_context_api from datasette.app import Datasette from datasette.database import ( @@ -1223,3 +1225,110 @@ async def test_database_close_is_idempotent(tmpdir): # Second call should be a no-op, not raise db.close() ds._internal_database.close() + + +_CONTEXT_LEAK_MARKER_KEY = "otel-context-leak-marker" + + +@pytest.mark.asyncio +@pytest.mark.parametrize("num_sql_threads", (0, 1)) +async def test_write_thread_context_is_detached_between_tasks( + tmp_path, monkeypatch, num_sql_threads +): + """ + The write thread attaches each task's otel Context and must detach it + again before picking up the next task. The thread is persistent and + shared, so a leaked token would grow that thread's context stack for the + rest of the process - and a *wrong*-token detach only logs a warning + rather than raising, so "does it throw" cannot catch either mistake. + + Two things are asserted, because neither alone is sufficient: + + 1. Each task observes the context value that was current on the event + loop when it was queued. This is what fails if the Context is not + carried on WriteTask, or is never attached. It does *not* catch a + missing detach: attach() replaces the current Context wholesale, so a + leftover one from a previous task is simply overwritten. + 2. The write thread's attach depth is identical at the same point in + every task. This is what fails if detach is missing - the stack grows + by one per task - and it holds across a task that raises, because the + detach lives in a `finally`. + + An otel context value is used rather than a plain contextvars.ContextVar: + a plain var set on the event loop never crosses into the write thread, so + the probe would read None every time and the test could not fail. + """ + name = f"context_leak_test_{num_sql_threads}" + db_path = tmp_path / f"{name}.db" + sqlite3.connect(db_path).close() + ds = Datasette([str(db_path)], settings={"num_sql_threads": num_sql_threads}) + db = ds.get_database(name) + await db.execute_write("create table t (id integer primary key)") + + write_thread_name = f"_execute_writes for database {name}" + depth = {"value": 0} + real_attach = otel_context_api.attach + real_detach = otel_context_api.detach + + def counting_attach(context): + token = real_attach(context) + if threading.current_thread().name == write_thread_name: + depth["value"] += 1 + return token + + def counting_detach(token): + real_detach(token) + if threading.current_thread().name == write_thread_name: + depth["value"] -= 1 + + # Patched on the opentelemetry.context module itself, which is what both + # database.py and opentelemetry.trace.use_span() look the functions up on. + monkeypatch.setattr(otel_context_api, "attach", counting_attach) + monkeypatch.setattr(otel_context_api, "detach", counting_detach) + + seen_markers = [] + seen_depths = [] + + def probe(conn): + seen_markers.append(otel_context_api.get_value(_CONTEXT_LEAK_MARKER_KEY)) + seen_depths.append(depth["value"]) + + def failing_probe(conn): + probe(conn) + # Exercises the write thread's exception path: the detach still has + # to happen, which is why it lives in a `finally`. + raise ValueError("deliberate failure inside a write task") + + try: + for i in range(5): + ctx = otel_context_api.set_value(_CONTEXT_LEAK_MARKER_KEY, f"marker-{i}") + token = real_attach(ctx) + try: + if i == 2: + with pytest.raises(ValueError): + await db.execute_write_fn(failing_probe) + else: + await db.execute_write_fn(probe) + finally: + real_detach(token) + + # Sanity check: no marker is active in *this* (event loop) context + # right now, so the final probe is a fair test of the write thread's + # own state rather than something this test forgot to clean up. + assert otel_context_api.get_value(_CONTEXT_LEAK_MARKER_KEY) is None + await db.execute_write_fn(probe) + finally: + db.close() + + assert seen_markers == [ + "marker-0", + "marker-1", + "marker-2", + "marker-3", + "marker-4", + None, + ] + assert len(set(seen_depths)) == 1, ( + f"write thread context stack grew across tasks: {seen_depths} - " + "a token was attached without being detached" + ) diff --git a/tests/test_telemetry.py b/tests/test_telemetry.py index d8e89446..f232a39d 100644 --- a/tests/test_telemetry.py +++ b/tests/test_telemetry.py @@ -2,12 +2,15 @@ import json import sqlite3 import subprocess import sys +import time import pytest +import sqlite_utils from opentelemetry.trace import StatusCode from datasette.app import Datasette -from datasette.telemetry import MAX_SQL_LENGTH, sql_attribute +from datasette.database import Database +from datasette.telemetry import MAX_SQL_LENGTH, sql_attribute, tracer SECRET_PARAM_VALUE = "SUPER_SECRET_PARAM_VALUE_XYZ_123" @@ -33,6 +36,26 @@ def _spans_for_namespace(otel_spans, namespace): ] +def _children_named(otel_spans, name, parent_span_context): + """ + Finished spans called `name` whose parent really is `parent_span_context`. + + Parentage is matched on span id, not on "a span with this name exists" - + a span can exist and still be an unparented root if a thread boundary + dropped the otel context, which is the exact failure these tests exist + to catch. + """ + return [ + span + for span in otel_spans.get_finished_spans() + if span.name == name + and span.parent is not None + and span.parent.span_id == parent_span_context.span_id + and span.parent.trace_id == parent_span_context.trace_id + and span.context.trace_id == parent_span_context.trace_id + ] + + def _all_attribute_values(otel_spans): "Every attribute value across every finished span, for the 'no leaked param values' test." values = [] @@ -267,3 +290,170 @@ async def test_execute_write_many_records_param_sets_not_rows_returned(otel_span # calling this a row count would be a lie. Asserted explicitly because the # attribute really was named datasette.rows_returned at one point. assert "datasette.rows_returned" not in span.attributes + + +# --- Context propagation across thread boundaries -------------------------- +# +# Every assertion below checks parentage (child.parent.span_id == +# expected_parent.span_id, in the same trace), not merely that spans exist. +# Spans can exist and still be wrongly parented - or be unparented roots - if +# a thread boundary drops the otel context, which is exactly the failure mode +# these tests exist to prevent. + + +@pytest.mark.asyncio +async def test_db_query_execute_parents_to_db_query(ds_client, otel_spans): + # execute_fn()'s executor.submit() is thread boundary #1. The + # db.query.execute span is created inside the worker thread; without the + # copy_context() propagation it comes back as an unparented root span + # rather than a child of db.query. + response = await ds_client.get("/fixtures/-/query.json?sql=select+1") + assert response.status_code == 200 + + query_spans = [ + span + for span in _spans_for_namespace(otel_spans, "fixtures") + if span.attributes["db.query.text"] == "select 1" + ] + assert query_spans, "expected a db.query span for 'select 1'" + query_span = query_spans[-1] + + assert [ + span + for span in otel_spans.get_finished_spans() + if span.name == "db.query.execute" + ], "expected at least one db.query.execute span" + children = _children_named(otel_spans, "db.query.execute", query_span.context) + assert len(children) == 1, "expected exactly one db.query.execute child of db.query" + # The execute span is strictly contained by the round-trip span, and the + # gap between the two is the thread-pool wait. + assert query_span.start_time <= children[0].start_time + assert children[0].end_time <= query_span.end_time + + +@pytest.mark.asyncio +async def test_immutable_database_propagates_context(tmp_path, otel_spans): + # Thread boundary #3, the easy one to miss: immutable databases route + # execute_isolated_fn() through loop.run_in_executor() directly rather + # than through the write thread. A span created inside that worker must + # still parent to whatever was current when execute_isolated_fn() was + # awaited, or every immutable-database operation emits orphan roots. + db_path = tmp_path / "t04_immutable.db" + sqlite_utils.Database(str(db_path))["t"].insert({"id": 1}, pk="id") + + ds = Datasette() + db = Database(ds, path=str(db_path), is_mutable=False) + ds.add_database(db, name="t04_immutable") + + def fn(conn): + with tracer.start_as_current_span("t04-child-in-isolated-worker"): + pass + + try: + with tracer.start_as_current_span("t04-parent-on-event-loop") as parent: + parent_context = parent.get_span_context() + await db.execute_isolated_fn(fn) + finally: + ds.remove_database("t04_immutable") + + assert [ + span + for span in otel_spans.get_finished_spans() + if span.name == "t04-child-in-isolated-worker" + ], "expected a span created inside execute_isolated_fn's worker thread" + children = _children_named( + otel_spans, "t04-child-in-isolated-worker", parent_context + ) + assert len(children) == 1 + + +@pytest.mark.asyncio +async def test_write_spans_parent_to_db_query(otel_spans): + # Thread boundary #2: WriteTask -> queue.Queue -> the write thread. + # db.write.queue_wait and db.write.execute are both direct children of + # the db.query span that was current on the event loop at enqueue time, + # so they are siblings rather than nested inside one another. + db = Datasette(memory=True).add_memory_database("t04_write_spans") + await db.execute_write("create table docs (id integer primary key)") + + query_spans = _spans_for_namespace(otel_spans, "t04_write_spans") + assert query_spans, "expected a db.query span from execute_write()" + query_span = query_spans[-1] + + queue_wait_children = _children_named( + otel_spans, "db.write.queue_wait", query_span.context + ) + execute_children = _children_named( + otel_spans, "db.write.execute", query_span.context + ) + assert len(queue_wait_children) == 1 + assert len(execute_children) == 1 + + execute_span = execute_children[0] + assert execute_span.attributes["datasette.isolated_connection"] is False + assert execute_span.attributes["datasette.transaction"] is True + # Siblings, not parent/child: the queue wait is over by the time the + # write begins. + assert queue_wait_children[0].end_time <= execute_span.start_time + + +@pytest.mark.asyncio +async def test_write_queue_wait_duration_reflects_real_wait(otel_spans): + # db.write.queue_wait is built from explicit start/end timestamps - + # task.enqueued_at_ns, captured on the event loop, through to the moment + # the write thread dequeued it. If it were a plain `with` block on the + # write thread it would instead measure the microseconds spent building + # the span object, and this assertion would fail. + ds = Datasette(memory=True) + db = ds.add_memory_database("t04_queue_wait") + await db.execute_write("create table docs (id integer primary key)") + + def slow_write(conn): + time.sleep(0.1) + + # Queue a deliberately slow write without waiting for it, then queue a + # second write immediately behind it: the second task sits in the queue + # for roughly the duration of the first. + _, slow_future = await db._send_to_write_thread(slow_write, block=False) + await db.execute_write("insert into docs (id) values (1)") + await slow_future + + query_spans = [ + span + for span in _spans_for_namespace(otel_spans, "t04_queue_wait") + if span.attributes["db.query.text"] == "insert into docs (id) values (1)" + ] + assert query_spans, "expected a db.query span for the queued-behind insert" + queue_wait_children = _children_named( + otel_spans, "db.write.queue_wait", query_spans[-1].context + ) + assert len(queue_wait_children) == 1 + duration_ns = queue_wait_children[0].end_time - queue_wait_children[0].start_time + # The slow write sleeps 100ms; anything above 10ms is far beyond the + # microseconds a mis-timestamped span would report. + assert duration_ns > 10_000_000, f"queue wait was only {duration_ns}ns" + + +@pytest.mark.asyncio +async def test_suppressed_error_does_not_mark_execute_span(ds_client, otel_spans): + """ + The inner db.query.execute span must honour log_sql_errors too. + + It is created inside the worker thread, so without record_exception / + set_status_on_exception being passed through it would mark every facet + suggestion probe as failed even though the outer db.query span correctly + reports the failure as suppressed. + """ + db = ds_client.ds.get_database("fixtures") + with pytest.raises(sqlite3.OperationalError): + await db.execute(INVALID_SQL, log_sql_errors=False) + + execute_spans = [ + span + for span in otel_spans.get_finished_spans() + if span.name == "db.query.execute" + ] + assert execute_spans + span = execute_spans[-1] + assert span.status.status_code == StatusCode.UNSET + assert not [event for event in span.events if event.name == "exception"] From 4ebec0b1ea4a9e49cbf062a8882d774cd2fa9a36 Mon Sep 17 00:00:00 2001 From: Alex Garcia Date: Thu, 30 Jul 2026 18:16:46 -0700 Subject: [PATCH 05/10] Give startup's ~20 orphan spans somewhere to belong invoke_startup() runs before any request exists, so nothing it does has an ambient span to nest under. Measured on a fresh instance: 19 distinct traces, 19 of them single- or few-span roots - the register_* hook dispatches, the internal catalog's db.query reads and its db.write.* catalog writes. In a trace UI that is nineteen pieces of noise sitting next to every real trace, which for an operator opening Jaeger for the first time is the difference between "this works" and "this is unusable". Bracketing the whole method body in one datasette.startup span takes that to 1. This is not a propagation fix - ticket 04's context propagation was already correct, it simply had nothing to propagate. The bulk of the app.py diff is re-indentation; `git diff -w` shows the real change (plus one line-length rewrap black applied to the StartupError raise). register_output_renderer and asgi_wrapper stay orphans deliberately: both are dispatched from Datasette.__init__ / .app(), before invoke_startup() exists to be called, and wrapping them would mean holding a span open across object construction in library code that may never serve a request. Suppressing instrumentation during warm-up was rejected as an alternative: a slow prepare_connection runs on every connection, not just at startup, and is exactly what tracing should reveal. Also corrects the stale write-thread warm-up comment in database.py. It is still a root, but for a reason worth stating precisely: a raw threading.Thread does not inherit the starting thread's context, so the datasette.startup span current on the event loop does not reach it. Read connections do warm up under copy_context() and nest correctly. Co-Authored-By: Claude Opus 5 --- datasette/app.py | 109 ++++++++++++++++++++++------------------ datasette/database.py | 11 +++- tests/test_telemetry.py | 86 +++++++++++++++++++++++++++++++ 3 files changed, 156 insertions(+), 50 deletions(-) diff --git a/datasette/app.py b/datasette/app.py index 42be7425..d3626ab2 100644 --- a/datasette/app.py +++ b/datasette/app.py @@ -49,6 +49,7 @@ from .events import Event from .plugins import DEFAULT_PLUGINS, get_plugins, pm from .renderer import json_renderer from .resources import DatabaseResource, TableResource +from .telemetry import tracer from .tokens import TokenInvalid from .tracer import AsgiTracer from .url_builder import Urls @@ -778,57 +779,69 @@ class Datasette: # This must be called for Datasette to be in a usable state if self._startup_invoked: return - # Register event classes - event_classes = [] - for hook in pm.hook.register_events(datasette=self): - extra_classes = await await_me_maybe(hook) - if extra_classes: - event_classes.extend(extra_classes) - self.event_classes = tuple(event_classes) + # invoke_startup() runs before any request exists, so every span its + # children create - the register_* hook dispatches, the internal + # catalog's db.query/db.write spans, and the prepare_connection + # warm-up of the read connections those touch - would otherwise be + # its own orphan root trace: around twenty of them on a fresh + # instance. Bracketing the whole thing gives them somewhere to belong. + # A connection warmed lazily later, by a request touching a new + # database for the first time, nests under that request instead: + # this span has already ended by then. + with tracer.start_as_current_span("datasette.startup"): + # Register event classes + event_classes = [] + for hook in pm.hook.register_events(datasette=self): + extra_classes = await await_me_maybe(hook) + if extra_classes: + event_classes.extend(extra_classes) + self.event_classes = tuple(event_classes) - # Register actions, but watch out for duplicate name/abbr - action_names = {} - action_abbrs = {} - for hook in pm.hook.register_actions(datasette=self): - if hook: - for action in hook: - if ( - action.name in action_names - and action != action_names[action.name] - ): - raise StartupError(f"Duplicate action name: {action.name}") - if ( - action.abbr - and action.abbr in action_abbrs - and