feat(http): add error handling for exporting (open-telemetry#4709)

* feat(http): add error handling for exporting

* feat(http_exporter): allow to run retry loop on connection errors

* feat(http): change error types that are caught

* refactor(http): introduce variables to unify logging

* feat(http_exporter): only retry on connection error

* test(http_exporter): add test case for connection errors while exporting

* refactor(http): simplify if statements

* docs(changelog): add changes

* fix(http_exporter): use correct class after rebase

* docs(changelog): update changelog

* refactor(http_exporter): add empty space in logs

* docs(tests): remove comments

* refactor(tests): simplify tests

* Update CHANGELOG.md

* Update exporter/opentelemetry-exporter-otlp-proto-http/src/opentelemetry/exporter/otlp/proto/http/_log_exporter/__init__.py

* Apply suggestions from code review

---------

Co-authored-by: Riccardo Magliocchetti <riccardo.magliocchetti@gmail.com>

Author: pafi-code

CHANGELOG.md

@@ -26,6 +26,8 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
   ([#4847](https://github.com/open-telemetry/opentelemetry-python/pull/4847))
 - Prevent possible endless recursion from happening in `SimpleLogRecordProcessor.on_emit`,
   ([#4799](https://github.com/open-telemetry/opentelemetry-python/pull/4799)) and ([#4867](https://github.com/open-telemetry/opentelemetry-python/pull/4867)).
+- `opentelemetry-exporter-otlp-proto-http`: fix retry logic and error handling for connection failures in trace, metric, and log exporters
+  ([#4709](https://github.com/open-telemetry/opentelemetry-python/pull/4709))
 
 ## Version 1.39.0/0.60b0 (2025-12-03)
 
@@ -93,7 +95,7 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 ([#4654](https://github.com/open-telemetry/opentelemetry-python/pull/4654)).
 - Fix type checking for built-in metric exporters
   ([#4820](https://github.com/open-telemetry/opentelemetry-python/pull/4820))
-  
+
 ## Version 1.38.0/0.59b0 (2025-10-16)
 
 - Add `rstcheck` to pre-commit to stop introducing invalid RST

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

@@ -186,26 +186,42 @@ def export(
         serialized_data = encode_logs(batch).SerializeToString()
         deadline_sec = time() + self._timeout
         for retry_num in range(_MAX_RETRYS):
-            resp = self._export(serialized_data, deadline_sec - time())
-            if resp.ok:
-                return LogRecordExportResult.SUCCESS
             # 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)
+            try:
+                resp = self._export(serialized_data, deadline_sec - time())
+                if resp.ok:
+                    return LogRecordExportResult.SUCCESS
+            except requests.exceptions.RequestException as error:
+                reason = error
+                retryable = isinstance(error, ConnectionError)
+                status_code = None
+            else:
+                reason = resp.reason
+                retryable = _is_retryable(resp)
+                status_code = resp.status_code
+
+            if not retryable:
+                _logger.error(
+                    "Failed to export logs batch code: %s, reason: %s",
+                    status_code,
+                    reason,
+                )
+                return LogRecordExportResult.FAILURE
+
             if (
-                not _is_retryable(resp)
-                or retry_num + 1 == _MAX_RETRYS
+                retry_num + 1 == _MAX_RETRYS
                 or backoff_seconds > (deadline_sec - time())
                 or self._shutdown
             ):
                 _logger.error(
-                    "Failed to export logs batch code: %s, reason: %s",
-                    resp.status_code,
-                    resp.text,
+                    "Failed to export logs batch due to timeout, "
+                    "max retries or shutdown."
                 )
                 return LogRecordExportResult.FAILURE
             _logger.warning(
                 "Transient error %s encountered while exporting logs batch, retrying in %.2fs.",
-                resp.reason,
+                reason,
                 backoff_seconds,
             )
             shutdown = self._shutdown_is_occuring.wait(backoff_seconds)

