Abstract out OTLP http exporter logic

CHANGELOG.md

@@ -12,6 +12,8 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ## Unreleased
 
+- `opentelemetry-exporter-otlp-proto-http`: refactor shared HTTP exporter logic into `_common` module, introduce `OTLPHttpClient` base class and `_SignalConfig` dataclass, eliminating duplicated `__init__`, `_export`, retry loop, and `shutdown` across the trace, log, and metric exporters.
+  ([#2990](https://github.com/open-telemetry/opentelemetry-python/pull/5160))
 - Apply fixes for `UP` ruff rule
   ([#5133](https://github.com/open-telemetry/opentelemetry-python/pull/5133))
 - Switch to SPDX license headers and add CI enforcement

exporter/opentelemetry-exporter-otlp-proto-http/src/opentelemetry/exporter/otlp/proto/http/_common/__init__.py

@@ -1,15 +1,56 @@
 # Copyright The OpenTelemetry Authors
 # SPDX-License-Identifier: Apache-2.0
 
+import gzip
+import logging
+import random
+import threading
+import zlib
+from dataclasses import dataclass
+from io import BytesIO
 from os import environ
+from time import time
+from typing import Any, Dict, Optional
+from urllib.parse import urlparse
 from typing import Literal
 
 import requests
+from requests.exceptions import ConnectionError
 
+from opentelemetry.exporter.otlp.proto.common._exporter_metrics import (
+    ExporterMetrics,
+)
+from opentelemetry.exporter.otlp.proto.http import (
+    _OTLP_HTTP_HEADERS,
+    Compression,
+)
+from opentelemetry.metrics import MeterProvider
 from opentelemetry.sdk.environment_variables import (
     _OTEL_PYTHON_EXPORTER_OTLP_HTTP_CREDENTIAL_PROVIDER,
+    OTEL_EXPORTER_OTLP_CERTIFICATE,
+    OTEL_EXPORTER_OTLP_CLIENT_CERTIFICATE,
+    OTEL_EXPORTER_OTLP_CLIENT_KEY,
+    OTEL_EXPORTER_OTLP_COMPRESSION,
+    OTEL_EXPORTER_OTLP_ENDPOINT,
+    OTEL_EXPORTER_OTLP_HEADERS,
+    OTEL_EXPORTER_OTLP_TIMEOUT,
+)
+from opentelemetry.semconv._incubating.attributes.otel_attributes import (
+    OtelComponentTypeValues,
+)
+from opentelemetry.semconv.attributes.http_attributes import (
+    HTTP_RESPONSE_STATUS_CODE,
 )
 from opentelemetry.util._importlib_metadata import entry_points
+from opentelemetry.util.re import parse_env_headers
+
+_logger = logging.getLogger(__name__)
+
+_MAX_RETRIES = 6
+
+DEFAULT_COMPRESSION = Compression.NoCompression
+DEFAULT_ENDPOINT = "http://localhost:4318/"
+DEFAULT_TIMEOUT = 10  # in seconds
 