action != action_abbrs[action.abbr] - ): - raise StartupError(f"Duplicate action abbr: {action.abbr}") - action_names[action.name] = action - if action.abbr: - action_abbrs[action.abbr] = action - self.actions[action.name] = action + # Register actions, but watch out for duplicate name/abbr + action_names = {} + action_abbrs = {} + for hook in pm.hook.register_actions(datasette=self): + if hook: + for action in hook: + if ( + action.name in action_names + and action != action_names[action.name] + ): + raise StartupError(f"Duplicate action name: {action.name}") + if ( + action.abbr + and action.abbr in action_abbrs + and action != action_abbrs[action.abbr] + ): + raise StartupError(f"Duplicate action abbr: {action.abbr}") + action_names[action.name] = action + if action.abbr: + action_abbrs[action.abbr] = action + self.actions[action.name] = action - # Register column types (classes, not instances) - self._column_types = {} - for hook in pm.hook.register_column_types(datasette=self): - if hook: - for ct_cls in hook: - if ct_cls.name in self._column_types: - raise StartupError(f"Duplicate column type name: {ct_cls.name}") - self._column_types[ct_cls.name] = ct_cls + # Register column types (classes, not instances) + self._column_types = {} + for hook in pm.hook.register_column_types(datasette=self): + if hook: + for ct_cls in hook: + if ct_cls.name in self._column_types: + raise StartupError( + f"Duplicate column type name: {ct_cls.name}" + ) + self._column_types[ct_cls.name] = ct_cls - for hook in pm.hook.prepare_jinja2_environment( - env=self._jinja_env, datasette=self - ): - await await_me_maybe(hook) - # Ensure internal tables and metadata are populated before startup hooks - await self._refresh_schemas() - await self._save_queries_from_config() - # Load column_types from config into internal DB - await self._apply_column_types_config() - for hook in pm.hook.startup(datasette=self): - await await_me_maybe(hook) - self._startup_invoked = True + for hook in pm.hook.prepare_jinja2_environment( + env=self._jinja_env, datasette=self + ): + await await_me_maybe(hook) + # Ensure internal tables and metadata are populated before startup hooks + await self._refresh_schemas() + await self._save_queries_from_config() + # Load column_types from config into internal DB + await self._apply_column_types_config() + for hook in pm.hook.startup(datasette=self): + await await_me_maybe(hook) + self._startup_invoked = True def sign(self, value, namespace="default"): return URLSafeSerializer(self._secret, namespace).dumps(value) diff --git a/datasette/database.py b/datasette/database.py index d7cbc99b..38b38f48 100644 --- a/datasette/database.py +++ b/datasette/database.py @@ -496,8 +496,15 @@ class Database: try: conn = self.connect(write=True) # This warm-up runs before any write has ever been queued, so - # there is no caller otel context yet to attach - any spans - # created by plugin hooks here are orphans (roots). + # there is no captured caller context to attach - and a raw + # threading.Thread does not inherit the context of whoever started + # it. Spans created by plugin hooks here are therefore roots even + # when the write thread is started from inside invoke_startup(): + # its datasette.startup span is current on the event loop but does + # not cross this thread boundary. Read connections differ - they + # warm up inside executor tasks submitted with copy_context(), so + # their prepare_connection spans do nest under whoever triggered + # them. self.ds._prepare_connection(conn, self.name) except Exception as e: # noqa: BLE001 # Stored and re-raised to whoever queues the next write diff --git a/tests/test_telemetry.py b/tests/test_telemetry.py index f232a39d..7894670c 100644 --- a/tests/test_telemetry.py +++ b/tests/test_telemetry.py @@ -6,6 +6,7 @@ import time import pytest import sqlite_utils +from opentelemetry import trace as otel_trace from opentelemetry.trace import StatusCode from datasette.app import Datasette @@ -56,6 +57,27 @@ def _children_named(otel_spans, name, parent_span_context): ] +def _descends_from(span, ancestor_span_context, by_span_id): + """ + True if `span` reaches `ancestor_span_context` by walking parent links. + + Walks real span ids rather than trusting a shared trace id: a span can + carry the right trace id and still hang off the wrong parent. + """ + seen = set() + current = span + while current.parent is not None: + if current.parent.span_id == ancestor_span_context.span_id: + return current.parent.trace_id == ancestor_span_context.trace_id + if current.parent.span_id in seen: + return False + seen.add(current.parent.span_id) + current = by_span_id.get(current.parent.span_id) + if current is None: + return False + return False + + def _all_attribute_values(otel_spans): "Every attribute value across every finished span, for the 'no leaked param values' test." values = [] @@ -457,3 +479,67 @@ async def test_suppressed_error_does_not_mark_execute_span(ds_client, otel_spans span = execute_spans[-1] assert span.status.status_code == StatusCode.UNSET assert not [event for event in span.events if event.name == "exception"] + + +@pytest.mark.asyncio +async def test_invoke_startup_produces_one_trace_not_dozens_of_orphans(otel_spans): + """ + invoke_startup() runs with no request, so nothing it does has an ambient + span to nest under. Without datasette.startup every register_* hook, every + internal-catalog read and every catalog write becomes its own single-span + root trace - around twenty of them per fresh instance. + """ + ds = Datasette(memory=True) + # Named in-memory databases are shared-cache, so this needs its own name. + ds.add_memory_database("t05_startup_db") + # Constructing a Datasette already touches the internal catalog, and that + # work is genuinely outside startup. Clear so the assertions below describe + # invoke_startup() alone. + otel_spans.clear() + + # Deliberately no ambient span: this mirrors the ASGI lifespan path, where + # startup runs before any request exists. If something did wrap this call + # the "one root" assertion below would pass for the wrong reason. + assert ( + not otel_trace.get_current_span().get_span_context().is_valid + ), "this test must run with no ambient span" + + await ds.invoke_startup() + + spans = otel_spans.get_finished_spans() + assert len(spans) > 10, f"expected startup to emit many spans, got {len(spans)}" + + startup_spans = [span for span in spans if span.name == "datasette.startup"] + assert len(startup_spans) == 1 + startup = startup_spans[0] + assert startup.parent is None, "datasette.startup should be a root span" + + trace_ids = {span.context.trace_id for span in spans} + assert trace_ids == {startup.context.trace_id}, ( + f"startup produced {len(trace_ids)} distinct traces; every span it " + "causes should share the datasette.startup trace" + ) + + roots = [span for span in spans if span.parent is None] + assert [span.name for span in roots] == ["datasette.startup"] + + by_span_id = {span.context.span_id: span for span in spans} + + # The internal catalog reads are what made up the bulk of the orphans. + internal_queries = [ + span + for span in spans + if span.name == "db.query" and span.attributes["db.namespace"] == "__INTERNAL__" + ] + assert internal_queries, "expected internal-catalog db.query spans during startup" + assert all( + _descends_from(span, startup.context, by_span_id) for span in internal_queries + ) + + # ...and the catalog writes, which reach the span through the write thread, + # so they also prove the ticket-04 context capture survives startup. + write_spans = [span for span in spans if span.name.startswith("db.write.")] + assert write_spans, "expected db.write.* spans during startup" + assert all( + _descends_from(span, startup.context, by_span_id) for span in write_spans + ) From 77b025be281b1215c51c2f62c75fb3d6b1c7092d Mon Sep 17 00:00:00 2001 From: Alex Garcia Date: Thu, 30 Jul 2026 18:29:16 -0700 Subject: [PATCH 06/10] Make db.query spans match OpenTelemetry semantic conventions Three corrections to the emitted data, bundled because changing what is on the wire after operators have built dashboards on it is a breaking change - so they belong in the first release that ships spans at all, not a later one. db.query is now SpanKind.CLIENT. Trace UIs key their database rendering off the span kind rather than off db.system, so the spans rendered as ordinary internal work despite carrying db.system and db.query.text. The three child spans stay INTERNAL on purpose: db.query.execute, db.write.execute and db.write.queue_wait are Datasette's decomposition of one logical query, not three database calls, and queue_wait touches no database at all - marking them CLIENT would make one query look like several to anything counting spans by kind. The instrumentation scope now carries the Datasette version and a schema URL, so a backend can tell which Datasette produced a span. The URL is 1.29.0 rather than the latest semconv release because that is the highest version at which every name emitted here is the current spelling: db.system was renamed to db.system.name in 1.30.0 and this code still emits the older form. Claiming a later schema would be false, and would stop a consumer translating that name forward, since the claim asserts the rename already happened. db.operation.name is the statement's leading keyword matched against a fixed allowlist, not a parse. On a public instance the SQL is attacker-controlled and this attribute is a candidate metric dimension in a later phase, so echoing back an arbitrary first token would let a visitor's typo mint a permanent series. Anything unrecognised gets no attribute rather than a wrong one. execute_write_script() does not set it at all, since semantic conventions say not to extract an operation name from query text that can hold several statements. db.collection.name comes only from a new table= argument on Database.execute(), and is never derived from the SQL: deriving it would be a parse, and on an instance where anyone can create a table the value set has no ceiling. It is passed from every query in the table and row views that targets exactly one user table. Internal-catalog reads and the row view's cross-table foreign key counts are deliberately left without it. Co-Authored-By: Claude Opus 5 --- datasette/database.py | 38 +++++-- datasette/telemetry.py | 89 ++++++++++++++++- datasette/views/row.py | 9 +- datasette/views/table.py | 16 ++- tests/test_telemetry.py | 209 ++++++++++++++++++++++++++++++++++++++- 5 files changed, 346 insertions(+), 15 deletions(-) diff --git a/datasette/database.py b/datasette/database.py index 38b38f48..1b9a5076 100644 --- a/datasette/database.py +++ b/datasette/database.py @@ -14,10 +14,10 @@ from pathlib import Path import sqlite_utils from opentelemetry import context as otel_context_api -from opentelemetry.trace import Status, StatusCode +from opentelemetry.trace import SpanKind, Status, StatusCode from .inspect import inspect_hash -from .telemetry import sql_attribute, tracer +from .telemetry import sql_attribute, sql_operation_name, tracer from .tracer import trace from .utils import ( call_with_supported_arguments, @@ -268,10 +268,13 @@ class Database: with trace( # noqa: SIM117 "sql", database=self.name, sql=sql.strip(), params=params ): - with tracer.start_as_current_span("db.query") as span: + with tracer.start_as_current_span("db.query", kind=SpanKind.CLIENT) as span: span.set_attribute("db.system", "sqlite") span.set_attribute("db.namespace", self.name) span.set_attribute("db.query.text", sql_attribute(sql)) + operation_name = sql_operation_name(sql) + if operation_name: + span.set_attribute("db.operation.name", operation_name) if params: span.set_attribute("datasette.param_count", len(params)) results = await self.execute_write_fn( @@ -289,7 +292,11 @@ class Database: with trace( # noqa: SIM117 "sql", database=self.name, sql=sql.strip(), executescript=True ): - with tracer.start_as_current_span("db.query") as span: + # No db.operation.name here, deliberately: executescript() runs + # several semicolon-separated statements, and semantic conventions + # say the attribute should not be extracted from query text that + # can hold more than one operation - see sql_operation_name(). + with tracer.start_as_current_span("db.query", kind=SpanKind.CLIENT) as span: span.set_attribute("db.system", "sqlite") span.set_attribute("db.namespace", self.name) span.set_attribute("db.query.text", sql_attribute(sql)) @@ -317,11 +324,16 @@ class Database: with trace( "sql", database=self.name, sql=sql.strip(), executemany=True ) as kwargs: - with tracer.start_as_current_span("db.query") as span: + with tracer.start_as_current_span("db.query", kind=SpanKind.CLIENT) as span: span.set_attribute("db.system", "sqlite") span.set_attribute("db.namespace", self.name) span.set_attribute("db.query.text", sql_attribute(sql)) span.set_attribute("datasette.executemany", True) + # A single statement run with many parameter sets, so unlike + # execute_write_script() there is exactly one operation to name. + operation_name = sql_operation_name(sql) + if operation_name: + span.set_attribute("db.operation.name", operation_name) results, count = await self.execute_write_fn( _inner, block=block, request=request ) @@ -632,8 +644,16 @@ class Database: custom_time_limit=None, page_size=None, log_sql_errors=True, + table=None, ): - """Executes sql against db_name in a thread""" + """Executes sql against db_name in a thread + + `table`, if passed, is recorded as the `db.collection.name` span + attribute. It exists for callers that already know which table the + query targets - the table and row views - and is never derived from + `sql` itself: deriving it would be a parse, and on an instance where + anyone can create a table the resulting value set has no ceiling. + """ self._check_not_closed() page_size = page_size or self.ds.page_size time_limit_ms = self.ds.sql_time_limit_ms @@ -698,6 +718,7 @@ class Database: # can be honoured - see the comment on the generic handler below. with tracer.start_as_current_span( "db.query", + kind=SpanKind.CLIENT, record_exception=False, set_status_on_exception=False, ) as span: @@ -705,6 +726,11 @@ class Database: span.set_attribute("db.namespace", self.name) span.set_attribute("db.query.text", sql_attribute(sql)) span.set_attribute("datasette.time_limit_ms", time_limit_ms) + operation_name = sql_operation_name(sql) + if operation_name: + span.set_attribute("db.operation.name", operation_name) + if table: + span.set_attribute("db.collection.name", table) if params: span.set_attribute("datasette.param_count", len(params)) try: diff --git a/datasette/telemetry.py b/datasette/telemetry.py index 6f5ed093..37195584 100644 --- a/datasette/telemetry.py +++ b/datasette/telemetry.py @@ -9,9 +9,34 @@ harness). With no provider installed every span produced here is a `NonRecordingSpan` and costs approximately nothing. """ +import re + from opentelemetry import trace as otel_trace -tracer = otel_trace.get_tracer("datasette") +from .version import __version__ + +# The semantic-convention version whose spellings this instrumentation +# actually emits. Deliberately NOT the latest release. +# +# A schema URL is a machine-readable claim: a consumer doing schema +# translation replays the renames between the declared version and the one +# it wants, so the claim has to name the version whose spellings are on the +# wire. A wrong one makes translation wrong rather than merely uninformative. +# +# Datasette emits `db.system`, which was renamed to `db.system.name` in +# semconv 1.30.0. Everything else it emits (`db.namespace`, `db.query.text`, +# `db.operation.name`, `db.collection.name`) has been current since 1.26.0. +# So 1.29.0 is the highest version at which every name emitted here is the +# current spelling. Everything under `datasette.