[V0][Metrics] Deprecate some KV/prefix cache metrics

vllm:cpu_cache_usage_perc and vllm:cpu_prefix_cache_hit_rate will
no longer be relevant in V1 since we no longer implement KV cache
offloading. So these metrics should be considered deprecated.

And as agreed in #12592, we have added prefix_cache_queries and
prefix_cache_hits counters to replace the prefix_cache_hit_rate
gauge as it allows the interval over which the hit rate is
calculated to be controlled in a Prometheus query like:

```
rate(prefix_cache_queries[5m]) / rate(prefix_cache_hits[5m])
```

In theory, we could ease the transition be implementing the
old hit rate metric in V1 and the new queries/hits metrics
in V0, but it's probably not worthwhile unless we learn the
hit rate metric is heavily used by V0 users.

Signed-off-by: Mark McLoughlin <[email protected]>

vllm/engine/metrics.py

@@ -79,26 +79,41 @@ def __init__(self, labelnames: List[str], vllm_config: VllmConfig):
             documentation="Number of requests swapped to CPU.",
             labelnames=labelnames,
             multiprocess_mode="sum")
+
         #   KV Cache Usage in %
         self.gauge_gpu_cache_usage = self._gauge_cls(
             name="vllm:gpu_cache_usage_perc",
             documentation="GPU KV-cache usage. 1 means 100 percent usage.",
             labelnames=labelnames,
             multiprocess_mode="sum")
+
+        # Deprecated in 0.8 - KV cache offloading is not used in V1
+        # TODO: in 0.9, only enable if show_hidden_metrics=True
         self.gauge_cpu_cache_usage = self._gauge_cls(
             name="vllm:cpu_cache_usage_perc",
-            documentation="CPU KV-cache usage. 1 means 100 percent usage.",
+            documentation=(
+                "CPU KV-cache usage. 1 means 100 percent usage. "
+                "DEPRECATED: KV cache offloading is not used in V1"),
             labelnames=labelnames,
             multiprocess_mode="sum")
-        #   Prefix caching block hit rate
+
+        # Deprecated in 0.8 - KV cache offloading is not used in V1
+        # TODO: in 0.9, only enable if show_hidden_metrics=True
         self.gauge_cpu_prefix_cache_hit_rate = self._gauge_cls(
             name="vllm:cpu_prefix_cache_hit_rate",
-            documentation="CPU prefix cache block hit rate.",
+            documentation=(
+                "CPU prefix cache block hit rate. "
+                "DEPRECATED: KV cache offloading is not used in V1"),
             labelnames=labelnames,
             multiprocess_mode="sum")
+
+        # Deprecated in 0.8 - replaced by queries+hits counters in V1
+        # TODO: in 0.9, only enable if show_hidden_metrics=True
         self.gauge_gpu_prefix_cache_hit_rate = self._gauge_cls(
             name="vllm:gpu_prefix_cache_hit_rate",
-            documentation="GPU prefix cache block hit rate.",
+            documentation=("GPU prefix cache block hit rate. "
+                           "DEPRECATED: use vllm:gpu_prefix_cache_queries and "
+                           "vllm:gpu_prefix_cache_queries in V1"),
             labelnames=labelnames,
             multiprocess_mode="sum")