 
 def _is_retryable(resp: requests.Response) -> bool:
@@ -53,3 +94,243 @@ def _load_session_from_envvar(
                 f" must be of type `requests.Session`."
             )
     return None
+
+
+def _compression_from_env(compression_envvar: str) -> Compression:
+    compression = (
+        environ.get(
+            compression_envvar,
+            environ.get(OTEL_EXPORTER_OTLP_COMPRESSION, "none"),
+        )
+        .lower()
+        .strip()
+    )
+    return Compression(compression)
+
+
+def _append_signal_path(endpoint: str, signal_path: str) -> str:
+    if endpoint.endswith("/"):
+        return endpoint + signal_path
+    return endpoint + f"/{signal_path}"
+
+
+@dataclass(frozen=True)
+class _SignalConfig:
+    endpoint_envvar: str
+    certificate_envvar: str
+    client_key_envvar: str
+    client_certificate_envvar: str
+    headers_envvar: str
+    timeout_envvar: str
+    compression_envvar: str
+    credential_envvar: Literal[
+        "OTEL_PYTHON_EXPORTER_OTLP_HTTP_LOGS_CREDENTIAL_PROVIDER",
+        "OTEL_PYTHON_EXPORTER_OTLP_HTTP_TRACES_CREDENTIAL_PROVIDER",
+        "OTEL_PYTHON_EXPORTER_OTLP_HTTP_METRICS_CREDENTIAL_PROVIDER",
+    ]
+    default_export_path: str
+    component_type: OtelComponentTypeValues
+    signal_name: Literal["traces", "metrics", "logs"]
+
+
+class OTLPHttpClient:
+    def __init__(
+        self,
+        endpoint: Optional[str],
+        certificate_file: Optional[str],
+        client_key_file: Optional[str],
+        client_certificate_file: Optional[str],
+        headers: Optional[Dict[str, str]],
+        timeout: Optional[float],
+        compression: Optional[Compression],
+        session: Optional[requests.Session],
+        meter_provider: Optional[MeterProvider],
+        signal_config: _SignalConfig,
+    ):
+        self._shutdown_in_progress = threading.Event()
+        self._endpoint = endpoint or environ.get(
+            signal_config.endpoint_envvar,
+            _append_signal_path(
+                environ.get(OTEL_EXPORTER_OTLP_ENDPOINT, DEFAULT_ENDPOINT),
+                signal_config.default_export_path,
+            ),
+        )
+        self._certificate_file = certificate_file or environ.get(
+            signal_config.certificate_envvar,
+            environ.get(OTEL_EXPORTER_OTLP_CERTIFICATE, True),
+        )
+        self._client_key_file = client_key_file or environ.get(
+            signal_config.client_key_envvar,
+            environ.get(OTEL_EXPORTER_OTLP_CLIENT_KEY, None),
+        )
+        self._client_certificate_file = client_certificate_file or environ.get(
+            signal_config.client_certificate_envvar,
+            environ.get(OTEL_EXPORTER_OTLP_CLIENT_CERTIFICATE, None),
+        )
+        self._client_cert = (
+            (self._client_certificate_file, self._client_key_file)
+            if self._client_certificate_file and self._client_key_file
+            else self._client_certificate_file
+        )
+        headers_string = environ.get(
+            signal_config.headers_envvar,
+            environ.get(OTEL_EXPORTER_OTLP_HEADERS, ""),
+        )
+        self._headers = headers or parse_env_headers(
+            headers_string, liberal=True
+        )
+        self._timeout = timeout or float(
+            environ.get(
+                signal_config.timeout_envvar,
+                environ.get(OTEL_EXPORTER_OTLP_TIMEOUT, DEFAULT_TIMEOUT),
+            )
+        )
+        self._compression = compression or _compression_from_env(
+            signal_config.compression_envvar
+        )
+        self._session = self._setup_session(
+            session, signal_config.credential_envvar
+        )
+        self._shutdown = False
+        self._metrics = ExporterMetrics(
+            signal_config.component_type,
+            signal_config.signal_name,
+            urlparse(self._endpoint),
+            meter_provider,
+        )
+
+    def _setup_session(
+        self,
+        session: Optional[requests.Session],
+        cred_envvar: Literal[
+        "OTEL_PYTHON_EXPORTER_OTLP_HTTP_LOGS_CREDENTIAL_PROVIDER",
+        "OTEL_PYTHON_EXPORTER_OTLP_HTTP_TRACES_CREDENTIAL_PROVIDER",
+        "OTEL_PYTHON_EXPORTER_OTLP_HTTP_METRICS_CREDENTIAL_PROVIDER",
+    ],
+    ) -> requests.Session:
+        configured_session = (
+            session
+            or _load_session_from_envvar(cred_envvar)
+            or requests.Session()
+        )
+        configured_session.headers.update(self._headers)
+        configured_session.headers.update(_OTLP_HTTP_HEADERS)