*` is Datasette's own and +# outside semconv, so it is unaffected either way. +# +# Declaring 1.43.0 would be false about `db.system`, and would actively STOP +# a consumer translating it forward, because it asserts the rename already +# happened. Bump this deliberately, in the same commit as the attribute +# renames it implies - it is a claim about the names, not decoration. +SCHEMA_URL = "https://opentelemetry.io/schemas/1.29.0" + +tracer = otel_trace.get_tracer("datasette", __version__, schema_url=SCHEMA_URL) MAX_SQL_LENGTH = 2048 @@ -22,3 +47,65 @@ def sql_attribute(sql: str) -> str: if len(sql) <= MAX_SQL_LENGTH: return sql return sql[:MAX_SQL_LENGTH] + "…[truncated]" + + +# db.operation.name is the leading keyword of a statement matched against a +# fixed allowlist - deliberately not a parse. +# +# This runs against arbitrary user-supplied SQL (the `?sql=` query string, +# canned queries, anything typed into the query editor), and the attribute is +# a candidate dimension on a query-duration metric in a later phase. A metric +# series is keyed by its attribute values, so echoing back an arbitrary first +# token would let one visitor's typo mint a new, permanent series. The +# allowlist bounds that at a fixed, small set regardless of what anyone sends. +DB_OPERATION_ALLOWLIST = frozenset( + { + "SELECT", + "INSERT", + "UPDATE", + "DELETE", + "CREATE", + "DROP", + "ALTER", + "PRAGMA", + "EXPLAIN", + "REPLACE", + "VACUUM", + "ANALYZE", + "WITH", + } +) + +_LEADING_KEYWORD = re.compile(r"^\s*([A-Za-z]+)") + + +def sql_operation_name(sql: str) -> str | None: + """ + The statement's leading keyword, if it is one we recognise. + + Returns None - never a guess - for anything not on the allowlist, + including a statement that opens with a comment or with punctuation such + as the "(" of a parenthesised SELECT. + + Known limitation: a statement beginning with a CTE reports `WITH` rather + than the operation inside it, and a substantial share of Datasette's own + reads take that form. Extracting more than the leading keyword means + handling comment stripping, parenthesised `(SELECT ...) UNION` and + compound names like `CREATE TABLE` - each a special case a hand-rolled + matcher would accrete and eventually get wrong. Omitting a name beats + guessing at one. + + Only safe to call with a single statement: `execute_write_script()` runs + several separated by semicolons, and semantic conventions say + `db.operation.name` "SHOULD NOT be extracted from db.query.text, when the + database system supports query text with multiple operations in non-batch + operations" - so that call site does not use this at all rather than + reporting only the first statement's operation. + """ + match = _LEADING_KEYWORD.match(sql) + if not match: + return None + keyword = match.group(1).upper() + if keyword in DB_OPERATION_ALLOWLIST: + return keyword + return None diff --git a/datasette/views/row.py b/datasette/views/row.py index b1388299..b5196a83 100644 --- a/datasette/views/row.py +++ b/datasette/views/row.py @@ -407,7 +407,7 @@ class RowView(BaseView): raise Forbidden("You do not have permission to view this table") results = await resolved.db.execute( - resolved.sql, resolved.params, truncate=True + resolved.sql, resolved.params, truncate=True, table=table ) columns = [r[0] for r in results.description] rows = list(results.rows) @@ -652,6 +652,9 @@ class RowView(BaseView): ] ) try: + # No table= here: this counts incoming references across every + # foreign key pointing at this row, so it spans many tables and + # there is no single value db.collection.name could take. rows = list(await db.execute(sql, {"id": pk_values[0]})) except QueryInterrupted: # Almost certainly hit the timeout @@ -840,7 +843,7 @@ class RowUpdateView(BaseView): returned_row = None if data.get("return"): results = await resolved.db.execute( - resolved.sql, resolved.params, truncate=True + resolved.sql, resolved.params, truncate=True, table=resolved.table ) returned_row = results.dicts()[0] result["rows"] = [returned_row] @@ -858,7 +861,7 @@ class RowUpdateView(BaseView): message_row = returned_row if message_row is None: results = await resolved.db.execute( - resolved.sql, resolved.params, truncate=True + resolved.sql, resolved.params, truncate=True, table=resolved.table ) message_row = results.first() self.ds.add_message( diff --git a/datasette/views/table.py b/datasette/views/table.py index 7c814b27..6d920fd8 100644 --- a/datasette/views/table.py +++ b/datasette/views/table.py @@ -1170,6 +1170,7 @@ class TableInsertView(BaseView): "rowid, " if pks == ["rowid"] else "", table_name, where_clause ), args, + table=table_name, ) result["rows"] = fetched_rows.dicts() else: @@ -1382,7 +1383,9 @@ class TableDropView(BaseView): "database": database_name, "table": table_name, "row_count": ( - await db.execute(f"select count(*) from [{table_name}]") + await db.execute( + f"select count(*) from [{table_name}]", table=table_name + ) ).single_value(), "message": 'Pass "confirm": true to confirm', }, @@ -1576,7 +1579,10 @@ class TableAutocompleteView(BaseView): try: results = await db.execute( - sql, params, custom_time_limit=AUTOCOMPLETE_TIME_LIMIT_MS + sql, + params, + custom_time_limit=AUTOCOMPLETE_TIME_LIMIT_MS, + table=table_name, ) except QueryInterrupted: fallback_where = _autocomplete_prefix_like(pks[0]) @@ -1597,6 +1603,7 @@ class TableAutocompleteView(BaseView): fallback_sql, params, custom_time_limit=AUTOCOMPLETE_TIME_LIMIT_MS, + table=table_name, ) except QueryInterrupted: return Response.json({"ok": True, "rows": []}) @@ -2163,7 +2170,9 @@ async def table_view_data( # Execute the main query! try: - results = await db.execute(sql, params, truncate=True, **extra_args) + results = await db.execute( + sql, params, truncate=True, table=table_name, **extra_args + ) except (sqlite3.OperationalError, InvalidSql) as e: raise DatasetteError(str(e), title="Invalid SQL", status=400) @@ -2439,6 +2448,7 @@ async def _next_value_and_url( await db.execute( prefix_lookup_sql, {**{f"pk{i}": rows[-2][pk] for i, pk in enumerate(pks)}}, + table=table_name, ) ).single_value() if isinstance(prefix, dict) and "value" in prefix: diff --git a/tests/test_telemetry.py b/tests/test_telemetry.py index 7894670c..c49a11cf 100644 --- a/tests/test_telemetry.py +++ b/tests/test_telemetry.py @@ -7,11 +7,18 @@ import time import pytest import sqlite_utils from opentelemetry import trace as otel_trace -from opentelemetry.trace import StatusCode +from opentelemetry.trace import SpanKind, StatusCode from datasette.app import Datasette from datasette.database import Database -from datasette.telemetry import MAX_SQL_LENGTH, sql_attribute, tracer +from datasette.telemetry import ( + MAX_SQL_LENGTH, + SCHEMA_URL, + sql_attribute, + sql_operation_name, + tracer, +) +from datasette.version import __version__ SECRET_PARAM_VALUE = "SUPER_SECRET_PARAM_VALUE_XYZ_123" @@ -543,3 +550,201 @@ async def test_invoke_startup_produces_one_trace_not_dozens_of_orphans(otel_span assert all( _descends_from(span, startup.context, by_span_id) for span in write_spans ) + + +# --- Semantic conventions: span kind, scope, db.operation/collection ------- + + +@pytest.mark.asyncio +async def test_db_query_is_client_kind_and_children_are_internal(otel_spans): + """ + db.query is a database client span; Datasette's decomposition of it is not. + + Trace UIs key their database rendering off the span kind rather than off + db.system, so db.query has to be CLIENT. db.query.execute, + db.write.execute and db.write.queue_wait deliberately stay INTERNAL: they + are parts of one logical query rather than three separate database calls, + and queue_wait touches no database at all - marking them CLIENT would + make one query look like several to anything counting spans by kind. + """ + # Named in-memory databases are shared-cache, so this needs its own name. + db = Datasette(memory=True).add_memory_database("t06_span_kind") + # All four db.query entry points, so a missed `kind=` on any one of them + # fails here - plus the write path (db.write.queue_wait, + # db.write.execute) and the read path (db.query.execute) children. + await db.execute_write("create table docs (id integer primary key)") + await db.execute_write_many( + "insert into docs (id) values (?)", [[i] for i in range(1, 4)] + ) + await db.execute_write_script("insert into docs (id) values (99);") + await db.execute("select id from docs") + + query_spans = _spans_for_namespace(otel_spans, "t06_span_kind") + assert len(query_spans) == 4, "expected a db.query span per entry point" + for span in query_spans: + text = span.attributes["db.query.text"] + assert span.kind == SpanKind.CLIENT, f"db.query for {text!r} should be CLIENT" + + for name in ("db.query.execute", "db.write.execute", "db.write.queue_wait"): + children = [ + span for span in otel_spans.get_finished_spans() if span.name == name + ] + assert children, f"expected at least one {name} span" + for span in children: + assert span.kind == SpanKind.INTERNAL, f"{name} should be INTERNAL" + + +@pytest.mark.asyncio +async def test_instrumentation_scope_declares_version_and_schema_url( + ds_client, otel_spans +): + """ + Spans say which Datasette produced them and which semconv version their + attribute names follow. + + Before get_tracer() was given a version and a schema URL every exported + scope was name='datasette' version='' schema_url='', so nothing + downstream could tell which Datasette a span came from, or whether + `db.system` meant `db.system` or the post-1.30.0 `db.system.name`. + """ + response = await ds_client.get("/fixtures/-/query.json?sql=select+1") + assert response.status_code == 200 + + spans = _db_query_spans(otel_spans) + assert spans, "expected at least one db.query span" + scope = spans[-1].instrumentation_scope + + assert scope.name == "datasette" + assert scope.version == __version__ + # The literal URL, not the SCHEMA_URL constant: comparing the span + # against the same constant the instrumentation is built from would only + # catch a dropped argument, never a wrong value. Bumping this is a claim + # about the attribute names on the wire - see SCHEMA_URL in telemetry.py. + assert scope.schema_url == "https://opentelemetry.io/schemas/1.29.0" + assert SCHEMA_URL == "https://opentelemetry.io/schemas/1.29.0" + assert __version__, "the scope version must not be empty" + + +def test_db_operation_name_from_leading_keyword(): + assert sql_operation_name("select 1") == "SELECT" + assert sql_operation_name(" insert into x (a) values (1)") == "INSERT" + # A leading CTE reports WITH rather than the operation inside it. That is + # the documented limitation, not an accident - see sql_operation_name(). + assert sql_operation_name("with foo as (select 1) select * from foo") == "WITH" + # Unrecognised leading keyword: no attribute rather than a wrong one, and + # no unbounded value set derived from attacker-supplied SQL. + assert sql_operation_name("gibberish 1") is None + # Not a parser: a parenthesised SELECT and a leading comment both yield + # nothing rather than a guess. + assert sql_operation_name("(select 1) union select 2") is None + assert sql_operation_name("-- a comment\nselect 1") is None + assert sql_operation_name("") is None + + +@pytest.mark.asyncio +async def test_db_operation_name_on_real_span(ds_client, otel_spans): + response = await ds_client.get("/fixtures/-/query.json?sql=select+1") + assert response.status_code == 200 + + spans = [ + span + for span in _spans_for_namespace(otel_spans, "fixtures") + if span.attributes["db.query.text"] == "select 1" + ] + assert spans, "expected a db.query span for 'select 1'" + assert spans[-1].attributes["db.operation.name"] == "SELECT" + + +@pytest.mark.asyncio +async def test_execute_write_sets_db_operation_name(otel_spans): + db = Datasette(memory=True).add_memory_database("t06_write_operation") + await db.execute_write("create table docs (id integer primary key)") + await db.execute_write_many( + "insert into docs (id) values (?)", [[i] for i in range(1, 4)] + ) + + spans = _spans_for_namespace(otel_spans, "t06_write_operation") + by_operation = { + span.attributes["db.query.text"]: span.attributes.get("db.operation.name") + for span in spans + } + assert by_operation["create table docs (id integer primary key)"] == "CREATE" + assert by_operation["insert into docs (id) values (?)"] == "INSERT" + + +@pytest.mark.asyncio +async def test_execute_write_script_has_no_operation_name(otel_spans): + """ + executescript() runs several statements, so naming the operation after + the first one would be a lie. Semantic conventions say db.operation.name + should not be extracted from query text that can hold more than one + operation, so the attribute is absent entirely. + + The script deliberately starts with `create`, which *is* on the + allowlist - so this fails if the call site ever starts calling + sql_operation_name(). + """ + db = Datasette(memory=True).add_memory_database("t06_script_operation") + await db.execute_write_script( + "create table docs (id integer primary key);\n" + "insert into docs (id) values (1);" + ) + + spans = _spans_for_namespace(otel_spans, "t06_script_operation") + script_spans = [ + span for span in spans if span.attributes.get("datasette.executescript") is True + ] + assert len(script_spans) == 1 + assert "db.operation.name" not in script_spans[0].attributes + + +@pytest.mark.asyncio +async def test_db_collection_name_set_from_table_argument(ds_client, otel_spans): + db = ds_client.ds.get_database("fixtures") + await db.execute("select pk from facetable limit 1", table="facetable") + + spans = _spans_for_namespace(otel_spans, "fixtures") + assert spans + assert spans[-1].attributes["db.collection.name"] == "facetable" + + +@pytest.mark.asyncio +async def test_db_collection_name_absent_without_table_argument(ds_client, otel_spans): + """ + db.collection.name comes only from an explicit table= argument and is + never derived from the SQL. + + Deriving it would be a parse, and on an instance where anybody can create + a table the value set has no ceiling. Without this test the one above + would still pass if the table name were being read out of the query text. + """ + db = ds_client.ds.get_database("fixtures") + await db.execute("select pk from facetable limit 1") + + spans = _spans_for_namespace(otel_spans, "fixtures") + assert spans + span = spans[-1] + assert span.attributes["db.query.text"] == "select pk from facetable limit 1" + assert "db.collection.name" not in span.attributes + + +@pytest.mark.parametrize( + "path,table", + ( + ("/fixtures/facetable.json", "facetable"), + ("/fixtures/simple_primary_key/1.json", "simple_primary_key"), + ), +) +@pytest.mark.asyncio +async def test_table_and_row_pages_set_db_collection_name( + ds_client, otel_spans, path, table +): + "The table and row views know their table, so their queries carry it." + response = await ds_client.get(path) + assert response.status_code == 200 + + spans = _spans_for_namespace(otel_spans, "fixtures") + assert spans + assert any( + span.attributes.get("db.collection.name") == table for span in spans + ), f"expected a db.query span from {path} carrying db.collection.name" From e24a2c122fb646d72f3db2f2b6cd50e78130726c Mon Sep 17 00:00:00 2001 From: Alex Garcia Date: Thu, 30 Jul 2026 18:36:49 -0700 Subject: [PATCH 07/10] Link block=False write spans to their enqueuer instead of parenting them A block=False write returns without awaiting the reply future, so the enclosing db.query span finishes - and exports - before db.write.queue_wait and db.write.execute even exist. They were still parented to it, which produced a child bar ending ~50ms after its already-closed parent: legal OpenTelemetry, but it renders as nonsense in a trace UI. Parenting asserts containment; a link asserts causation without containment. The enqueueing request causes the write without containing it, which is exactly what a span link is for. So for block=False both write spans are now roots - started with an explicit empty Context, so the write thread's ambient context cannot supply a parent either - each carrying one link back to the enqueueing span. block=True is untouched, since there the caller really does await the reply and containment is accurate. The link carries no attributes. There is only one kind of link here, so naming the relationship would be a constant conveying nothing the link's existence does not already say. Accepted trade-off: a linked span will not appear inside the request's waterfall in most trace UIs. It shows up as its own trace with a "linked from" reference rather than a bar under the request. For a fire-and-forget write whose latency the request never pays, that is the right trade - correctness over at-a-glance nesting for a case the request-latency view was never accurate for anyway. This does add root traces, which looks like it cuts against the startup span work that spent its whole diff removing them. The difference is reachability: those roots were orphans, whereas these are reachable from the request that caused them via the link. Nothing in core issues block=False writes today - it is a plugin-facing path - so this changes no trace Datasette produces on its own. Co-Authored-By: Claude Opus 5 --- datasette/database.py | 85 +++++++++++++++++---- tests/test_telemetry.py | 161 ++++++++++++++++++++++++++++++++++++++++ 2 files changed, 231 insertions(+), 15 deletions(-) diff --git a/datasette/database.py b/datasette/database.py index 1b9a5076..3c14c1c0 100644 --- a/datasette/database.py +++ b/datasette/database.py @@ -14,7 +14,7 @@ from pathlib import Path import sqlite_utils from opentelemetry import context as otel_context_api -from opentelemetry.trace import SpanKind, Status, StatusCode +from opentelemetry.trace import Link, SpanKind, Status, StatusCode, get_current_span from .inspect import inspect_hash from .telemetry import sql_attribute, sql_operation_name, tracer @@ -482,7 +482,9 @@ class Database: # Captured here, on