Add OTLP payload byte-size chunking to Agent365Exporter

libraries/microsoft-agents-a365-observability-core/microsoft_agents_a365/observability/core/exporters/agent365_exporter.py

@@ -18,7 +18,10 @@
 from opentelemetry.trace import StatusCode
 
 from .utils import (
+    DEFAULT_MAX_PAYLOAD_BYTES,
     build_export_url,
+    chunk_by_size,
+    estimate_span_bytes,
     get_validated_domain_override,
     hex_span_id,
     hex_trace_id,
@@ -56,6 +59,7 @@ def __init__(
         token_resolver: Callable[[str, str], str | None],
         cluster_category: str = "prod",
         use_s2s_endpoint: bool = False,
+        max_payload_bytes: int = DEFAULT_MAX_PAYLOAD_BYTES,
     ):
         if token_resolver is None:
             raise ValueError("token_resolver must be provided.")
@@ -65,6 +69,7 @@ def __init__(
         self._token_resolver = token_resolver
         self._cluster_category = cluster_category
         self._use_s2s_endpoint = use_s2s_endpoint
+        self._max_payload_bytes = max_payload_bytes
         # Read domain override once at initialization
         self._domain_override = get_validated_domain_override()
 
@@ -89,8 +94,20 @@ def export(self, spans: Sequence[ReadableSpan]) -> SpanExportResult:
 
             any_failure = False
             for (tenant_id, agent_id), activities in groups.items():
-                payload = self._build_export_request(activities)
-                body = json.dumps(payload, separators=(",", ":"), ensure_ascii=False)
+                # Map and truncate spans first, then chunk by estimated byte size
+                mapped_spans = self._map_and_truncate_spans(activities)
+                resource_attrs = self._get_resource_attributes(activities)
+                chunks = chunk_by_size(
+                    mapped_spans,
+                    lambda ms: estimate_span_bytes(ms[0]),
+                    self._max_payload_bytes,
+                )
+
+                if len(chunks) > 1:
+                    logger.info(
+                        f"Split {len(activities)} spans into {len(chunks)} chunks "
+                        f"for tenantId: {tenant_id}, agentId: {agent_id}"
+                    )
 
                 # Resolve endpoint: domain override > default URL
                 if self._domain_override:
@@ -128,11 +145,29 @@ def export(self, spans: Sequence[ReadableSpan]) -> SpanExportResult:
                     any_failure = True
                     continue
 
-                # Basic retry loop
-                ok = self._post_with_retries(url, body, headers)
+                # Send each chunk (all-or-nothing: fail group on first chunk failure)
+                group_failed = False
+                for i, chunk in enumerate(chunks):
+                    payload = self._build_envelope(chunk, resource_attrs)
+                    body = json.dumps(payload, separators=(",", ":"), ensure_ascii=False)
+                    body_bytes = len(body.encode("utf-8"))
+                    logger.info(
+                        f"Sending chunk {i + 1} of {len(chunks)} "
+                        f"({len(chunk)} spans, {body_bytes} bytes)"
+                    )
 
-                if not ok:
-                    any_failure = True
+                    ok = self._post_with_retries(url, body, headers)
+                    if not ok:
+                        logger.error(
+                            f"Chunk {i + 1} of {len(chunks)} failed for "
+                            f"tenant {tenant_id}, agent {agent_id}"
+                        )
+                        any_failure = True
+                        group_failed = True
+                        break
+
+                if group_failed:
+                    continue
 
             return SpanExportResult.FAILURE if any_failure else SpanExportResult.SUCCESS
 
@@ -231,32 +266,47 @@ def _post_with_retries(self, url: str, body: str, headers: dict[str, str]) -> bo
 
     # ------------- Payload mapping ------------------
 
-    def _build_export_request(self, spans: Sequence[ReadableSpan]) -> dict[str, Any]:
-        # Group by instrumentation scope (name, version)
-        scope_map: dict[tuple[str, str | None], list[dict[str, Any]]] = {}
+    def _map_and_truncate_spans(
+        self, spans: Sequence[ReadableSpan]
+    ) -> list[tuple[dict[str, Any], str, str | None]]:
+        """Map ReadableSpans to OTLP dicts and apply per-span truncation.
 
+        Returns a list of (mapped_span, scope_name, scope_version) tuples so
+        that envelope grouping by instrumentation scope can be performed
+        efficiently after byte-size chunking.
+        """
+        result: list[tuple[dict[str, Any], str, str | None]] = []
         for sp in spans:
             scope = sp.instrumentation_scope
-            scope_key = (scope.name, scope.version)
-            scope_map.setdefault(scope_key, []).append(self._map_span(sp))
+            scope_name = scope.name if scope is not None else "unknown"
+            scope_version = scope.version if scope is not None else None
+            result.append((self._map_span(sp), scope_name, scope_version))
+        return result
 