+        # let users override our defaults
+        configured_session.headers.update(self._headers)
+        if self._compression is not Compression.NoCompression:
+            configured_session.headers.update(
+                {"Content-Encoding": self._compression.value}
+            )
+        return configured_session
+
+    def _export(
+        self, serialized_data: bytes, timeout_sec: float
+    ) -> requests.Response:
+        data = serialized_data
+        if self._compression == Compression.Gzip:
+            gzip_data = BytesIO()
+            with gzip.GzipFile(fileobj=gzip_data, mode="w") as gzip_stream:
+                gzip_stream.write(serialized_data)
+            data = gzip_data.getvalue()
+        elif self._compression == Compression.Deflate:
+            data = zlib.compress(serialized_data)
+
+        # By default, keep-alive is enabled in Session's request
+        # headers. Backends may choose to close the connection
+        # while a post happens which causes an unhandled
+        # exception. This try/except will retry the post on such exceptions
+        try:
+            resp = self._session.post(
+                url=self._endpoint,
+                data=data,
+                verify=self._certificate_file,
+                timeout=timeout_sec,
+                cert=self._client_cert,
+            )
+        except ConnectionError:
+            resp = self._session.post(
+                url=self._endpoint,
+                data=data,
+                verify=self._certificate_file,
+                timeout=timeout_sec,
+                cert=self._client_cert,
+            )
+        return resp
+
+    def _export_with_retries(
+        self,
+        serialized_data: bytes,
+        result: Any,
+        batch_name: str,
+    ) -> bool:
+        deadline_sec = time() + self._timeout
+        for retry_num in range(_MAX_RETRIES):
+            # multiplying by a random number between .8 and 1.2 introduces a +/20% jitter to each backoff.
+            backoff_seconds = 2**retry_num * random.uniform(0.8, 1.2)
+            export_error: Optional[Exception] = None
+            try:
+                if (
+                    resp := self._export(
+                        serialized_data, deadline_sec - time()
+                    )
+                ).ok:
+                    return True
+            except requests.exceptions.RequestException as error:
+                reason = error
+                export_error = error
+                retryable = isinstance(error, ConnectionError)
+                status_code = None
+            else:
+                reason = resp.reason
+                retryable = _is_retryable(resp)
+                status_code = resp.status_code
+
+            error_attrs = (
+                {HTTP_RESPONSE_STATUS_CODE: status_code}
+                if status_code is not None
+                else None
+            )
+
+            if not retryable:
+                _logger.error(
+                    "Failed to export %s batch code: %s, reason: %s",
+                    batch_name,
+                    status_code,
+                    reason,
+                )
+                result.error = export_error
+                result.error_attrs = error_attrs
+                return False
+
+            if (
+                retry_num + 1 == _MAX_RETRIES
+                or backoff_seconds > (deadline_sec - time())
+                or self._shutdown_in_progress.is_set()
+            ):
+                _logger.error(
+                    "Failed to export %s batch due to timeout, "
+                    "max retries or shutdown.",
+                    batch_name,
+                )
+                result.error = export_error
+                result.error_attrs = error_attrs
+                return False
+
+            _logger.warning(
+                "Transient error %s encountered while exporting %s batch, retrying in %.2fs.",
+                reason,
+                batch_name,
+                backoff_seconds,
+            )
+            if self._shutdown_in_progress.wait(backoff_seconds):
+                _logger.warning("Shutdown in progress, aborting retry.")
+                break
+        return False
+
+    def shutdown(self, timeout_millis: float = 30000, **kwargs) -> None:
+        if self._shutdown:
+            _logger.warning("Exporter already shutdown, ignoring call")
+            return
+        self._shutdown = True
+        self._shutdown_in_progress.set()
+        self._session.close()