[V1][Metrics] Fix http metrics middleware (#15894)

docs/source/design/v1/metrics.md

@@ -86,6 +86,17 @@ See [the PR which added this Dashboard](gh-pr:2316) for interesting and useful b
 
 Prometheus support was initially added [using the aioprometheus library](gh-pr:1890), but a switch was made quickly to [prometheus_client](gh-pr:2730). The rationale is discussed in both linked PRs.
 
+With the switch to `aioprometheus`, we lost a `MetricsMiddleware` to track HTTP metrics, but this was reinstated [using prometheus_fastapi_instrumentator](gh-pr:15657):
+
+```bash
+$ curl http://0.0.0.0:8000/metrics 2>/dev/null  | grep -P '^http_(?!.*(_bucket|_created|_sum)).*'
+http_requests_total{handler="/v1/completions",method="POST",status="2xx"} 201.0
+http_request_size_bytes_count{handler="/v1/completions"} 201.0
+http_response_size_bytes_count{handler="/v1/completions"} 201.0
+http_request_duration_highr_seconds_count 201.0
+http_request_duration_seconds_count{handler="/v1/completions",method="POST"} 201.0
+```
+
 ### Multi-process Mode
 
 In v0, metrics are collected in the engine core process and we use multi-process mode to make them available in the API server process. See <gh-pr:7279>.

vllm/entrypoints/openai/api_server.py

@@ -310,33 +310,33 @@ def mount_metrics(app: FastAPI):
     # We need to set PROMETHEUS_MULTIPROC_DIR environment variable
     # before prometheus_client is imported.
     # See https://prometheus.github.io/client_python/multiprocess/
-    from prometheus_client import (CollectorRegistry, make_asgi_app,
+    from prometheus_client import (REGISTRY, CollectorRegistry, make_asgi_app,
                                    multiprocess)
     from prometheus_fastapi_instrumentator import Instrumentator
 
+    registry = REGISTRY
+
     prometheus_multiproc_dir_path = os.getenv("PROMETHEUS_MULTIPROC_DIR", None)
     if prometheus_multiproc_dir_path is not None:
         logger.debug("vLLM to use %s as PROMETHEUS_MULTIPROC_DIR",
                      prometheus_multiproc_dir_path)
         registry = CollectorRegistry()
         multiprocess.MultiProcessCollector(registry)
-        Instrumentator(
-            excluded_handlers=[
-                "/metrics",
-                "/health",
-                "/load",
-                "/ping",
-                "/version",
-                "/server_info",
-            ],
-            registry=registry,
-        ).add().instrument(app).expose(app)
-
-        # Add prometheus asgi middleware to route /metrics requests
-        metrics_route = Mount("/metrics", make_asgi_app(registry=registry))
-    else:
-        # Add prometheus asgi middleware to route /metrics requests
-        metrics_route = Mount("/metrics", make_asgi_app())
+
+    Instrumentator(
+        excluded_handlers=[
+            "/metrics",
+            "/health",
+            "/load",
+            "/ping",
+            "/version",
+            "/server_info",
+        ],
+        registry=registry,
+    ).add().instrument(app).expose(app)
+
+    # Add prometheus asgi middleware to route /metrics requests
+    metrics_route = Mount("/metrics", make_asgi_app(registry=registry))
 
     # Workaround for 307 Redirect for /metrics
     metrics_route.path_regex = re.compile("^/metrics(?P<path>.*)$")