-        scope_spans: list[dict[str, Any]] = []
-        for (name, version), mapped_spans in scope_map.items():
-            scope_spans.append(
-                {
-                    "scope": {
-                        "name": name,
-                        "version": version,
-                    },
-                    "spans": mapped_spans,
-                }
-            )
-
-        # Resource attributes (from the first span – all spans in a batch usually share resource)
-        # If you need to merge across spans, adapt accordingly.
-        resource_attrs = {}
+    @staticmethod
+    def _get_resource_attributes(spans: Sequence[ReadableSpan]) -> dict[str, Any]:
+        """Extract resource attributes from the first span in the batch."""
         if spans:
-            resource_attrs = dict(getattr(spans[0].resource, "attributes", {}) or {})
+            return dict(getattr(spans[0].resource, "attributes", {}) or {})
+        return {}
+
+    def _build_envelope(
+        self,
+        mapped_spans: Sequence[tuple[dict[str, Any], str, str | None]],
+        resource_attrs: dict[str, Any],
+    ) -> dict[str, Any]:
+        """Build an OTLP export request envelope from pre-mapped spans."""
+        scope_map: dict[tuple[str, str | None], list[dict[str, Any]]] = {}
+        for mapped_span, scope_name, scope_version in mapped_spans:
+            scope_map.setdefault((scope_name, scope_version), []).append(mapped_span)
+
+        scope_spans: list[dict[str, Any]] = [
+            {
+                "scope": {"name": name, "version": version},
+                "spans": spans,
+            }
+            for (name, version), spans in scope_map.items()
+        ]
 
         return {
             "resourceSpans": [

libraries/microsoft-agents-a365-observability-core/microsoft_agents_a365/observability/core/exporters/utils.py

@@ -6,8 +6,8 @@
 import json
 import logging
 import os
-from collections.abc import Sequence
-from typing import Any
+from collections.abc import Callable, Sequence
+from typing import Any, TypeVar
 from urllib.parse import urlparse
 
 from opentelemetry.sdk.trace import ReadableSpan
@@ -262,3 +262,119 @@ def is_agent365_exporter_enabled() -> bool:
     # Check environment variable
     enable_exporter = os.getenv(ENABLE_A365_OBSERVABILITY_EXPORTER, "").lower()
     return (enable_exporter) in ("true", "1", "yes", "on")
+
+
+# ---------------------------------------------------------------------------
+# Span size estimation and byte-level chunking
+# ---------------------------------------------------------------------------
+
+# Default upper bound on HTTP request body size in bytes. Provides ~100 KB
+# headroom under the A365 1 MB server limit for estimator error and JSON/
+# envelope overhead (e.g. resource attributes and scope wrappers).
+DEFAULT_MAX_PAYLOAD_BYTES = 900_000
+
+# Overhead constant for OTLP JSON span fixed fields (traceId, spanId,
+# parentSpanId, kind, timestamps, status, scope wrapper, etc.). Intentionally
+# generous to account for serializer variance.
+_SPAN_BASE_OVERHEAD = 2000
+
+# Overhead per attribute in OTLP JSON format. Covers key/value JSON wrapping.
+_ATTR_OVERHEAD = 80
+
+# Overhead per event in OTLP JSON.
+_EVENT_OVERHEAD = 200
+
+
+def _utf8_len(s: str) -> int:
+    return len(s.encode("utf-8"))
+
+
+def estimate_value_bytes(value: Any) -> int:
+    """Estimate the serialized byte size of a single attribute value in OTLP/HTTP JSON."""
+    if isinstance(value, str):
+        return 40 + _utf8_len(value)
+    # bool is a subclass of int; check before sequence/list handling below
+    if isinstance(value, bool):
+        return 40
+    if isinstance(value, (list, tuple)):
+        if len(value) == 0:
+            return 60
+        first = value[0]
+        if isinstance(first, str):
+            total = 60
+            for s in value:
+                total += 40 + _utf8_len(str(s))
+            return total
+        return 60 + 50 * len(value)
+    return 40  # int/float/None/other
+
+
+def estimate_span_bytes(span: dict[str, Any]) -> int:
+    """Heuristic estimator for the serialized size of an OTLP span in HTTP JSON.
+
+    Uses generous constants tuned to over-estimate by ~25-50%, providing
+    headroom for JSON serializer variance (whitespace, enum representation,
+    integer-as-string).
+    """
+    total = _SPAN_BASE_OVERHEAD
+    name = span.get("name")
+    if isinstance(name, str):
+        total += _utf8_len(name)
+
+    attributes = span.get("attributes")
+    if attributes:
+        for key, value in attributes.items():
+            total += _ATTR_OVERHEAD
+            total += _utf8_len(str(key))
+            total += estimate_value_bytes(value)
+
+    events = span.get("events")
+    if events:
+        for ev in events:
+            total += _EVENT_OVERHEAD
+            ev_name = ev.get("name") if isinstance(ev, dict) else None
+            if isinstance(ev_name, str):
+                total += _utf8_len(ev_name)
+            ev_attrs = ev.get("attributes") if isinstance(ev, dict) else None
+            if ev_attrs:
+                for key, value in ev_attrs.items():
+                    total += _ATTR_OVERHEAD
+                    total += _utf8_len(str(key))
+                    total += estimate_value_bytes(value)
+    return total
+
+
+T = TypeVar("T")
+
+
+def chunk_by_size(
+    items: Sequence[T],
+    get_size: Callable[[T], int],
+    max_chunk_bytes: int,
+) -> list[list[T]]:
+    """Split items into sub-batches whose cumulative estimated size stays under ``max_chunk_bytes``.
+
+    Invariants:
+    - Input order is preserved across chunks.
+    - Empty input produces empty output.
+    - A single item whose size exceeds ``max_chunk_bytes`` forms its own
+      single-item chunk (never silently dropped).
+    - No chunk is ever empty.
+    """
+    chunks: list[list[T]] = []
+    current: list[T] = []
+    current_bytes = 0
+
+    for item in items:
+        item_bytes = get_size(item)
+        if current and current_bytes + item_bytes > max_chunk_bytes:
+            chunks.append(current)
+            current = []
+            current_bytes = 0
+        current.append(item)
+        current_bytes += item_bytes
+
+    if current:
+        chunks.append(current)
+
+    return chunks