Merge pull request #2264 from anik120/fix-otel-instrumentation

LCORE-1801: Fix Otel intrumentation to respect OTEL_SDK_DISABLED flag

Author: tisnik

Makefile

@@ -36,7 +36,11 @@ CONTAINER_RUNTIME ?= $(shell command -v podman 2>/dev/null || command -v docker
 	docs/models/database.svg
 
 run-stack: ## Run lightspeed-stack directly, without building dependent service/s
-	uv run opentelemetry-instrument python3.12 src/lightspeed_stack.py -c $(CONFIG)
+	@if [ "$${OTEL_SDK_DISABLED:-true}" = "false" ]; then \
+		uv run opentelemetry-instrument python3.12 src/lightspeed_stack.py -c $(CONFIG); \
+	else \
+		uv run python3.12 src/lightspeed_stack.py -c $(CONFIG); \
+	fi
 
 run: start-llama-stack-container ## Run the service locally with dependent services
 	@echo "Starting Lightspeed Core Stack..."

deploy/lightspeed-stack/Containerfile

@@ -133,9 +133,13 @@ ENV PATH="/app-root/.venv/bin:$PATH"
 # We place them at /app-root/providers.d. YAMLs there reference lightspeed_stack_providers.*, so that package must be on PYTHONPATH.
 ENV PYTHONPATH="/app-root"
 
+# Copy entrypoint script
+COPY ${LSC_SOURCE_DIR}/scripts/entrypoint.sh /app-root/entrypoint.sh
+RUN chmod +x /app-root/entrypoint.sh
+
 # Run the application
 EXPOSE 8080
-ENTRYPOINT ["opentelemetry-instrument", "python3.12", "src/lightspeed_stack.py"]
+ENTRYPOINT ["/app-root/entrypoint.sh"]
 
 LABEL vendor="Red Hat, Inc." \
     name="lightspeed-core/lightspeed-stack-rhel9" \

scripts/entrypoint.sh

@@ -0,0 +1,10 @@
+#!/bin/bash
+set -e
+
+# Only use OpenTelemetry instrumentation if explicitly enabled
+# Use explicit venv paths to ensure dependencies are found
+if [ "${OTEL_SDK_DISABLED:-true}" = "false" ]; then
+    exec /app-root/.venv/bin/opentelemetry-instrument /app-root/.venv/bin/python src/lightspeed_stack.py "$@"
+else
+    exec /app-root/.venv/bin/python src/lightspeed_stack.py "$@"
+fi

tests/e2e-prow/rhoai/manifests/lightspeed/lightspeed-stack.yaml

@@ -63,7 +63,7 @@ spec:
               key: key
               optional: true
       image: ${LIGHTSPEED_STACK_IMAGE}
-      command: ["/bin/sh", "-c", "mkdir -p /tmp/data && exec opentelemetry-instrument python3.12 src/lightspeed_stack.py"]
+      command: ["/bin/sh", "-c", "mkdir -p /tmp/data && exec /app-root/entrypoint.sh"]
       ports:
         - containerPort: 8080
       # TCP probes avoid HTTP/auth. LCS + Llama handshake and large images can take 60–120s before :8080 listens;