feat(span-streaming): Add span batcher (#5398)

- Add a specialized `SpanBatcher` that batches and sends spans by trace
ID. Hook it up to the client.
- Add a skeleton of a new `StreamedSpan` class that will eventually
replace the current `Span` class.

Chipping away at #5317 to
transform it into reviewable and mergeable PRs.

Author: sentrivana

sentry_sdk/_span_batcher.py

@@ -0,0 +1,118 @@
+import threading
+from collections import defaultdict
+from datetime import datetime, timezone
+from typing import TYPE_CHECKING
+
+from sentry_sdk._batcher import Batcher
+from sentry_sdk.consts import SPANSTATUS
+from sentry_sdk.envelope import Envelope, Item, PayloadRef
+from sentry_sdk.utils import format_timestamp, serialize_attribute, safe_repr
+
+if TYPE_CHECKING:
+    from typing import Any, Callable, Optional
+    from sentry_sdk.traces import StreamedSpan
+    from sentry_sdk._types import SerializedAttributeValue
+
+
+class SpanBatcher(Batcher["StreamedSpan"]):
+    # TODO[span-first]: size-based flushes
+    # TODO[span-first]: adjust flush/drop defaults
+    MAX_BEFORE_FLUSH = 1000
+    MAX_BEFORE_DROP = 5000
+    FLUSH_WAIT_TIME = 5.0
+
+    TYPE = "span"
+    CONTENT_TYPE = "application/vnd.sentry.items.span.v2+json"
+
+    def __init__(
+        self,
+        capture_func: "Callable[[Envelope], None]",
+        record_lost_func: "Callable[..., None]",
+    ) -> None:
+        # Spans from different traces cannot be emitted in the same envelope
+        # since the envelope contains a shared trace header. That's why we bucket
+        # by trace_id, so that we can then send the buckets each in its own
+        # envelope.
+        # trace_id -> span buffer
+        self._span_buffer: dict[str, list["StreamedSpan"]] = defaultdict(list)
+        self._capture_func = capture_func
+        self._record_lost_func = record_lost_func
+        self._running = True
+        self._lock = threading.Lock()
+
+        self._flush_event: "threading.Event" = threading.Event()
+
+        self._flusher: "Optional[threading.Thread]" = None
+        self._flusher_pid: "Optional[int]" = None
+
+    def get_size(self) -> int:
+        # caller is responsible for locking before checking this
+        return sum(len(buffer) for buffer in self._span_buffer.values())
+
+    def add(self, span: "StreamedSpan") -> None:
+        if not self._ensure_thread() or self._flusher is None:
+            return None
+
+        with self._lock:
+            size = self.get_size()
+            if size >= self.MAX_BEFORE_DROP:
+                self._record_lost_func(
+                    reason="queue_overflow",
+                    data_category="span",
+                    quantity=1,
+                )
+                return None
+
+            self._span_buffer[span.trace_id].append(span)
+            if size + 1 >= self.MAX_BEFORE_FLUSH:
+                self._flush_event.set()
+
+    @staticmethod
+    def _to_transport_format(item: "StreamedSpan") -> "Any":
+        # TODO[span-first]
+        res: "dict[str, Any]" = {}
+        return res
+
+    def _flush(self) -> None:
+        with self._lock:
+            if len(self._span_buffer) == 0:
+                return None
+
+            envelopes = []
+            for trace_id, spans in self._span_buffer.items():
+                if spans:
+                    # TODO[span-first]
+                    # dsc = spans[0].dynamic_sampling_context()
+                    dsc = None
+
+                    envelope = Envelope(
+                        headers={
+                            "sent_at": format_timestamp(datetime.now(timezone.utc)),
+                            "trace": dsc,
+                        }
+                    )
+
+                    envelope.add_item(
+                        Item(
+                            type="span",
+                            content_type="application/vnd.sentry.items.span.v2+json",
+                            headers={
+                                "item_count": len(spans),
+                            },
+                            payload=PayloadRef(
+                                json={
+                                    "items": [
+                                        self._to_transport_format(span)
+                                        for span in spans
+                                    ]
+                                }
+                            ),
+                        )
+                    )
+
+                    envelopes.append(envelope)
+
+            self._span_buffer.clear()
+
+        for envelope in envelopes:
+            self._capture_func(envelope)

