Simplify prometheus metrics (#1981)

Co-authored-by: Mohit Reddy <mohitreddy1996@users.noreply.github.com>

test/srt/test_metrics.py

@@ -22,23 +22,41 @@ class TestEnableMetrics(unittest.TestCase):
         )
 
         try:
-            # Make a request to generate some metrics
+            # Make some requests to generate some metrics
             response = requests.get(f"{DEFAULT_URL_FOR_TEST}/health_generate")
             self.assertEqual(response.status_code, 200)
 
+            response = requests.post(
+                f"{DEFAULT_URL_FOR_TEST}/generate",
+                json={
+                    "text": "The capital of France is",
+                    "sampling_params": {
+                        "temperature": 0,
+                        "max_new_tokens": 32,
+                    },
+                    "stream": True,
+                },
+                stream=True,
+            )
+            for _ in response.iter_lines(decode_unicode=False):
+                pass
+
             # Get metrics
             metrics_response = requests.get(f"{DEFAULT_URL_FOR_TEST}/metrics")
             self.assertEqual(metrics_response.status_code, 200)
             metrics_content = metrics_response.text
 
-            print(f"{metrics_content=}")
+            print(f"metrics_content=\n{metrics_content}")
 
             # Verify essential metrics are present
             essential_metrics = [
+                "sglang:num_running_reqs",
+                "sglang:token_usage",
+                "sglang:gen_throughput",
+                "sglang:cache_hit_rate",
+                "sglang:func_latency_seconds",
                 "sglang:prompt_tokens_total",
                 "sglang:generation_tokens_total",
-                "sglang:max_total_num_tokens",
-                "sglang:context_len",
                 "sglang:time_to_first_token_seconds",
                 "sglang:time_per_output_token_seconds",
                 "sglang:e2e_request_latency_seconds",
@@ -50,6 +68,7 @@ class TestEnableMetrics(unittest.TestCase):
             # Verify model name label is present and correct
             expected_model_name = DEFAULT_SMALL_MODEL_NAME_FOR_TEST
             self.assertIn(f'model_name="{expected_model_name}"', metrics_content)
+
             # Verify metrics have values (not empty)
             self.assertIn("_sum{", metrics_content)
             self.assertIn("_count{", metrics_content)