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

@@ -231,26 +231,41 @@ def export(
         serialized_data = encode_metrics(metrics_data).SerializeToString()
         deadline_sec = time() + self._timeout
         for retry_num in range(_MAX_RETRYS):
-            resp = self._export(serialized_data, deadline_sec - time())
-            if resp.ok:
-                return MetricExportResult.SUCCESS
             # 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)
+            try:
+                resp = self._export(serialized_data, deadline_sec - time())
+                if resp.ok:
+                    return MetricExportResult.SUCCESS
+            except requests.exceptions.RequestException as error:
+                reason = error
+                retryable = isinstance(error, ConnectionError)
+                status_code = None
+            else:
+                reason = resp.reason
+                retryable = _is_retryable(resp)
+                status_code = resp.status_code
+
+            if not retryable:
+                _logger.error(
+                    "Failed to export metrics batch code: %s, reason: %s",
+                    status_code,
+                    reason,
+                )
+                return MetricExportResult.FAILURE
             if (
-                not _is_retryable(resp)
-                or retry_num + 1 == _MAX_RETRYS
+                retry_num + 1 == _MAX_RETRYS
                 or backoff_seconds > (deadline_sec - time())
                 or self._shutdown
             ):
                 _logger.error(
-                    "Failed to export metrics batch code: %s, reason: %s",
-                    resp.status_code,
-                    resp.text,
+                    "Failed to export metrics batch due to timeout, "
+                    "max retries or shutdown."
                 )
                 return MetricExportResult.FAILURE
             _logger.warning(
                 "Transient error %s encountered while exporting metrics batch, retrying in %.2fs.",
-                resp.reason,
+                reason,
                 backoff_seconds,
             )
             shutdown = self._shutdown_in_progress.wait(backoff_seconds)

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

@@ -179,26 +179,42 @@ def export(self, spans: Sequence[ReadableSpan]) -> SpanExportResult:
         serialized_data = encode_spans(spans).SerializePartialToString()
         deadline_sec = time() + self._timeout
         for retry_num in range(_MAX_RETRYS):
-            resp = self._export(serialized_data, deadline_sec - time())
-            if resp.ok:
-                return SpanExportResult.SUCCESS
             # 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)
+            try:
+                resp = self._export(serialized_data, deadline_sec - time())
+                if resp.ok:
+                    return SpanExportResult.SUCCESS
+            except requests.exceptions.RequestException as error:
+                reason = error
+                retryable = isinstance(error, ConnectionError)
+                status_code = None
+            else:
+                reason = resp.reason
+                retryable = _is_retryable(resp)
+                status_code = resp.status_code
+
+            if not retryable:
+                _logger.error(
+                    "Failed to export span batch code: %s, reason: %s",
+                    status_code,
+                    reason,
+                )
+                return SpanExportResult.FAILURE
+
             if (
-                not _is_retryable(resp)
-                or retry_num + 1 == _MAX_RETRYS
+                retry_num + 1 == _MAX_RETRYS
                 or backoff_seconds > (deadline_sec - time())
                 or self._shutdown
             ):
                 _logger.error(
-                    "Failed to export span batch code: %s, reason: %s",
-                    resp.status_code,
-                    resp.text,
+                    "Failed to export span batch due to timeout, "
+                    "max retries or shutdown."
                 )
                 return SpanExportResult.FAILURE
             _logger.warning(
                 "Transient error %s encountered while exporting span batch, retrying in %.2fs.",
-                resp.reason,
+                reason,
                 backoff_seconds,
             )
             shutdown = self._shutdown_in_progress.wait(backoff_seconds)

exporter/opentelemetry-exporter-otlp-proto-http/tests/metrics/test_otlp_metrics_exporter.py

@@ -19,7 +19,9 @@
 from unittest import TestCase
 from unittest.mock import ANY, MagicMock, Mock, patch
 
+import requests
 from requests import Session
+from requests.exceptions import ConnectionError
 from requests.models import Response
 
 from opentelemetry.exporter.otlp.proto.common.metrics_encoder import (
@@ -555,6 +557,40 @@ def test_retry_timeout(self, mock_post):
                 warning.records[0].message,
             )
 
+    @patch.object(Session, "post")
+    def test_export_no_collector_available_retryable(self, mock_post):
+        exporter = OTLPMetricExporter(timeout=1.5)
+        msg = "Server not available."
+        mock_post.side_effect = ConnectionError(msg)
+        with self.assertLogs(level=WARNING) as warning:
+            self.assertEqual(
+                exporter.export(self.metrics["sum_int"]),
+                MetricExportResult.FAILURE,
+            )
+            # Check for greater 2 because the request is on each retry
+            # done twice at the moment.
+            self.assertGreater(mock_post.call_count, 2)
+            self.assertIn(
+                f"Transient error {msg} encountered while exporting metrics batch, retrying in",
+                warning.records[0].message,
+            )
+
+    @patch.object(Session, "post")
+    def test_export_no_collector_available(self, mock_post):
+        exporter = OTLPMetricExporter(timeout=1.5)
+
+        mock_post.side_effect = requests.exceptions.RequestException()
+        with self.assertLogs(level=WARNING) as warning:
+            self.assertEqual(
+                exporter.export(self.metrics["sum_int"]),
+                MetricExportResult.FAILURE,
+            )
+            self.assertEqual(mock_post.call_count, 1)
+            self.assertIn(
+                "Failed to export metrics batch code",
+                warning.records[0].message,
+            )
+
     @patch.object(Session, "post")
     def test_timeout_set_correctly(self, mock_post):
         resp = Response()

