Initial commit

libs/observability/opentelemetry_setup.py

@@ -0,0 +1,99 @@
+"""OpenTelemetry tracing and metrics initialization."""
+
+import os
+from typing import Any
+
+from opentelemetry import metrics, trace
+from opentelemetry.exporter.otlp.proto.grpc.metric_exporter import OTLPMetricExporter
+from opentelemetry.exporter.otlp.proto.grpc.trace_exporter import OTLPSpanExporter
+from opentelemetry.instrumentation.fastapi import FastAPIInstrumentor
+from opentelemetry.instrumentation.httpx import HTTPXClientInstrumentor
+from opentelemetry.instrumentation.psycopg2 import Psycopg2Instrumentor
+from opentelemetry.instrumentation.redis import RedisInstrumentor
+from opentelemetry.sdk.metrics import MeterProvider
+from opentelemetry.sdk.metrics.export import (
+    MetricExporter,
+    PeriodicExportingMetricReader,
+)
+from opentelemetry.sdk.resources import Resource
+from opentelemetry.sdk.trace import TracerProvider
+from opentelemetry.sdk.trace.export import BatchSpanProcessor, SpanExporter
+
+
+def init_opentelemetry(
+    service_name: str,
+    service_version: str = "1.0.0",
+    otlp_endpoint: str | None = None,
+) -> tuple[Any, Any]:
+    """Initialize OpenTelemetry tracing and metrics"""
+
+    # Create resource
+    resource = Resource.create(
+        {
+            "service.name": service_name,
+            "service.version": service_version,
+            "service.instance.id": os.getenv("HOSTNAME", "unknown"),
+        }
+    )
+
+    # Configure tracing
+    span_exporter: SpanExporter
+    if otlp_endpoint:
+        span_exporter = OTLPSpanExporter(endpoint=otlp_endpoint)
+        span_processor = BatchSpanProcessor(span_exporter)
+    else:
+        # Use console exporter for development
+        try:
+            # pylint: disable=import-outside-toplevel
+            from opentelemetry.sdk.trace.export import ConsoleSpanExporter
+
+            span_exporter = ConsoleSpanExporter()
+        except ImportError:
+            # Fallback to logging exporter
+            # pylint: disable=import-outside-toplevel
+            from opentelemetry.sdk.trace.export import ConsoleSpanExporter
+
+            span_exporter = ConsoleSpanExporter()
+        span_processor = BatchSpanProcessor(span_exporter)
+
+    tracer_provider = TracerProvider(resource=resource)
+    tracer_provider.add_span_processor(span_processor)
+    trace.set_tracer_provider(tracer_provider)
+
+    # Configure metrics
+    metric_exporter: MetricExporter
+    if otlp_endpoint:
+        metric_exporter = OTLPMetricExporter(endpoint=otlp_endpoint)
+        metric_reader = PeriodicExportingMetricReader(
+            metric_exporter, export_interval_millis=30000
+        )
+    else:
+        # Use console exporter for development
+        try:
+            # pylint: disable=import-outside-toplevel
+            from opentelemetry.sdk.metrics.export import ConsoleMetricExporter
+
+            metric_exporter = ConsoleMetricExporter()
+        except ImportError:
+            # Fallback to logging exporter
+            from opentelemetry.sdk.metrics.export import ConsoleMetricExporter
+
+            metric_exporter = ConsoleMetricExporter()
+        metric_reader = PeriodicExportingMetricReader(
+            metric_exporter, export_interval_millis=30000
+        )
+
+    meter_provider = MeterProvider(resource=resource, metric_readers=[metric_reader])
+    metrics.set_meter_provider(meter_provider)
+
+    # Auto-instrument common libraries
+    try:
+        FastAPIInstrumentor().instrument()
+        HTTPXClientInstrumentor().instrument()
+        Psycopg2Instrumentor().instrument()
+        RedisInstrumentor().instrument()
+    except Exception:  # pylint: disable=broad-exception-caught
+        # Ignore instrumentation errors in tests
+        pass
+
+    return trace.get_tracer(service_name), metrics.get_meter(service_name)