sentry_sdk/client.py

@@ -11,6 +11,7 @@
 import sentry_sdk
 from sentry_sdk._compat import PY37, check_uwsgi_thread_support
 from sentry_sdk._metrics_batcher import MetricsBatcher
+from sentry_sdk._span_batcher import SpanBatcher
 from sentry_sdk.utils import (
     AnnotatedValue,
     ContextVar,
@@ -31,6 +32,7 @@
 )
 from sentry_sdk.serializer import serialize
 from sentry_sdk.tracing import trace
+from sentry_sdk.tracing_utils import has_span_streaming_enabled
 from sentry_sdk.transport import BaseHttpTransport, make_transport
 from sentry_sdk.consts import (
     SPANDATA,
@@ -188,6 +190,7 @@ def __init__(self, options: "Optional[Dict[str, Any]]" = None) -> None:
         self.monitor: "Optional[Monitor]" = None
         self.log_batcher: "Optional[LogBatcher]" = None
         self.metrics_batcher: "Optional[MetricsBatcher]" = None
+        self.span_batcher: "Optional[SpanBatcher]" = None
         self.integrations: "dict[str, Integration]" = {}
 
     def __getstate__(self, *args: "Any", **kwargs: "Any") -> "Any":
@@ -399,6 +402,13 @@ def _record_lost_event(
                     record_lost_func=_record_lost_event,
                 )
 
+            self.span_batcher = None
+            if has_span_streaming_enabled(self.options):
+                self.span_batcher = SpanBatcher(
+                    capture_func=_capture_envelope,
+                    record_lost_func=_record_lost_event,
+                )
+
             max_request_body_size = ("always", "never", "small", "medium")
             if self.options["max_request_body_size"] not in max_request_body_size:
                 raise ValueError(
@@ -909,7 +919,10 @@ def capture_event(
         return return_value
 
     def _capture_telemetry(
-        self, telemetry: "Optional[Union[Log, Metric]]", ty: str, scope: "Scope"
+        self,
+        telemetry: "Optional[Union[Log, Metric]]",
+        ty: str,
+        scope: "Scope",
     ) -> None:
         # Capture attributes-based telemetry (logs, metrics, spansV2)
         if telemetry is None:
@@ -993,6 +1006,8 @@ def close(
                 self.log_batcher.kill()
             if self.metrics_batcher is not None:
                 self.metrics_batcher.kill()
+            if self.span_batcher is not None:
+                self.span_batcher.kill()
             if self.monitor:
                 self.monitor.kill()
             self.transport.kill()
@@ -1018,6 +1033,8 @@ def flush(
                 self.log_batcher.flush()
             if self.metrics_batcher is not None:
                 self.metrics_batcher.flush()
+            if self.span_batcher is not None:
+                self.span_batcher.flush()
             self.transport.flush(timeout=timeout, callback=callback)
 
     def __enter__(self) -> "_Client":

sentry_sdk/traces.py

@@ -0,0 +1,39 @@
+"""
+The API in this file is only meant to be used in span streaming mode.
+
+You can enable span streaming mode via
+sentry_sdk.init(_experiments={"trace_lifecycle": "stream"}).
+"""
+
+import uuid
+from typing import TYPE_CHECKING
+
+if TYPE_CHECKING:
+    from typing import Optional
+
+
+class StreamedSpan:
+    """
+    A span holds timing information of a block of code.
+
+    Spans can have multiple child spans thus forming a span tree.
+
+    This is the Span First span implementation. The original transaction-based
+    span implementation lives in tracing.Span.
+    """
+
+    __slots__ = ("_trace_id",)
+
+    def __init__(
+        self,
+        *,
+        trace_id: "Optional[str]" = None,
+    ):
+        self._trace_id = trace_id
+
+    @property
+    def trace_id(self) -> str:
+        if not self._trace_id:
+            self._trace_id = uuid.uuid4().hex
+
+        return self._trace_id