the event loop, at enqueue time: the otel # Context (carrying the enclosing db.query span, if any) and the # timestamp used to build the db.write.queue_wait span once this - # task is dequeued on the write thread. + # task is dequeued on the write thread. `block` travels with the + # task too, because it decides whether that context is this task's + # parent or only a link target - see `_execute_writes`. self._write_queue.put( WriteTask( fn, @@ -493,6 +495,7 @@ class Database: transaction, otel_context_api.get_current(), time.time_ns(), + block, ) ) if block: @@ -531,15 +534,52 @@ class Database: # Best-effort close as the write thread exits pass return - # Restore the caller's otel context (captured on the event loop - # at enqueue time) so spans created while processing this task - # parent correctly to the request that queued it. Must be - # detached below in `finally` - a leaked token silently poisons - # this thread's ambient context for every write processed after - # it, and a *wrong*-token detach only logs a warning rather than - # raising, so this pairing is load-bearing and easy to get wrong - # silently. - token = otel_context_api.attach(task.otel_context) + # `task.block` decides how this task's spans relate to the + # context captured at enqueue time: + # + # - block=True: the caller genuinely awaits the reply, so + # containment is accurate. Restore that context as current + # (attach below) so db.write.queue_wait/db.write.execute parent + # normally to the request that queued them. The token must be + # detached below in `finally` - a leaked token silently + # poisons this thread's ambient context for every write + # processed after it, and a *wrong*-token detach only logs a + # warning rather than raising, so this pairing is load-bearing + # and easy to get wrong silently. + # - block=False: the caller returned already without awaiting, + # so the enqueueing span may already have closed (and + # exported) before this task's spans even start - parenting to + # it would make a child appear to outlive its already-closed + # parent, which OTel allows but which renders badly in most + # trace UIs. The enqueueing request *caused* this write + # without *containing* it, so nothing is attached here - + # instead each write span is started as its own root (explicit + # empty `context=`, so the write thread's ambient context + # cannot supply a parent either) carrying one `Link` back to + # the enqueueing span's context, built once into + # `write_span_kwargs` and spread into every start_span call + # below. + token = None + write_span_kwargs = {} + if task.block: + token = otel_context_api.attach(task.otel_context) + else: + enqueueing_span_context = get_current_span( + task.otel_context + ).get_span_context() + # No attributes on the link: there is only one kind of link + # here, so naming the relationship would be a constant that + # carries no information a consumer does not already have + # from the link's existence. + links = ( + [Link(enqueueing_span_context)] + if enqueueing_span_context.is_valid + else [] + ) + write_span_kwargs = { + "context": otel_context_api.Context(), + "links": links, + } try: exception = None result = None @@ -548,7 +588,9 @@ class Database: # waiting in the queue (enqueue -> dequeue), not the near- # zero time spent constructing/ending the span object here. tracer.start_span( - "db.write.queue_wait", start_time=task.enqueued_at_ns + "db.write.queue_wait", + start_time=task.enqueued_at_ns, + **write_span_kwargs, ).end(end_time=time.time_ns()) if conn_exception is not None: # fn never runs in this branch, so there is nothing to @@ -556,7 +598,9 @@ class Database: exception = conn_exception elif task.isolated_connection: try: - with tracer.start_as_current_span("db.write.execute") as span: + with tracer.start_as_current_span( + "db.write.execute", **write_span_kwargs + ) as span: span.set_attribute( "datasette.isolated_connection", task.isolated_connection, @@ -583,7 +627,9 @@ class Database: exception = e else: try: - with tracer.start_as_current_span("db.write.execute") as span: + with tracer.start_as_current_span( + "db.write.execute", **write_span_kwargs + ) as span: span.set_attribute( "datasette.isolated_connection", task.isolated_connection, @@ -603,7 +649,8 @@ class Database: exception = e _deliver_write_result(task, result, exception) finally: - otel_context_api.detach(token) + if token is not None: + otel_context_api.detach(token) async def execute_fn(self, fn): self._check_not_closed() @@ -1043,6 +1090,7 @@ def _apply_write_wrapper(fn, wrapper_factory, track_event): class WriteTask: __slots__ = ( + "block", "enqueued_at_ns", "fn", "isolated_connection", @@ -1063,6 +1111,7 @@ class WriteTask: transaction, otel_context, enqueued_at_ns, + block, ): self.fn = fn self.task_id = task_id @@ -1072,6 +1121,12 @@ class WriteTask: self.transaction = transaction self.otel_context = otel_context self.enqueued_at_ns = enqueued_at_ns + # Whether the enqueueing caller awaits the reply future. Decides how + # `_execute_writes` relates this task's spans to `otel_context`: + # parent (block=True) or span-link target (block=False). See the + # comment at the WriteTask construction site in + # `_send_to_write_thread`. + self.block = block def _deliver_write_result(task, result, exception): diff --git a/tests/test_telemetry.py b/tests/test_telemetry.py index c49a11cf..0a6d0192 100644 --- a/tests/test_telemetry.py +++ b/tests/test_telemetry.py @@ -2,10 +2,12 @@ import json import sqlite3 import subprocess import sys +import threading import time import pytest import sqlite_utils +from opentelemetry import context as otel_context_api from opentelemetry import trace as otel_trace from opentelemetry.trace import SpanKind, StatusCode @@ -463,6 +465,165 @@ async def test_write_queue_wait_duration_reflects_real_wait(otel_spans): assert duration_ns > 10_000_000, f"queue wait was only {duration_ns}ns" +async def _write_spans_from_one_enqueue(otel_spans, name, block): + """ + Run exactly one write through the write thread from inside a span of our + own, and return (enqueueing span context, {span name: span}). + + `_send_to_write_thread` is called directly rather than `execute_write()` + because `execute_write()` opens its own db.query span, which would then + be the span current at enqueue time - so the parent/link would point at + that span rather than at the one this test controls. + + The exporter is cleared immediately before the enqueue so the write spans + collected here can only have come from this one write. + """ + db = Datasette(memory=True).add_memory_database(name) + await db.execute_write("create table docs (id integer primary key)") + + def insert(conn): + conn.execute("insert into docs (id) values (1)") + + otel_spans.clear() + with tracer.start_as_current_span("enqueueing-span") as enqueuer: + enqueuer_context = enqueuer.get_span_context() + queued = await db._send_to_write_thread(insert, block=block) + if not block: + # The point of block=False is that the write happens after the + # caller has returned and the enqueueing span above has closed. + # Awaiting the reply future outside that `with` waits for the write + # thread deterministically - it is resolved only after both write + # spans have ended and been exported. + _, reply_future = queued + await reply_future + + spans = {} + for span in otel_spans.get_finished_spans(): + if span.name in ("db.write.queue_wait", "db.write.execute"): + assert span.name not in spans, f"more than one {span.name} span" + spans[span.name] = span + assert set(spans) == {"db.write.queue_wait", "db.write.execute"} + return enqueuer_context, spans + + +@pytest.mark.asyncio +async def test_blocking_write_spans_still_parent_normally(otel_spans): + # Regression guard for ticket 07: block=True genuinely has containment - + # the caller awaits the reply future - so those spans must keep parenting + # to the enqueueing span, and must not grow links. + enqueuer_context, spans = await _write_spans_from_one_enqueue( + otel_spans, "t07_blocking_write", block=True + ) + for name, span in spans.items(): + assert span.parent is not None, f"{name} lost its parent" + assert span.parent.span_id == enqueuer_context.span_id, name + assert span.parent.trace_id == enqueuer_context.trace_id, name + assert span.context.trace_id == enqueuer_context.trace_id, name + assert span.links == (), f"{name} should be parented, not linked" + + +@pytest.mark.asyncio +async def test_nonblocking_write_spans_are_roots_with_a_link(otel_spans): + # block=False returns before the write runs, so the enqueueing span has + # already ended (and exported) by the time these spans start. Parenting + # them to it would draw a child outliving its closed parent, so they are + # roots in their own traces, linked back to the span that caused them. + enqueuer_context, spans = await _write_spans_from_one_enqueue( + otel_spans, "t07_nonblocking_write", block=False + ) + assert enqueuer_context.is_valid, "test's own enqueueing span was not recorded" + for name, span in spans.items(): + assert span.parent is None, f"{name} is still parented" + # A link does not join the linked trace: each of these is its own + # root trace, which is the correct shape and not a workaround. + assert span.context.trace_id != enqueuer_context.trace_id, name + assert len(span.links) == 1, f"{name} has links {span.links}" + link_context = span.links[0].context + assert link_context.trace_id == enqueuer_context.trace_id, name + assert link_context.span_id == enqueuer_context.span_id, name + # The two write spans are independent roots, not nested in one another. + assert ( + spans["db.write.queue_wait"].context.trace_id + != spans["db.write.execute"].context.trace_id + ) + + +@pytest.mark.asyncio +async def test_nonblocking_write_link_has_no_attributes(otel_spans): + # There is only one kind of link here, so a relationship-name attribute + # would be a constant conveying nothing the link's existence does not. + _, spans = await _write_spans_from_one_enqueue( + otel_spans, "t07_nonblocking_link_attrs", block=False + ) + for name, span in spans.items(): + assert len(span.links) == 1, name + assert dict(span.links[0].attributes or {}) == {}, name + + +@pytest.mark.asyncio +async def test_nonblocking_write_spans_ignore_the_write_threads_ambient_context( + otel_spans, +): + """ + block=False spans pass an explicit empty Context, not merely "no attach". + + Nothing is attached for a block=False task, but "nothing attached" is not + the same as "no ambient context": the write thread is persistent, and + anything running on it - a prepare_connection plugin hook, say - can + attach a context and never detach it. Without the explicit `context=` + these spans would silently parent to that leftover span instead of being + roots, and no other test here would notice, because in every other test + the write thread's ambient context happens to be empty. + + So this test leaks exactly such a context on the write thread, the way a + careless plugin would, and then checks the write spans are still roots. + """ + ds = Datasette(memory=True) + db = ds.add_memory_database("t07_ambient_write_thread") + write_thread_name = "_execute_writes for database t07_ambient_write_thread" + real_prepare_connection = ds._prepare_connection + leaked = {} + + def prepare_connection(conn, database): + if threading.current_thread().name == write_thread_name: + # Runs once, on the write thread, before any task is dequeued - + # and never detaches, which is the whole point. + span = tracer.start_span("leaked-write-thread-ambient-span") + leaked["span_id"] = span.get_span_context().span_id + otel_context_api.attach(otel_trace.set_span_in_context(span)) + return real_prepare_connection(conn, database) + + ds._prepare_connection = prepare_connection + try: + await db.execute_write("create table docs (id integer primary key)") + + def insert(conn): + conn.execute("insert into docs (id) values (1)") + + otel_spans.clear() + with tracer.start_as_current_span("enqueueing-span") as enqueuer: + enqueuer_context = enqueuer.get_span_context() + _, reply_future = await db._send_to_write_thread(insert, block=False) + await reply_future + finally: + ds._prepare_connection = real_prepare_connection + db.close() + + assert "span_id" in leaked, "the ambient context was never leaked - test is vacuous" + write_spans = [ + span + for span in otel_spans.get_finished_spans() + if span.name in ("db.write.queue_wait", "db.write.execute") + ] + assert len(write_spans) == 2 + for span in write_spans: + assert span.parent is None, ( + f"{span.name} parented to the write thread's leftover ambient " + "context instead of being a root" + ) + assert span.links[0].context.span_id == enqueuer_context.span_id + + @pytest.mark.asyncio async def test_suppressed_error_does_not_mark_execute_span(ds_client, otel_spans): """ From 8d32eac89508453eb519ae6d4377950aa03a19a0 Mon Sep 17 00:00:00 2001 From: Alex Garcia Date: Thu, 30 Jul 2026 18:46:34 -0700 Subject: [PATCH 08/10] Name every span and attribute once, in a registry the docs are built from The span and attribute names were string literals spread across four call sites in database.py and one in app.py, with a hand-written reference page that would have been true only on the day it was written. That drift is not hypothetical: an earlier iteration of this work carried a README asserting parameter values were never recorded for two branches after that had stopped being true. datasette/telemetry_registry.py now holds each name once, with its documentation. Attribute and SpanName subclass str, so a registry entry *is* the string OpenTelemetry wants - no wrapper API over the OTel calls, no parallel structure to keep in step, and a typo becomes an ImportError rather than a silently misnamed attribute. docs/internals.rst renders the span reference from it via cog, and `cog --check docs/*.rst` already runs in CI, so the reference cannot drift from the definitions. Nothing changes on the wire: the emitted span names and attribute keys are byte-identical before and after, verified by diffing a dump of both. tests/test_telemetry_registry.py exercises a real workload and compares it against the registry in both directions - emitted-but-unregistered catches instrumentation added without documentation, registered-but-never-emitted catches documentation that has outlived its code. Because the call sites now take their names from the registry, neither direction can catch a rename: move DB_NAMESPACE to "db.namespace2" and code and registry still agree while every dashboard breaks. So the literal names are also written out in the test and asserted against the registry and against the wire separately. That pair is the only comparison in the file not derived from the registry itself. Co-Authored-By: Claude Opus 5 --- datasette/app.py | 3 +- datasette/database.py | 106 ++++++----- datasette/telemetry_registry.py | 269 +++++++++++++++++++++++++++ docs/internals.rst | 72 +++++++ docs/telemetry_doc.py | 37 ++++ tests/test_telemetry_registry.py | 309 +++++++++++++++++++++++++++++++ 6 files changed, 751 insertions(+), 45 deletions(-) create mode 100644 datasette/telemetry_registry.py create mode 100644 docs/telemetry_doc.py create mode 100644 tests/test_telemetry_registry.py diff --git a/datasette/app.py b/datasette/app.py index d3626ab2..7499c6f9 100644 --- a/datasette/app.py +++ b/datasette/app.py @@ -50,6 +50,7 @@ from .plugins import DEFAULT_PLUGINS, get_plugins, pm from .renderer import json_renderer from .resources import DatabaseResource, TableResource from .telemetry import tracer +from .telemetry_registry import STARTUP from .tokens import TokenInvalid from .tracer import AsgiTracer from .url_builder import Urls @@ -788,7 +789,7 @@ class Datasette: # A connection warmed lazily later, by a request touching a new # database for the first time, nests under that request instead: # this span has already ended by then. - with tracer.start_as_current_span("datasette.startup"): + with tracer.start_as_current_span(STARTUP): # Register event classes event_classes = [] for hook in pm.hook.register_events(datasette=self): diff --git a/datasette/database.py b/datasette/database.py index 3c14c1c0..22984d6f 100644 --- a/datasette/database.py +++ b/datasette/database.py @@ -14,10 +14,32 @@ from pathlib import Path import sqlite_utils from opentelemetry import context as otel_context_api -from opentelemetry.trace import Link, SpanKind, Status, StatusCode, get_current_span +from opentelemetry.trace import Link, Status, StatusCode, get_current_span from .inspect import inspect_hash from .telemetry import sql_attribute, sql_operation_name, tracer +from .telemetry_registry import ( + DB_COLLECTION_NAME, + DB_NAMESPACE, + DB_OPERATION_NAME, + DB_QUERY, + DB_QUERY_EXECUTE, + DB_QUERY_TEXT, + DB_SYSTEM, + DB_WRITE_EXECUTE, + DB_WRITE_QUEUE_WAIT, + EXECUTEMANY, + EXECUTESCRIPT, + INTERRUPTED, + ISOLATED_CONNECTION, + PARAM_COUNT, + PARAM_SETS, + ROWS_RETURNED, + SQL_ERROR_SUPPRESSED, + TIME_LIMIT_MS, + TRANSACTION, + TRUNCATED, +) from .tracer import trace from .utils import ( call_with_supported_arguments, @@ -268,15 +290,15 @@ class Database: with trace( # noqa: SIM117 "sql", database=self.name, sql=sql.strip(), params=params ): - with tracer.start_as_current_span("db.query", kind=SpanKind.CLIENT) as span: - span.set_attribute("db.system", "sqlite") - span.set_attribute("db.namespace", self.name) - span.set_attribute("db.query.text", sql_attribute(sql)) + with tracer.start_as_current_span(DB_QUERY, kind=DB_QUERY.kind) as span: + span.set_attribute(DB_SYSTEM, "sqlite") + span.set_attribute(DB_NAMESPACE, self.name) + span.set_attribute(DB_QUERY_TEXT, sql_attribute(sql)) operation_name = sql_operation_name(sql) if operation_name: - span.set_attribute("db.operation.name", operation_name) + span.set_attribute(DB_OPERATION_NAME, operation_name) if params: - span.set_attribute("datasette.param_count", len(params)) + span.set_attribute(PARAM_COUNT, len(params)) results = await self.execute_write_fn( _inner, block=block, request=request, transaction=transaction ) @@ -296,11 +318,11 @@ class Database: # several semicolon-separated statements, and semantic conventions # say the attribute should not be extracted from query text that # can hold more than one operation - see sql_operation_name(). - with tracer.start_as_current_span("db.query", kind=SpanKind.CLIENT) as span: - span.set_attribute("db.system", "sqlite") - span.set_attribute("db.namespace", self.name) - span.set_attribute("db.query.text", sql_attribute(sql)) - span.set_attribute("datasette.executescript", True) + with tracer.start_as_current_span(DB_QUERY, kind=DB_QUERY.kind) as span: + span.set_attribute(DB_SYSTEM, "sqlite") + span.set_attribute(DB_NAMESPACE, self.name) + span.set_attribute(DB_QUERY_TEXT, sql_attribute(sql)) + span.set_attribute(EXECUTESCRIPT, True) results = await self.execute_write_fn( _inner, block=block, transaction=False, request=request ) @@ -324,22 +346,22 @@ class Database: with trace( "sql", database=self.name, sql=sql.strip(), executemany=True ) as kwargs: - with tracer.start_as_current_span("db.query", kind=SpanKind.CLIENT) as span: - span.set_attribute("db.system", "sqlite") - span.set_attribute("db.namespace", self.name) - span.set_attribute("db.query.text", sql_attribute(sql)) - span.set_attribute("datasette.executemany", True) + with tracer.start_as_current_span(DB_QUERY, kind=DB_QUERY.kind) as span: + span.set_attribute(DB_SYSTEM, "sqlite") + span.set_attribute(DB_NAMESPACE, self.name) + span.set_attribute(DB_QUERY_TEXT, sql_attribute(sql)) + span.set_attribute(EXECUTEMANY, True) # A single statement run with many parameter sets, so unlike # execute_write_script() there is exactly one operation to name. operation_name = sql_operation_name(sql) if operation_name: - span.set_attribute("db.operation.name", operation_name) + span.set_attribute(DB_OPERATION_NAME, operation_name) results, count = await self.execute_write_fn( _inner, block=block, request=request ) # count is the number of parameter *sets* consumed by # executemany(), not a row count - executemany returns no rows. - span.set_attribute("datasette.param_sets", count) + span.set_attribute(PARAM_SETS, count) kwargs["count"] = count return results @@ -588,7 +610,7 @@ class Database: # waiting in the queue (enqueue -> dequeue), not the near- # zero time spent constructing/ending the span object here. tracer.start_span( - "db.write.queue_wait", + DB_WRITE_QUEUE_WAIT, start_time=task.enqueued_at_ns, **write_span_kwargs, ).end(end_time=time.time_ns()) @@ -599,15 +621,13 @@ class Database: elif task.isolated_connection: try: with tracer.start_as_current_span( - "db.write.execute", **write_span_kwargs + DB_WRITE_EXECUTE, **write_span_kwargs ) as span: span.set_attribute( - "datasette.isolated_connection", + ISOLATED_CONNECTION, task.isolated_connection, ) - span.set_attribute( - "datasette.transaction", task.transaction - ) + span.set_attribute(TRANSACTION, task.transaction) isolated_connection = self.connect(write=True) try: result = task.fn(isolated_connection) @@ -628,15 +648,13 @@ class Database: else: try: with tracer.start_as_current_span( - "db.write.execute", **write_span_kwargs + DB_WRITE_EXECUTE, **write_span_kwargs ) as span: span.set_attribute( - "datasette.isolated_connection", + ISOLATED_CONNECTION, task.isolated_connection, ) - span.set_attribute( - "datasette.transaction", task.transaction - ) + span.set_attribute(TRANSACTION, task.transaction) if task.transaction: with conn: conn.execute("BEGIN IMMEDIATE") @@ -720,7 +738,7 @@ class Database: # db.query span in execute(). Without this, facet suggestion marks # two spans per text column as failed on every table page. with tracer.start_as_current_span( - "db.query.execute", + DB_QUERY_EXECUTE, record_exception=log_sql_errors, set_status_on_exception=log_sql_errors, ): @@ -764,27 +782,27 @@ class Database: # manager's defaults, so that callers passing log_sql_errors=False # can be honoured - see the comment on the generic handler below. with tracer.start_as_current_span( - "db.query", - kind=SpanKind.CLIENT, + DB_QUERY, + kind=DB_QUERY.kind, record_exception=False, set_status_on_exception=False, ) as span: - span.set_attribute("db.system", "sqlite") - span.set_attribute("db.namespace", self.name) - span.set_attribute("db.query.text", sql_attribute(sql)) - span.set_attribute("datasette.time_limit_ms", time_limit_ms) + span.set_attribute(DB_SYSTEM, "sqlite") + span.set_attribute(DB_NAMESPACE, self.name) + span.set_attribute(DB_QUERY_TEXT, sql_attribute(sql)) + span.set_attribute(TIME_LIMIT_MS, time_limit_ms) operation_name = sql_operation_name(sql) if operation_name: - span.set_attribute("db.operation.name", operation_name) + span.set_attribute(DB_OPERATION_NAME, operation_name) if table: - span.set_attribute("db.collection.name", table) + span.set_attribute(DB_COLLECTION_NAME, table) if params: - span.set_attribute("datasette.param_count", len(params)) + span.set_attribute(PARAM_COUNT, len(params)) try: results = await self.execute_fn(sql_operation_in_thread) except QueryInterrupted as e: span.set_status(Status(StatusCode.ERROR, str(e))) - span.set_attribute("datasette.interrupted", True) + span.set_attribute(INTERRUPTED, True) span.record_exception(e) raise except Exception as e: @@ -799,10 +817,10 @@ class Database: span.record_exception(e) span.set_status(Status(StatusCode.ERROR, str(e))) else: - span.set_attribute("datasette.sql_error_suppressed", True) + span.set_attribute(SQL_ERROR_SUPPRESSED, True) raise - span.set_attribute("datasette.truncated", results.truncated) - span.set_attribute("datasette.rows_returned", len(results.rows)) + span.set_attribute(TRUNCATED, results.truncated) + span.set_attribute(ROWS_RETURNED, len(results.rows)) return results @property diff --git a/datasette/telemetry_registry.py b/datasette/telemetry_registry.py new file mode 100644 index 00000000..40b32cc5 --- /dev/null +++ b/datasette/telemetry_registry.py @@ -0,0 +1,269 @@ +""" +The single source of truth for every span and span attribute that Datasette +core emits. + +Three things read this module, which is the point of it existing: + +1. **The instrumentation itself.** `Attribute` and `SpanName` subclass `str`, + so a registry entry *is* the string OpenTelemetry wants. Call sites pass + `DB_NAMESPACE` where they used to pass `"db.namespace"` - no wrapper API + over the OTel calls, no parallel structure to keep in step, and a typo is + now an `ImportError` instead of a silently misnamed attribute. + +2. **The documentation.** `docs/telemetry_doc.py` renders the span reference + in `docs/internals.rst` from these definitions using cog, and + `cog --check` runs in CI - so the docs cannot drift from the code. + +3. **A conformance test.** `tests/test_telemetry_registry.py` makes real + requests, collects every span and attribute actually emitted, and compares + both directions: emitted-but-unregistered catches instrumentation added + without documentation, registered-but-never-emitted catches documentation + describing something that no longer exists. Neither the type system nor + the generated docs can catch that second case. +""" + +from opentelemetry.trace import SpanKind + + +class Attribute(str): + """ + A span attribute key, carrying its own documentation. + + Subclasses `str` so it can be handed straight to `set_attribute()`. + """ + + __slots__ = ("description", "optional") + + def __new__(cls, name, description, optional=False): + self = super().__new__(cls, name) + self.description = description + self.optional = optional + return self + + def __repr__(self): + return f"Attribute({str(self)!r})" + + +class SpanName(str): + "A span name, carrying its documentation and the attributes it may set." + + __slots__ = ("attributes", "description", "kind", "prefix") + + def __new__( + cls, name, description, attributes=(), prefix=False, kind=SpanKind.INTERNAL + ): + self = super().__new__(cls, name) + self.description = description + self.attributes = tuple(attributes) + # True for a span family whose emitted names carry a variable suffix, + # so the conformance test matches by prefix rather than equality. + # Nothing sets it yet. + self.prefix = prefix + # SpanKind.INTERNAL by default - every span Datasette emits describes + # its own internal work. db.query is the one exception: it is a real + # database call, so semantic conventions (and trace UIs, which key + # their database styling off this) expect SpanKind.CLIENT. + self.kind = kind + return self + + def __repr__(self): + return f"SpanName({str(self)!r})" + + +# --- Attributes ----------------------------------------------------------- +# +# Shared attributes are defined once and referenced by every span that sets +# them, so "which spans carry db.namespace?" is answerable by grep. + +DB_SYSTEM = Attribute("db.system", "Always ``sqlite``.") +DB_NAMESPACE = Attribute("db.namespace", "Name of the database being queried.") +DB_QUERY_TEXT = Attribute( + "db.query.text", + "The SQL, truncated to 2048 characters. Never the parameter values.", +) +DB_OPERATION_NAME = Attribute( + "db.operation.name", + "The statement's leading keyword - ``SELECT``, ``INSERT``, ``CREATE``, and " + "so on - matched against a small fixed allowlist. Omitted rather than set " + "to an arbitrary value: the allowlist exists because this attribute is a " + "candidate dimension for a query-duration metric in a later phase, and " + "echoing an unrecognised first token from user-supplied SQL would be an " + "unbounded-cardinality hazard. Also omitted for " + "``execute_write_script()``, which runs multiple statements - per " + "semantic conventions, the operation name should not be extracted from " + "query text that can contain more than one operation. Note that a " + "statement beginning with a CTE reports ``WITH``, not the operation " + "inside it - a substantial share of Datasette's own reads take that " + "form. Resolving it further would mean parsing.", + optional=True, +) +DB_COLLECTION_NAME = Attribute( + "db.collection.name", + "The primary table, set only where the view already knows it - the table " + "and row pages. Omitted for arbitrary ``?sql=`` queries, where determining " + "the table would mean parsing the query.", + optional=True, +) + +PARAM_COUNT = Attribute( + "datasette.param_count", + "Number of bound parameters. Recorded instead of the values themselves.", + optional=True, +) +PARAM_SETS = Attribute( + "datasette.param_sets", + "Number of parameter sets consumed by ``execute_write_many()``. Not a row " + "count - ``executemany()`` returns no rows. The parameter values " + "themselves are never recorded: that sequence can hold thousands of rows.", + optional=True, +) +TIME_LIMIT_MS = Attribute( + "datasette.time_limit_ms", + "The :ref:`setting_sql_time_limit_ms` value this query ran under. Set on " + "reads, which are the queries that time limit applies to.", + optional=True, +) +ROWS_RETURNED = Attribute( + "datasette.rows_returned", + "Number of rows a read returned. Set on the read path only, and only when " + "the read succeeded.", + optional=True, +) +TRUNCATED = Attribute( + "datasette.truncated", + "True if the result was cut short by :ref:`setting_max_returned_rows`.", + optional=True, +) +INTERRUPTED = Attribute( + "datasette.interrupted", + "True if the query was cancelled for exceeding the time limit. The span " + "status is also set to ``ERROR``.", + optional=True, +) +SQL_ERROR_SUPPRESSED = Attribute( + "datasette.sql_error_suppressed", + "True when the query failed but the caller passed ``log_sql_errors=False``, " + "meaning it was probing and treats failure as an expected answer. Facet " + "suggestion does this against every column.", + optional=True, +) +EXECUTESCRIPT = Attribute( + "datasette.executescript", + "True for ``execute_write_script()``, which runs multiple statements.", + optional=True, +) +EXECUTEMANY = Attribute( + "datasette.executemany", + "True for ``execute_write_many()``, which runs one statement against many " + "parameter sets.", + optional=True, +) +ISOLATED_CONNECTION = Attribute( + "datasette.isolated_connection", + "True if the write ran on its own connection rather than the shared write " + "connection.", +) +TRANSACTION = Attribute( + "datasette.transaction", + "False for statements such as ``VACUUM`` that cannot run inside a transaction.", +) + + +# --- Spans ---------------------------------------------------------------- + +DB_QUERY = SpanName( + "db.query", + "A SQL operation issued by Datasette, covering the full round trip " + "including any time spent queued for a thread.", + ( + DB_SYSTEM, + DB_NAMESPACE, + DB_QUERY_TEXT, + DB_OPERATION_NAME, + DB_COLLECTION_NAME, + PARAM_COUNT, + PARAM_SETS, + TIME_LIMIT_MS, + ROWS_RETURNED, + TRUNCATED, + INTERRUPTED, + SQL_ERROR_SUPPRESSED, + EXECUTESCRIPT, + EXECUTEMANY, + ), + kind=SpanKind.CLIENT, +) + +DB_QUERY_EXECUTE = SpanName( + "db.query.execute", + "The read executing inside a SQL worker thread. Child of ``db.query``; the " + "gap between the two is time spent waiting for a thread.", +) + +DB_WRITE_QUEUE_WAIT = SpanName( + "db.write.queue_wait", + "Time a write spent waiting in its database's write queue before the write " + "thread picked it up. Child of ``db.query`` for a ``block=True`` write, " + "where the caller awaits the write and containment is accurate. For a " + "``block=False`` write the caller does not await it - the enqueueing " + "request *caused* the write without *containing* it, and the write's " + "spans can outlive the request's own - so this is a root span instead, " + "carrying an OpenTelemetry link back to the enqueueing span rather than " + "a parent. A link records causation without asserting containment, which " + "is exactly the distinction here.", +) + +DB_WRITE_EXECUTE = SpanName( + "db.write.execute", + "The write executing on the write thread. Child of ``db.query`` for a " + "``block=True`` write; for ``block=False`` a root span with a link back " + "to the enqueueing span instead - see ``db.write.queue_wait`` above.", + (ISOLATED_CONNECTION, TRANSACTION), +) + +STARTUP = SpanName( + "datasette.startup", + "``invoke_startup()`` running: ``register_events``, ``register_actions``, " + "``register_column_types``, ``prepare_jinja2_environment``, internal-database " + "schema catalog refresh (including the ``prepare_connection`` warm-up this " + "triggers for each database touched for the first time), saved queries, " + "column type config and the ``startup`` hook. Runs once per process, before " + "any request exists, so without this span every child it creates would be " + "its own orphan root trace. A connection warmed later - lazily, the first " + "time a *request* touches a new database or thread - nests under that " + "request's own span instead, not under this one, since this span has " + "already ended by then.", +) + +SPANS = ( + DB_QUERY, + DB_QUERY_EXECUTE, + DB_WRITE_QUEUE_WAIT, + DB_WRITE_EXECUTE, + STARTUP, +) + + +def span_for(emitted_name): + """ + Resolve an emitted span name to its registry entry, or None. + + Handles span families whose emitted names carry a suffix that is not + knowable in advance - `prefix=True` entries. Phase 1 has none, but the + lookup is what the conformance test calls, so it lives here rather than + in the test. + """ + for span in SPANS: + if span.prefix: + if emitted_name.startswith(span): + return span + elif emitted_name == span: + return span + return None + + +def attribute_allowed(span, emitted_key): + "Whether `emitted_key` is a registered attribute of `span`." + if span is None: + return False + return emitted_key in span.attributes diff --git a/docs/internals.rst b/docs/internals.rst index d2bd46ef..b2215d9d 100644 --- a/docs/internals.rst +++ b/docs/internals.rst @@ -2313,6 +2313,78 @@ The ``Database`` class also provides properties and methods for introspecting th } } +.. _internals_telemetry: + +OpenTelemetry +============= + +Datasette core depends on `opentelemetry-api `__ only. It never creates a ``TracerProvider``, never configures an exporter and never sets a sampler. With no OpenTelemetry SDK provider installed, every span described below is a no-op ``NonRecordingSpan`` - the overhead is close to zero and nothing is recorded or exported anywhere. + +Turning tracing on is entirely an operational decision made outside of Datasette itself: run Datasette under the standard ``opentelemetry-instrument`` agent, or embed Datasette inside a host application that installs its own provider. + +Everything Datasette emits carries the instrumentation scope ``datasette``, versioned with the running Datasette version and declaring the `semantic conventions schema `__ its attribute names follow. + +Span reference +-------------- + +Attribute names use the ``datasette.