exporter/opentelemetry-exporter-otlp-proto-http/tests/test_proto_log_exporter.py

@@ -24,6 +24,7 @@
 import requests
 from google.protobuf.json_format import MessageToDict
 from requests import Session
+from requests.exceptions import ConnectionError
 from requests.models import Response
 
 from opentelemetry._logs import LogRecord, SeverityNumber
@@ -483,6 +484,40 @@ def test_retry_timeout(self, mock_post):
                 warning.records[0].message,
             )
 
+    @patch.object(Session, "post")
+    def test_export_no_collector_available_retryable(self, mock_post):
+        exporter = OTLPLogExporter(timeout=1.5)
+        msg = "Server not available."
+        mock_post.side_effect = ConnectionError(msg)
+        with self.assertLogs(level=WARNING) as warning:
+            self.assertEqual(
+                exporter.export(self._get_sdk_log_data()),
+                LogRecordExportResult.FAILURE,
+            )
+            # Check for greater 2 because the request is on each retry
+            # done twice at the moment.
+            self.assertGreater(mock_post.call_count, 2)
+            self.assertIn(
+                f"Transient error {msg} encountered while exporting logs batch, retrying in",
+                warning.records[0].message,
+            )
+
+    @patch.object(Session, "post")
+    def test_export_no_collector_available(self, mock_post):
+        exporter = OTLPLogExporter(timeout=1.5)
+
+        mock_post.side_effect = requests.exceptions.RequestException()
+        with self.assertLogs(level=WARNING) as warning:
+            self.assertEqual(
+                exporter.export(self._get_sdk_log_data()),
+                LogRecordExportResult.FAILURE,
+            )
+            self.assertEqual(mock_post.call_count, 1)
+            self.assertIn(
+                "Failed to export logs batch code",
+                warning.records[0].message,
+            )
+
     @patch.object(Session, "post")
     def test_timeout_set_correctly(self, mock_post):
         resp = Response()

exporter/opentelemetry-exporter-otlp-proto-http/tests/test_proto_span_exporter.py

@@ -20,6 +20,7 @@
 
 import requests
 from requests import Session
+from requests.exceptions import ConnectionError
 from requests.models import Response
 
 from opentelemetry.exporter.otlp.proto.http import Compression
@@ -303,6 +304,40 @@ def test_retry_timeout(self, mock_post):
                 warning.records[0].message,
             )
 
+    @patch.object(Session, "post")
+    def test_export_no_collector_available_retryable(self, mock_post):
+        exporter = OTLPSpanExporter(timeout=1.5)
+        msg = "Server not available."
+        mock_post.side_effect = ConnectionError(msg)
+        with self.assertLogs(level=WARNING) as warning:
+            self.assertEqual(
+                exporter.export([BASIC_SPAN]),
+                SpanExportResult.FAILURE,
+            )
+            # Check for greater 2 because the request is on each retry
+            # done twice at the moment.
+            self.assertGreater(mock_post.call_count, 2)
+            self.assertIn(
+                f"Transient error {msg} encountered while exporting span batch, retrying in",
+                warning.records[0].message,
+            )
+
+    @patch.object(Session, "post")
+    def test_export_no_collector_available(self, mock_post):
+        exporter = OTLPSpanExporter(timeout=1.5)
+
+        mock_post.side_effect = requests.exceptions.RequestException()
+        with self.assertLogs(level=WARNING) as warning:
+            self.assertEqual(
+                exporter.export([BASIC_SPAN]),
+                SpanExportResult.FAILURE,
+            )
+            self.assertEqual(mock_post.call_count, 1)
+            self.assertIn(
+                "Failed to export span batch code",
+                warning.records[0].message,
+            )
+
     @patch.object(Session, "post")
     def test_timeout_set_correctly(self, mock_post):
         resp = Response()