WIP - Add vLLM readiness probe script to sim container image

Dockerfile

@@ -49,10 +49,14 @@ WORKDIR /
 USER root
 RUN microdnf install -y dnf && \
     dnf install -y 'https://dl.fedoraproject.org/pub/epel/epel-release-latest-9.noarch.rpm' && \
-    dnf install -y zeromq && \
+    dnf install -y zeromq curl && \
     dnf clean all && \
     rm -rf /var/cache/dnf /var/lib/dnf
 
+# Copy readiness probe script
+COPY scripts/readiness_probe.sh /usr/local/bin/readiness_probe.sh
+RUN chmod +x /usr/local/bin/readiness_probe.sh
+
 COPY --from=builder /workspace/bin/llm-d-inference-sim /app/llm-d-inference-sim
 
 USER 65532:65532

scripts/readiness_probe.sh

@@ -0,0 +1,85 @@
+#!/bin/bash
+#
+# vLLM Readiness Probe Script
+#
+# This script checks if the vLLM inference server is fully ready by verifying:
+# 1. The basic health endpoint is responding
+# 2. The /v1/models endpoint is available (indicates model is loaded)
+# 3. The model metadata is properly returned
+#
+# This addresses the three stages of readiness:
+# - Container is running (handled by Kubernetes)
+# - vLLM API server is up (checked via /health)
+# - Model-specific API routes are ready (checked via /v1/models)
+#
+# Exit codes:
+#   0 - Service is ready
+#   1 - Service is not ready
+#
+# Usage:
+#   readiness_probe.sh PORT [HOST]
+#
+# Arguments:
+#   PORT - Port where vLLM server is listening (required)
+#   HOST - Host to connect to (optional, default: localhost)
+
+set -euo pipefail
+
+# Validate required arguments
+if [ -z "${1:-}" ]; then
+    echo "[readiness_probe] ERROR: PORT argument is required" >&2
+    echo "Usage: readiness_probe.sh PORT [HOST]" >&2
+    exit 1
+fi
+
+# Configuration
+PORT="${1}"
+HOST="${2:-localhost}"
+TIMEOUT="${READINESS_TIMEOUT:-5}"
+HEALTH_ENDPOINT="http://${HOST}:${PORT}/health"
+MODELS_ENDPOINT="http://${HOST}:${PORT}/v1/models"
+
+# Helper function for logging
+log() {
+    echo "[readiness_probe] $*" >&2
+}
+
+# Check if curl is available
+if ! command -v curl &> /dev/null; then
+    log "ERROR: curl command not found"
+    exit 1
+fi
+
+# Step 1: Check basic health endpoint
+log "Checking health endpoint: ${HEALTH_ENDPOINT}"
+if ! curl -sf --max-time "${TIMEOUT}" "${HEALTH_ENDPOINT}" > /dev/null 2>&1; then
+    log "Health endpoint not responding"
+    exit 1
+fi
+log "Health endpoint is responding"
+
+# Step 2: Check /v1/models endpoint (indicates model is loaded)
+log "Checking models endpoint: ${MODELS_ENDPOINT}"
+MODELS_RESPONSE=$(curl -sf --max-time "${TIMEOUT}" "${MODELS_ENDPOINT}" 2>&1)
+CURL_EXIT_CODE=$?
+
+if [ ${CURL_EXIT_CODE} -ne 0 ]; then
+    log "Models endpoint not responding (curl exit code: ${CURL_EXIT_CODE})"
+    exit 1
+fi
+
+# Step 3: Verify response contains model data
+if ! echo "${MODELS_RESPONSE}" | grep -q '"data"'; then
+    log "Models endpoint response does not contain expected 'data' field"
+    exit 1
+fi
+
+# Check if data array is not empty
+if echo "${MODELS_RESPONSE}" | grep -q '"data":\s*\[\s*\]'; then
+    log "Models endpoint returned empty data array - model not loaded yet"
+    exit 1
+fi
+
+log "vLLM server is ready - model is loaded and API is responding"
+exit 0
+