*`` prefix for Datasette-specific data, alongside standard OpenTelemetry attributes such as ``db.system``. + +This reference is generated from ``datasette/telemetry_registry.py``, the single source of truth for every span and attribute Datasette emits. A conformance test makes real requests and compares what is actually emitted against that registry in both directions, so nothing here is hand-maintained and nothing can silently drift out of date. + +Spans are ``SpanKind.INTERNAL`` unless a kind is listed below. Only ``db.query`` is ``CLIENT``: it is the one span that represents a call to a database rather than Datasette's own work, and trace UIs use the kind to decide whether to render a span as a database call. Its children stay ``INTERNAL`` because they are Datasette's decomposition of that one query - marking them ``CLIENT`` too would make a single query look like several database calls to anything counting by kind. + +.. [[[cog + from telemetry_doc import spans + spans(cog) +.. ]]] + +``db.query`` + A SQL operation issued by Datasette, covering the full round trip including any time spent queued for a thread. + + Kind: ``CLIENT``. + + Attributes: + + - ``db.system`` - Always ``sqlite``. + - ``db.namespace`` - Name of the database being queried. + - ``db.query.text`` - The SQL, truncated to 2048 characters. Never the parameter values. + - ``db.operation.name`` *(optional)* - The statement's leading keyword - ``SELECT``, ``INSERT``, ``CREATE``, and so on - matched against a small fixed allowlist. Omitted rather than set to an arbitrary value: the allowlist exists because this attribute is a candidate dimension for a query-duration metric in a later phase, and echoing an unrecognised first token from user-supplied SQL would be an unbounded-cardinality hazard. Also omitted for ``execute_write_script()``, which runs multiple statements - per semantic conventions, the operation name should not be extracted from query text that can contain more than one operation. Note that a statement beginning with a CTE reports ``WITH``, not the operation inside it - a substantial share of Datasette's own reads take that form. Resolving it further would mean parsing. + - ``db.collection.name`` *(optional)* - The primary table, set only where the view already knows it - the table and row pages. Omitted for arbitrary ``?sql=`` queries, where determining the table would mean parsing the query. + - ``datasette.param_count`` *(optional)* - Number of bound parameters. Recorded instead of the values themselves. + - ``datasette.param_sets`` *(optional)* - Number of parameter sets consumed by ``execute_write_many()``. Not a row count - ``executemany()`` returns no rows. The parameter values themselves are never recorded: that sequence can hold thousands of rows. + - ``datasette.time_limit_ms`` *(optional)* - The :ref:`setting_sql_time_limit_ms` value this query ran under. Set on reads, which are the queries that time limit applies to. + - ``datasette.rows_returned`` *(optional)* - Number of rows a read returned. Set on the read path only, and only when the read succeeded. + - ``datasette.truncated`` *(optional)* - True if the result was cut short by :ref:`setting_max_returned_rows`. + - ``datasette.interrupted`` *(optional)* - True if the query was cancelled for exceeding the time limit. The span status is also set to ``ERROR``. + - ``datasette.sql_error_suppressed`` *(optional)* - True when the query failed but the caller passed ``log_sql_errors=False``, meaning it was probing and treats failure as an expected answer. Facet suggestion does this against every column. + - ``datasette.executescript`` *(optional)* - True for ``execute_write_script()``, which runs multiple statements. + - ``datasette.executemany`` *(optional)* - True for ``execute_write_many()``, which runs one statement against many parameter sets. + +``db.query.execute`` + The read executing inside a SQL worker thread. Child of ``db.query``; the gap between the two is time spent waiting for a thread. + + No attributes. + +``db.write.queue_wait`` + Time a write spent waiting in its database's write queue before the write thread picked it up. Child of ``db.query`` for a ``block=True`` write, where the caller awaits the write and containment is accurate. For a ``block=False`` write the caller does not await it - the enqueueing request *caused* the write without *containing* it, and the write's spans can outlive the request's own - so this is a root span instead, carrying an OpenTelemetry link back to the enqueueing span rather than a parent. A link records causation without asserting containment, which is exactly the distinction here. + + No attributes. + +``db.write.execute`` + The write executing on the write thread. Child of ``db.query`` for a ``block=True`` write; for ``block=False`` a root span with a link back to the enqueueing span instead - see ``db.write.queue_wait`` above. + + Attributes: + + - ``datasette.isolated_connection`` - True if the write ran on its own connection rather than the shared write connection. + - ``datasette.transaction`` - False for statements such as ``VACUUM`` that cannot run inside a transaction. + +``datasette.startup`` + ``invoke_startup()`` running: ``register_events``, ``register_actions``, ``register_column_types``, ``prepare_jinja2_environment``, internal-database schema catalog refresh (including the ``prepare_connection`` warm-up this triggers for each database touched for the first time), saved queries, column type config and the ``startup`` hook. Runs once per process, before any request exists, so without this span every child it creates would be its own orphan root trace. A connection warmed later - lazily, the first time a *request* touches a new database or thread - nests under that request's own span instead, not under this one, since this span has already ended by then. + + No attributes. + +.. [[[end]]] + .. _internals_csrf: CSRF protection diff --git a/docs/telemetry_doc.py b/docs/telemetry_doc.py new file mode 100644 index 00000000..3551ef0e --- /dev/null +++ b/docs/telemetry_doc.py @@ -0,0 +1,37 @@ +""" +Render the span reference in ``internals.rst`` from +``datasette/telemetry_registry.py``. + +Driven by cog, and ``cog --check docs/*.rst`` runs in CI - so adding a span +without documenting it, or documenting one that no longer exists, is a build +failure rather than something a reader discovers later. +""" + + +def _attribute_lines(cog, attributes): + if not attributes: + cog.out(" No attributes.\n\n") + return + cog.out(" Attributes:\n\n") + for attribute in attributes: + suffix = " *(optional)*" if attribute.optional else "" + cog.out(f" - ``{attribute}``{suffix} - {attribute.description}\n") + cog.out("\n") + + +def spans(cog): + from opentelemetry.trace import SpanKind + + from datasette.telemetry_registry import SPANS + + cog.out("\n") + for span in SPANS: + title = f"{span}*" if span.prefix else str(span) + cog.out(f"``{title}``\n") + cog.out(f" {span.description}\n\n") + # INTERNAL is the default and the overwhelming majority of spans - + # printing it on every one would be noise. Only the exceptional case, + # a real database call, is worth calling out. + if span.kind != SpanKind.INTERNAL: + cog.out(f" Kind: ``{span.kind.name}``.\n\n") + _attribute_lines(cog, span.attributes) diff --git a/tests/test_telemetry_registry.py b/tests/test_telemetry_registry.py new file mode 100644 index 00000000..335eb946 --- /dev/null +++ b/tests/test_telemetry_registry.py @@ -0,0 +1,309 @@ +""" +Two-way conformance between `datasette/telemetry_registry.py` and what +Datasette actually emits. + +This is the test that makes the generated documentation trustworthy. cog +guarantees the docs match the registry; this guarantees the registry matches +the code. Without it, both could agree with each other and be wrong. + +It checks both directions, and the second one is the one nothing else catches: + +- **emitted but not registered** - instrumentation was added without + documenting it, so the reference page silently omits it. +- **registered but never emitted** - the reference page describes a span or + attribute that no longer exists, which is worse than omitting it, because a + reader will build a dashboard on it. + +Both of those directions compare the code against the registry. Neither can +catch a *rename*, because the call sites now take their names from the +registry - move `DB_NAMESPACE` to `"db.namespace2"` and code and registry +still agree with each other, while every existing dashboard breaks. So the +literal names live here too, spelled out, and are asserted against both the +registry and the wire. That is the one comparison in this file that is not +made against a value derived from the registry itself. +""" + +import itertools + +import pytest +import pytest_asyncio + +pytest.importorskip("opentelemetry.sdk") + +from datasette import telemetry_registry as reg +from datasette.app import Datasette +from datasette.database import QueryInterrupted +from datasette.utils.sqlite import sqlite3 + +# The names as they appear on the wire, written out rather than read from the +# registry. If a change to the registry makes one of these fail, that change +# is renaming something a user's dashboards and saved queries depend on - +# which is a decision to take deliberately, here, not a line to re-derive. +EXPECTED_ATTRIBUTES = { + "db.query": { + "db.system", + "db.namespace", + "db.query.text", + "db.operation.name", + "db.collection.name", + "datasette.param_count", + "datasette.param_sets", + "datasette.time_limit_ms", + "datasette.rows_returned", + "datasette.truncated", + "datasette.interrupted", + "datasette.sql_error_suppressed", + "datasette.executescript", + "datasette.executemany", + }, + "db.query.execute": set(), + "db.write.queue_wait": set(), + "db.write.execute": { + "datasette.isolated_connection", + "datasette.transaction", + }, + "datasette.startup": set(), +} +EXPECTED_SPANS = set(EXPECTED_ATTRIBUTES) + +# Named in-memory databases are shared-cache, so two Datasette instances using +# the same name share one SQLite database - and the second `create table` +# fails. Every workload below therefore gets its own name. +_names = itertools.count() + + +def _unique(prefix): + return f"{prefix}{next(_names)}" + + +async def exercise(): + """ + Drive enough of Datasette to emit every span and attribute the registry + claims exists. + + Each call is here because it is the only thing that produces some span or + attribute - see the comments. If you add instrumentation on a path this + does not reach, add the path rather than loosening the assertions. + + Returns the instance so the caller can close it; startup happens inside + so that the `datasette.startup` span lands in the collected set. + """ + name = _unique("registry") + ds = Datasette(memory=True) + ds.add_memory_database(name) + # datasette.startup - and the internal catalog work nested under it + await ds.invoke_startup() + db = ds.get_database(name) + + # Writes: db.write.queue_wait, db.write.execute, db.query + await db.execute_write("create table t (id integer primary key, v text)") + # datasette.executemany, datasette.param_sets + await db.execute_write_many( + "insert into t (id, v) values (?, ?)", [[i, f"v{i}"] for i in range(30)] + ) + # datasette.executescript + await db.execute_write_script("create table t2 (id integer); drop table t2;") + # datasette.transaction=False - VACUUM cannot run inside a transaction + await db.execute_write("vacuum", transaction=False) + # datasette.isolated_connection=True + await db.execute_isolated_fn(lambda conn: conn.execute("select 1").fetchone()) + + # Reads: db.query.execute, datasette.rows_returned, datasette.truncated, + # datasette.param_count, datasette.time_limit_ms + await db.execute("select * from t where id > :n", {"n": 5}) + await db.execute("select * from t", truncate=True) + + # datasette.sql_error_suppressed - the caller is probing and treats + # failure as an expected answer + with pytest.raises(sqlite3.OperationalError): + await db.execute("select nope from t", log_sql_errors=False) + + # datasette.interrupted - only ever set when a query exceeds its time + # limit, so the workload has to force one rather than exempt it. An + # unbounded recursive CTE cannot finish, so 1ms is always exceeded. + with pytest.raises(QueryInterrupted): + await db.execute( + "with recursive c(x) as (select 0 union all select x+1 from c) " + "select * from c", + custom_time_limit=1, + ) + + # db.collection.name - set only by views that already know their table + assert (await ds.client.get(f"/{name}/t?_facet=v")).status_code == 200 + assert (await ds.client.get(f"/{name}/t/1.json")).status_code == 200 + return ds + + +@pytest_asyncio.fixture +async def emitted(otel_spans): + "Every span name and (span name, attribute key) pair a broad workload emits." + # otel_spans has already cleared the exporter, and nothing is cleared + # after this point: the workload's own startup emits datasette.startup. + ds = await exercise() + spans = otel_spans.get_finished_spans() + assert spans, "no spans captured - the fixture is not exercising anything" + names = set() + pairs = set() + for span in spans: + # str() because span.name is the registry's SpanName instance, and a + # set of those would compare equal to literals but read confusingly + # in a failure message. + names.add(str(span.name)) + for key in span.attributes or {}: + pairs.add((str(span.name), str(key))) + ds.close() + return {"names": names, "pairs": pairs} + + +def _keys_by_span(pairs): + by_span = {} + for span_name, key in pairs: + by_span.setdefault(span_name, set()).add(key) + return by_span + + +@pytest.mark.asyncio +async def test_workload_emits_exactly_the_expected_names(emitted): + """ + The wire format, pinned to literals. + + Not derived from the registry, so this is what catches a rename that the + registry and the call sites make together. + """ + assert emitted["names"] == EXPECTED_SPANS + by_span = _keys_by_span(emitted["pairs"]) + assert {name: by_span.get(name, set()) for name in emitted["names"]} == ( + EXPECTED_ATTRIBUTES + ) + + +def test_registry_matches_the_expected_names(): + "The other half of the rename check: the registry against the same literals." + assert {str(span) for span in reg.SPANS} == EXPECTED_SPANS + for span in reg.SPANS: + assert {str(attribute) for attribute in span.attributes} == EXPECTED_ATTRIBUTES[ + str(span) + ], f"{span} attributes have drifted" + + +@pytest.mark.asyncio +async def test_every_emitted_span_is_registered(emitted): + "A span added without a registry entry would be missing from the docs." + unregistered = sorted( + name for name in emitted["names"] if reg.span_for(name) is None + ) + assert ( + not unregistered + ), f"these spans are emitted but not in telemetry_registry.SPANS: {unregistered}" + + +@pytest.mark.asyncio +async def test_every_emitted_attribute_is_registered(emitted): + "An attribute added without a registry entry would be missing from the docs." + unregistered = sorted( + f"{span_name} -> {key}" + for span_name, key in emitted["pairs"] + if not reg.attribute_allowed(reg.span_for(span_name), key) + ) + assert ( + not unregistered + ), "these span attributes are emitted but not registered: " + ", ".join( + unregistered + ) + + +@pytest.mark.asyncio +async def test_every_registered_span_is_emitted(emitted): + """ + The direction nothing else catches: the docs must not describe a span that + no longer exists. + """ + missing = sorted( + str(span) + for span in reg.SPANS + if not any(reg.span_for(name) is span for name in emitted["names"]) + ) + assert not missing, ( + f"these spans are documented but never emitted by the workload: {missing}. " + "Either the instrumentation was removed, or exercise() no longer reaches it." + ) + + +@pytest.mark.asyncio +async def test_every_registered_attribute_is_emitted(emitted): + """ + Every registered attribute, optional or not, must actually be set at least + once by the workload. + + `optional` describes whether a reader should expect it on every span, not + whether the code still sets it - so an attribute deleted from the code but + left in the docs has to fail here even when it is marked optional. If a + new attribute only appears in some rare case, extend exercise() to reach + that case. + """ + by_span = _keys_by_span(emitted["pairs"]) + missing = [] + for span in reg.SPANS: + emitted_keys = by_span.get(str(span), set()) + for attribute in span.attributes: + if attribute not in emitted_keys: + missing.append(f"{span} -> {attribute}") + assert not missing, ( + "these attributes are documented but never emitted by the workload: " + + ", ".join(sorted(missing)) + ) + + +def test_registry_has_no_duplicate_names(): + assert len(set(reg.SPANS)) == len(reg.SPANS) + for span in reg.SPANS: + assert len(set(span.attributes)) == len( + span.attributes + ), f"{span} lists an attribute twice" + + +def test_registry_entries_are_documented(): + "Every entry carries a description - the docs are generated from these." + for span in reg.SPANS: + assert span.description.strip(), f"{span} has no description" + for attribute in span.attributes: + assert attribute.description.strip(), f"{span} -> {attribute} has none" + + +def test_registry_entries_are_usable_as_plain_strings(): + "The str subclassing is the whole reason call sites need no wrapper API." + assert isinstance(reg.DB_QUERY, str) + assert isinstance(reg.DB_NAMESPACE, str) + assert reg.DB_QUERY == "db.query" + assert reg.DB_NAMESPACE == "db.namespace" + assert f"{reg.DB_QUERY}.execute" == "db.query.execute" + + +def test_span_and_attribute_lookup(): + assert reg.span_for("db.query") is reg.DB_QUERY + assert reg.span_for("datasette.startup") is reg.STARTUP + assert reg.span_for("not.a.datasette.span") is None + assert reg.attribute_allowed(reg.DB_QUERY, "db.namespace") + assert not reg.attribute_allowed(reg.DB_QUERY, "db.namespace.extra") + assert not reg.attribute_allowed(reg.DB_QUERY, "datasette.isolated_connection") + assert not reg.attribute_allowed(None, "db.namespace") + + +def test_prefix_span_lookup(): + """ + `prefix=True` matching, exercised directly. + + Phase 1 registers no prefix spans, so without this the branch in + `span_for()` would be untested code that the conformance tests silently + never reach. + """ + hook = reg.SpanName("datasette.hook.", "A hypothetical span family", prefix=True) + original = reg.SPANS + reg.SPANS = original + (hook,) + try: + assert reg.span_for("datasette.hook.render_cell") is hook + assert reg.span_for("datasette.hook.anything") is hook + assert reg.span_for("datasette.hookish") is None + assert reg.span_for("db.query") is reg.DB_QUERY + finally: + reg.SPANS = original From b74231c190ce4ca71be5ce7f69513c63d30625fe Mon Sep 17 00:00:00 2001 From: Alex Garcia Date: Thu, 30 Jul 2026 18:58:53 -0700 Subject: [PATCH 09/10] Stop marking a deliberately-short query budget as a span error Datasette has a family of callers that run a query under a tiny time limit and treat "did not finish" as a usable answer. table_counts() is the loudest: the homepage counts every table with a 10ms budget and stores None for the ones that blow it. The QueryInterrupted handler on the db.query span was unconditional, so on a two-table database that produced four ERROR spans - two db.query and two db.query.execute - on every homepage hit. Measured on a 30MB two-table database: 4 red spans before, 0 after. Honouring log_sql_errors here would have silenced none of it. Only the three ArrayFacet json_type() probes pass log_sql_errors=False, and they are not the queries that time out; table_counts() and ColumnFacet.suggest both leave it at its True default. The signal that does separate the two cases is the budget itself: a caller asking for less time than sql_time_limit_ms is saying the query may not finish. Keying off that needs no new API and no changes outside database.py. A query that runs out the instance-wide limit is still an error. datasette.interrupted is still set in every case - it is the signal worth having, and only the ERROR status becomes conditional. Its registry description said the status is "also set to ERROR" full stop, which is now wrong, and that string is published in docs/internals.rst. The inner db.query.execute span carried the same bug through set_status_on_exception=log_sql_errors, so its exception handling is now explicit, matching the db.query span above it. The context manager's flags apply to every exception type alike and this span has to tell two apart. test_query_interrupted_sets_error_status forced its timeout with ?_timelimit=5, which is exactly the signal now reclassified as expected. It now forces one via sql_time_limit_ms so it still tests what it was written to test. Also documents, at the copy_context() sites, that context propagation carries Datasette's non-OTel ContextVars into worker threads too. Verified harmless: nothing reads _skip_permission_checks, _permission_check_cache or _in_datasette_client off the event loop, and Context.run() restores the thread's previous context on return, so no value can reach the next task on the shared pool. Co-Authored-By: Claude Opus 5 --- datasette/database.py | 116 +++++++++++++++++++++--------- datasette/telemetry_registry.py | 6 +- docs/internals.rst | 2 +- tests/test_telemetry.py | 120 +++++++++++++++++++++++++++++--- 4 files changed, 201 insertions(+), 43 deletions(-) diff --git a/datasette/database.py b/datasette/database.py index 22984d6f..7114a691 100644 --- a/datasette/database.py +++ b/datasette/database.py @@ -394,6 +394,9 @@ class Database: # Context raises "RuntimeError: cannot enter context ... already # entered". This propagates the caller's otel context (e.g. the # enclosing db.query span) onto the worker thread. + # + # It also propagates every *other* ContextVar - see the note in + # execute_fn() for why that is safe. ctx = contextvars.copy_context() return await asyncio.get_running_loop().run_in_executor( self.ds.executor, ctx.run, _run @@ -695,6 +698,22 @@ class Database: # Context raises "RuntimeError: cannot enter context ... # already entered". This propagates the caller's otel context # (e.g. the enclosing db.query span) onto the worker thread. + # + # copy_context() is not selective: it also carries Datasette's own + # ContextVars - _skip_permission_checks and _permission_check_cache + # (datasette/permissions.py), _in_datasette_client (app.py) and, + # until the hand-rolled tracer goes, trace_task_id (tracer.py) - + # into worker threads, where they previously took their defaults. + # That is safe, for two reasons. Nothing reads them on a worker + # thread: the permission code that reads the first two is async and + # only ever runs on the event loop. And Context.run() restores the + # thread's previous context when the callable returns, so a value + # cannot outlive the submit that carried it and reach the next task + # on this shared pool - "skip permission checks" in particular can + # never bleed from one request into another's query. Where a value + # would be read - a plugin calling datasette.in_client() or trace() + # from inside an execute_fn callable - seeing the submitting + # request's value is the more accurate answer, not a leak. ctx = contextvars.copy_context() future = self.ds.executor.submit(ctx.run, in_thread) self._pending_execute_futures.add(future) @@ -722,7 +741,19 @@ class Database: self._check_not_closed() page_size = page_size or self.ds.page_size time_limit_ms = self.ds.sql_time_limit_ms - if custom_time_limit and custom_time_limit < time_limit_ms: + # A caller that hands in a budget shorter than the instance-wide + # sql_time_limit_ms is saying "this may not finish, and that is an + # answer I can use" - and every such caller in core does treat the + # timeout as normal: table_counts() stores None per table, facet + # suggestion moves on to the next column, autocomplete falls back to a + # prefix query. Those timeouts are therefore not span errors. Without + # this, the homepage alone emits one red span per table (it counts + # every table under a 10ms budget) on every single hit. + # + # A query that runs out the instance-wide limit is a different event - + # nobody asked for a short budget, so it stays an error. + timeout_expected = bool(custom_time_limit) and custom_time_limit < time_limit_ms + if timeout_expected: time_limit_ms = custom_time_limit def sql_operation_in_thread(conn): @@ -733,38 +764,53 @@ class Database: # databases) - so it parents correctly to the enclosing # db.query span despite running on a different thread. # - # Callers passing log_sql_errors=False are probing and treat a - # failure as an expected answer - see the matching handling on the - # db.query span in execute(). Without this, facet suggestion marks - # two spans per text column as failed on every table page. + # Exception handling is explicit rather than left to the context + # manager's flags, which apply to every exception type alike. This + # span needs to tell two apart: an expected timeout is never an + # error, while a genuine SQL failure is one unless the caller + # passed log_sql_errors=False, meaning it was probing and treats + # failure as an expected answer. Without the latter, facet + # suggestion marks two spans per text column as failed on every + # table page; without the former, so does every homepage hit. with tracer.start_as_current_span( DB_QUERY_EXECUTE, - record_exception=log_sql_errors, - set_status_on_exception=log_sql_errors, - ): - with sqlite_timelimit(conn, time_limit_ms): - try: - cursor = conn.cursor() - cursor.execute(sql, params if params is not None else {}) - max_returned_rows = self.ds.max_returned_rows - if max_returned_rows == page_size: - max_returned_rows += 1 - if max_returned_rows and truncate: - rows = cursor.fetchmany(max_returned_rows + 1) - truncated = len(rows) > max_returned_rows - rows = rows[:max_returned_rows] - else: - rows = cursor.fetchall() - truncated = False - except (sqlite3.OperationalError, sqlite3.DatabaseError) as e: - if e.args == ("interrupted",): - raise QueryInterrupted(e, sql, params) - if log_sql_errors: - sys.stderr.write( - f"ERROR: conn={conn}, sql = {sql!r}, params = {params}: {e}\n" - ) - sys.stderr.flush() - raise + record_exception=False, + set_status_on_exception=False, + ) as execute_span: + try: + with sqlite_timelimit(conn, time_limit_ms): + try: + cursor = conn.cursor() + cursor.execute(sql, params if params is not None else {}) + max_returned_rows = self.ds.max_returned_rows + if max_returned_rows == page_size: + max_returned_rows += 1 + if max_returned_rows and truncate: + rows = cursor.fetchmany(max_returned_rows + 1) + truncated = len(rows) > max_returned_rows + rows = rows[:max_returned_rows] + else: + rows = cursor.fetchall() + truncated = False + except (sqlite3.OperationalError, sqlite3.DatabaseError) as e: + if e.args == ("interrupted",): + raise QueryInterrupted(e, sql, params) + if log_sql_errors: + sys.stderr.write( + f"ERROR: conn={conn}, sql = {sql!r}, params = {params}: {e}\n" + ) + sys.stderr.flush() + raise + except QueryInterrupted as e: + if not timeout_expected: + execute_span.record_exception(e) + execute_span.set_status(Status(StatusCode.ERROR, str(e))) + raise + except Exception as e: + if log_sql_errors: + execute_span.record_exception(e) + execute_span.set_status(Status(StatusCode.ERROR, str(e))) + raise if truncate: return Results(rows, truncated, cursor.description) @@ -801,9 +847,13 @@ class Database: try: results = await self.execute_fn(sql_operation_in_thread) except QueryInterrupted as e: - span.set_status(Status(StatusCode.ERROR, str(e))) + # datasette.interrupted is set either way - it is the + # signal worth having. Only the ERROR status is + # conditional; see the timeout_expected comment above. span.set_attribute(INTERRUPTED, True) - span.record_exception(e) + if not timeout_expected: + span.set_status(Status(StatusCode.ERROR, str(e))) + span.record_exception(e) raise except Exception as e: # log_sql_errors=False means the caller is probing and diff --git a/datasette/telemetry_registry.py b/datasette/telemetry_registry.py index 40b32cc5..2d0692a8 100644 --- a/datasette/telemetry_registry.py +++ b/datasette/telemetry_registry.py @@ -137,7 +137,11 @@ TRUNCATED = Attribute( INTERRUPTED = Attribute( "datasette.interrupted", "True if the query was cancelled for exceeding the time limit. The span " - "status is also set to ``ERROR``.", + "status is also set to ``ERROR``, unless the caller asked for a budget " + "shorter than :ref:`setting_sql_time_limit_ms` - as table counts, facet " + "suggestion and autocomplete all do - in which case running out of time " + "is an expected answer rather than a failure and the status is left " + "unset.", optional=True, ) SQL_ERROR_SUPPRESSED = Attribute( diff --git a/docs/internals.rst b/docs/internals.rst index b2215d9d..e7028637 100644 --- a/docs/internals.rst +++ b/docs/internals.rst @@ -2355,7 +2355,7 @@ Spans are ``SpanKind.INTERNAL`` unless a kind is listed below. Only ``db.query`` - ``datasette.time_limit_ms`` *(optional)* - The :ref:`setting_sql_time_limit_ms` value this query ran under. Set on reads, which are the queries that time limit applies to. - ``datasette.rows_returned`` *(optional)* - Number of rows a read returned. Set on the read path only, and only when the read succeeded. - ``datasette.truncated`` *(optional)* - True if the result was cut short by :ref:`setting_max_returned_rows`. - - ``datasette.interrupted`` *(optional)* - True if the query was cancelled for exceeding the time limit. The span status is also set to ``ERROR``. + - ``datasette.interrupted`` *(optional)* - True if the query was cancelled for exceeding the time limit. The span status is also set to ``ERROR``, unless the caller asked for a budget shorter than :ref:`setting_sql_time_limit_ms` - as table counts, facet suggestion and autocomplete all do - in which case running out of time is an expected answer rather than a failure and the status is left unset. - ``datasette.sql_error_suppressed`` *(optional)* - True when the query failed but the caller passed ``log_sql_errors=False``, meaning it was probing and treats failure as an expected answer. Facet suggestion does this against every column. - ``datasette.executescript`` *(optional)* - True for ``execute_write_script()``, which runs multiple statements. - ``datasette.executemany`` *(optional)* - True for ``execute_write_many()``, which runs one statement against many parameter sets. diff --git a/tests/test_telemetry.py b/tests/test_telemetry.py index 0a6d0192..0022883a 100644 --- a/tests/test_telemetry.py +++ b/tests/test_telemetry.py @@ -12,7 +12,7 @@ from opentelemetry import trace as otel_trace from opentelemetry.trace import SpanKind, StatusCode from datasette.app import Datasette -from datasette.database import Database +from datasette.database import Database, QueryInterrupted from datasette.telemetry import ( MAX_SQL_LENGTH, SCHEMA_URL, @@ -26,6 +26,15 @@ SECRET_PARAM_VALUE = "SUPER_SECRET_PARAM_VALUE_XYZ_123" INVALID_SQL = "select this_is_not_valid_sql from nowhere" +# Bounded so a broken time limit fails the test instead of hanging it, but far +# too long to finish inside any of the millisecond budgets used below. +SLOW_SQL = """ +with recursive counter(x) as ( + select 1 union all select x + 1 from counter where x < 50000000 +) +select max(x) from counter +""" + def _db_query_spans(otel_spans): return [span for span in otel_spans.get_finished_spans() if span.name == "db.query"] @@ -212,14 +221,22 @@ async def test_no_span_attribute_ever_contains_a_parameter_value(ds_client, otel @pytest.mark.asyncio -async def test_query_interrupted_sets_error_status(ds_client, otel_spans): - response = await ds_client.get( - "/fixtures/-/query.json", - params={"sql": "select sleep(0.05)", "_timelimit": 5}, - ) - assert response.status_code == 400 +async def test_query_interrupted_sets_error_status(otel_spans): + """ + A query that runs out the instance-wide sql_time_limit_ms is an error. - spans = _db_query_spans(otel_spans) + This used to force the timeout with `?_timelimit=5`, but a caller-supplied + budget shorter than the instance limit is now the signal that the timeout + was expected - see test_expected_timeout_is_not_a_span_error - so the + timeout has to come from the setting for this to still test what it was + written to test. + """ + ds = Datasette(memory=True, settings={"sql_time_limit_ms": 20}) + db = ds.add_memory_database("t09_instance_limit_timeout") + with pytest.raises(QueryInterrupted): + await db.execute(SLOW_SQL) + + spans = _spans_for_namespace(otel_spans, "t09_instance_limit_timeout") assert spans span = spans[-1] assert span.status.status_code == StatusCode.ERROR @@ -228,6 +245,93 @@ async def test_query_interrupted_sets_error_status(ds_client, otel_spans): assert all(event.name == "exception" for event in span.events) +async def _expected_timeout_count_span(otel_spans, database_name): + """ + Drive the real table_counts() path into a timeout; return its db.query span. + + table_counts() is where the headline instance of this lives: the homepage + counts every table under a 10ms budget and stores None for any table that + does not finish in time. Before this was fixed, a two-table database + produced four ERROR spans - two db.query and two db.query.execute - on + every single homepage hit. + """ + db = Datasette(memory=True).add_memory_database(database_name) + await db.execute_write("create table big (id integer primary key, t text)") + await db.execute_write_many( + "insert into big (t) values (?)", [["x" * 50] for _ in range(11000)] + ) + # count_limit caps the scan at 10001 rows, and below 20ms sqlite_timelimit() + # runs its progress handler on every VM instruction, so 1ms is not a close + # call - a scan of that size takes single-digit milliseconds at best. + counts = await db.table_counts(1) + assert counts == { + "big": None + }, "the count did not actually time out, so the rest of this test is vacuous" + + spans = [ + span + for span in _spans_for_namespace(otel_spans, database_name) + if "count(*)" in span.attributes["db.query.text"] + ] + assert len(spans) == 1 + return spans[0] + + +@pytest.mark.asyncio +async def test_expected_timeout_is_not_a_span_error(otel_spans): + span = await _expected_timeout_count_span(otel_spans, "t09_expected_timeout") + # The useful signal survives; only the red status goes away. + assert span.attributes["datasette.interrupted"] is True + assert span.status.status_code != StatusCode.ERROR + assert not [event for event in span.events if event.name == "exception"] + + +@pytest.mark.asyncio +async def test_expected_timeout_does_not_error_the_inner_execute_span(otel_spans): + """ + The same fix has to reach db.query.execute, which sets its own status. + + Half of the original bug lived here: the inner span passed + set_status_on_exception=log_sql_errors, and table_counts() leaves + log_sql_errors at its True default, so it went ERROR too. + """ + span = await _expected_timeout_count_span(otel_spans, "t09_expected_timeout_inner") + children = _children_named(otel_spans, "db.query.execute", span.context) + assert len(children) == 1 + child = children[0] + assert child.status.status_code != StatusCode.ERROR + assert not [event for event in child.events if event.name == "exception"] + + +@pytest.mark.asyncio +async def test_unexpected_timeout_is_still_a_span_error(otel_spans): + """ + A custom_time_limit *above* sql_time_limit_ms is not a short budget. + + This is the half of the rule that stops the fix collapsing into "never + report timeouts": the caller asked for 5 seconds, the instance overruled it + at 20ms, and nobody expected that. + """ + ds = Datasette(memory=True, settings={"sql_time_limit_ms": 20}) + db = ds.add_memory_database("t09_custom_limit_ignored") + with pytest.raises(QueryInterrupted): + await db.execute(SLOW_SQL, custom_time_limit=5000) + + spans = _spans_for_namespace(otel_spans, "t09_custom_limit_ignored") + assert spans + span = spans[-1] + # Proves the caller's larger budget really was discarded - otherwise this + # would be asserting on a query that ran under a 5s limit. + assert span.attributes["datasette.time_limit_ms"] == 20 + assert span.attributes["datasette.interrupted"] is True + assert span.status.status_code == StatusCode.ERROR + assert any(event.name == "exception" for event in span.events) + + children = _children_named(otel_spans, "db.query.execute", span.context) + assert len(children) == 1 + assert children[0].status.status_code == StatusCode.ERROR + + @pytest.mark.asyncio async def test_unsuppressed_sql_error_is_a_span_error(ds_client, otel_spans): db = ds_client.ds.get_database("fixtures") From 1052dc5c7b676193d175810f8f97580aae6b4493 Mon Sep 17 00:00:00 2001 From: Alex Garcia Date: Thu, 30 Jul 2026 19:10:04 -0700 Subject: [PATCH 10/10] Document what the database-layer spans emit, and how to turn them on The span reference itself is generated from the registry, so this adds the prose the generated list cannot supply: how to actually see a span, what is deliberately never recorded, and where the instrumentation stops short. The "how to turn it on" part is the part people get wrong. Core installs no provider, so OTEL_TRACES_EXPORTER=console against a plain `datasette` process emits nothing at all - that variable is read by the SDK auto-configuration which only runs under `opentelemetry-instrument`. Documented as a warning because it reads like a bug when you hit it. Two more measured facts get the same treatment: the SDK's BatchSpanProcessor default schedule delay is 5000ms (checked, not assumed - `BatchSpanProcessor._default_schedule_delay_millis()` on opentelemetry-sdk 1.44), so nothing appears for five seconds; and without OTEL_SERVICE_NAME the default resource reports service.name=unknown_service. Privacy properties are stated positively rather than left implicit: SQL truncated at 2048 characters, parameter values never recorded, no actor identifiers, table names only from an explicit `table=` argument. The last of those is now documented on db.execute() itself, since it is public API. The limitations section claims only what was measured. An earlier draft said two traces per process are orphaned by the register_output_renderer and asgi_wrapper hooks; measuring it showed a default install emits zero spans from either, because Datasette queries no database there - it is a plugin that would produce the orphan. Corrected to say that. It also deliberately does NOT say an embedder must install its provider before Datasette's first span or get nothing. That claim is false: ProxyTracer._tracer returns the no-op tracer without caching it when no provider is set, so early spans are dropped and nothing is poisoned. The telemetry.py docstring said no-op spans "cost approximately nothing". The benchmark for this diff does not support a claim that strong - a table page emits ~58 spans - so it now states the measurement instead: median 9.80ms to 9.98ms across 15 runs, inside a 1.4ms run-to-run spread. Co-Authored-By: Claude Opus 5 --- datasette/telemetry.py | 10 +++++-- docs/changelog.rst | 10 +++++++ docs/internals.rst | 62 ++++++++++++++++++++++++++++++++++++++++-- 3 files changed, 78 insertions(+), 4 deletions(-) diff --git a/datasette/telemetry.py b/datasette/telemetry.py index 37195584..3a205cee 100644 --- a/datasette/telemetry.py +++ b/datasette/telemetry.py @@ -5,8 +5,14 @@ Core depends on `opentelemetry-api` only. It never creates a `TracerProvider`, never configures an exporter, and never touches sampling - that is the responsibility of whoever is running Datasette (an `opentelemetry-instrument` agent, a future plugin, or a test -harness). With no provider installed every span produced here is a -`NonRecordingSpan` and costs approximately nothing. +harness). + +With no provider installed every span produced here is a +`NonRecordingSpan`. That is not free - a table page emits ~58 spans - +but it is below what an end-to-end page benchmark can resolve: measured +across 15 runs of a 5,000-row table page, the median moved 9.80ms to +9.98ms while run-to-run spread was 1.4ms. Installing an SDK provider is +what costs something measurable. """ import re diff --git a/docs/changelog.rst b/docs/changelog.rst index 66a7caab..cec8e92d 100644 --- a/docs/changelog.rst +++ b/docs/changelog.rst @@ -4,6 +4,16 @@ Changelog ========= +.. _v_unreleased: + +Unreleased +---------- + +- Datasette's database layer now emits `OpenTelemetry `__ spans: one per query, covering the full round trip including time spent waiting for a SQL worker thread, plus separate child spans for the execution itself and for time spent in the write queue. Datasette core depends on ``opentelemetry-api`` only and never installs an SDK provider, an exporter or a sampler, so there is no effect and no measurable overhead unless tracing is switched on externally - normally with the standard ``opentelemetry-instrument`` agent. See :ref:`internals_telemetry`. (:issue:`1730`) +- :ref:`db.execute(sql, ..., table=None) ` has a new optional ``table=`` parameter, naming the table a query is about so it can be recorded on that query's OpenTelemetry span. It has no effect on query execution, and Datasette never derives it from the SQL. (:issue:`1730`) + +Nothing is removed by this change: the ``?_trace=1`` query string parameter, the ``trace_debug`` setting and the :ref:`internals_tracer` module all continue to work as before. + .. _v1_0_a38: 1.0a38 (2026-08-06) diff --git a/docs/internals.rst b/docs/internals.rst index e7028637..31666e0e 100644 --- a/docs/internals.rst +++ b/docs/internals.rst @@ -1963,6 +1963,9 @@ Executes a SQL query against the database and returns the resulting rows (see :r ``log_sql_errors`` - boolean Should any SQL errors be logged to the console in addition to being raised as an error? Defaults to ``True``. +``table`` - string + The name of the table this query is about, if the caller already knows it. This has no effect on how the query executes - it is recorded as the ``db.collection.name`` attribute on the :ref:`OpenTelemetry span ` for the query. Datasette never derives this from the SQL, so leave it unset for queries that do not have one obvious table. + .. _database_results: Results @@ -2318,16 +2321,47 @@ The ``Database`` class also provides properties and methods for introspecting th OpenTelemetry ============= -Datasette core depends on `opentelemetry-api `__ only. It never creates a ``TracerProvider``, never configures an exporter and never sets a sampler. With no OpenTelemetry SDK provider installed, every span described below is a no-op ``NonRecordingSpan`` - the overhead is close to zero and nothing is recorded or exported anywhere. +Datasette core depends on `opentelemetry-api `__ only. It never creates a ``TracerProvider``, never configures an exporter and never sets a sampler. With no OpenTelemetry SDK provider installed, every span described below is a no-op ``NonRecordingSpan``: nothing is recorded, nothing is exported, and the cost does not show up in page latency. Benchmarking a table page with and without this instrumentation, the median moved by less than the run-to-run variation of the benchmark itself. Turning tracing on is entirely an operational decision made outside of Datasette itself: run Datasette under the standard ``opentelemetry-instrument`` agent, or embed Datasette inside a host application that installs its own provider. Everything Datasette emits carries the instrumentation scope ``datasette``, versioned with the running Datasette version and declaring the `semantic conventions schema `__ its attribute names follow. +This is separate from, and does not replace, the built-in :ref:`internals_tracer` mechanism behind ``?_trace=1`` and the :ref:`setting_trace_debug` setting. Both continue to work exactly as before. + +.. _internals_telemetry_turning_on: + +Turning tracing on +------------------ + +Install an OpenTelemetry SDK, an exporter and the instrumentation agent, then launch Datasette through ``opentelemetry-instrument``: + +.. code-block:: bash + + pip install opentelemetry-distro opentelemetry-exporter-otlp + OTEL_SERVICE_NAME=datasette \ + OTEL_EXPORTER_OTLP_ENDPOINT=http://localhost:4317 \ + OTEL_METRICS_EXPORTER=none \ + OTEL_LOGS_EXPORTER=none \ + opentelemetry-instrument datasette mydb.db + +Point ``OTEL_EXPORTER_OTLP_ENDPOINT`` at whichever tracing backend you use. To print spans straight to the terminal instead, with no backend at all, drop that variable and set ``OTEL_TRACES_EXPORTER=console`` in its place. + +A few things catch people out the first time: + +.. warning:: + ``OTEL_TRACES_EXPORTER=console datasette mydb.db`` produces **nothing**. That environment variable is read by the OpenTelemetry SDK's auto-configuration, which only runs when the ``opentelemetry-instrument`` agent wraps the process. Datasette core installs no provider, so a plain ``datasette`` process emits nothing at all, whatever ``OTEL_`` variables are set. + +Spans do not appear immediately. The SDK's default ``BatchSpanProcessor`` flushes on a timer, every 5 seconds. Either wait, or stop the process - shutdown triggers a final flush - or set ``OTEL_BSP_SCHEDULE_DELAY=1000`` while you are experimenting. That last one is for demos, not for production. + +Always set ``OTEL_SERVICE_NAME``. Without it the SDK's default resource reports a ``service.name`` of ``unknown_service``, and your traces will be filed under that instead of under a name you can search for. + +Setting ``OTEL_METRICS_EXPORTER=none`` and ``OTEL_LOGS_EXPORTER=none`` is worth doing unless your backend accepts those signals too - ``opentelemetry-distro`` defaults every signal to OTLP, and a traces-only backend will reject the other two noisily. Datasette itself emits no metrics and no logs through OpenTelemetry. + Span reference -------------- -Attribute names use the ``datasette.*`` prefix for Datasette-specific data, alongside standard OpenTelemetry attributes such as ``db.system``. +Datasette emits five spans. Four of them describe the database layer - one per query, one for the work that query does inside a SQL worker thread, and two more for the write queue - and the fifth covers startup. Attribute names use the ``datasette.*`` prefix for Datasette-specific data, alongside standard OpenTelemetry attributes such as ``db.system``. This reference is generated from ``datasette/telemetry_registry.py``, the single source of truth for every span and attribute Datasette emits. A conformance test makes real requests and compares what is actually emitted against that registry in both directions, so nothing here is hand-maintained and nothing can silently drift out of date. @@ -2385,6 +2419,30 @@ Spans are ``SpanKind.INTERNAL`` unless a kind is listed below. Only ``db.query`` .. [[[end]]] +.. _internals_telemetry_privacy: + +Privacy and safety +------------------ + +Spans leave your infrastructure whenever you configure an exporter, so what goes into them is a security decision. Datasette's rules are: + +- **SQL text is truncated to 2048 characters.** On a public instance the SQL is supplied by visitors and is unbounded in length, so ``db.query.text`` is cut off - with a ``…[truncated]`` marker - rather than allowed to set the size of a span. +- **SQL parameter values are never recorded.** Only ``datasette.param_count``, a count. Parameter values are the part of a query most likely to hold something sensitive, and separating them from the SQL is the reason bound parameters exist. +- **No actor identifiers are recorded.** No actor ID, no actor JSON, no client IP address. Nothing on a span identifies who made the request. +- **Table names come only from an explicit** ``table=`` **argument.** ``db.collection.name`` is set by callers that already know which table they are working with, and is never derived from the SQL. Deriving it would mean parsing, and on an instance where visitors can create tables the set of possible values has no ceiling. + +The SQL itself, though, *is* recorded, and on a public instance that means anything a visitor types into the query editor or passes as ``?sql=`` will be exported along with the span. That is the trade-off tracing a query engine makes. + +.. _internals_telemetry_limitations: + +Known limitations +----------------- + +- **Datasette does not create a span for the HTTP request itself.** Every span listed above is therefore a root span unless something above Datasette - an ASGI instrumentation layer, or the web framework embedding it - has already started one for the request, in which case Datasette's spans nest underneath it correctly. +- **Two plugin hooks run outside the** ``datasette.startup`` **span.** ``register_output_renderer`` is dispatched from ``Datasette.__init__()`` and ``asgi_wrapper`` from ``Datasette.app()``, both of which happen before ``invoke_startup()``. Datasette itself queries no database in either, so a default install emits nothing there - but a plugin that does will produce a root trace. Covering these would mean holding a span open across object construction, which is worse than the orphan. +- ``db.operation.name`` **reports** ``WITH`` **for a statement that opens with a common table expression**, rather than the operation inside it, and a substantial share of Datasette's own reads take that form. The attribute is a leading-keyword match against a fixed allowlist, deliberately not a parse. +- **Spans emitted before a provider is installed are not recorded.** If you are embedding Datasette in a host application, install your ``TracerProvider`` before serving traffic. This is ordinary OpenTelemetry behaviour rather than anything Datasette controls; nothing is permanently affected, those particular spans are simply dropped. + .. _internals_csrf: CSRF protection