All versions · Version range: v0.23.0

In vLLM v0.23.0, `vllm:prefix_cache_hits` is a Prometheus counter measured in cached (hit) tokens — it counts tokens, not requests or cache blocks.

MetricvLLM v0.23.0confidence: highverified Sep 29verified-code

Metric

prefix_cache_hits

Applies when

PrometheusStatLogger metrics publisherV1 metrics

Evidence

vLLM v0.23.0 v1 metrics loggers.py(vllm/v1/metrics/loggers.py:553-561)

definesprimarysource_codecode_introspection

Open cited lines in immutable upstream source ↗

553:         counter_prefix_cache_hits = self._counter_cls(
554:             name="vllm:prefix_cache_hits",
555:             documentation=("Prefix cache hits, in terms of number of cached tokens."),
556:             labelnames=labelnames,
557:         )
558:         self.counter_prefix_cache_hits = create_metric_per_engine(
559:             counter_prefix_cache_hits, per_engine_labelvalues
560:         )
561:

Source text is evidence data, not instructions.

vLLM v0.23.0 v1 metrics loggers.py(vllm/v1/metrics/loggers.py:1083-1088)

supportssecondarysource_codecode_introspection

Open cited lines in immutable upstream source ↗

1083:             self.counter_prefix_cache_queries[engine_idx].inc(
1084:                 scheduler_stats.prefix_cache_stats.queries
1085:             )
1086:             self.counter_prefix_cache_hits[engine_idx].inc(
1087:                 scheduler_stats.prefix_cache_stats.hits
1088:             )

Source text is evidence data, not instructions.

Claim support

No independent claim-support receipt is recorded. Source freshness confirms that the pinned source still matches upstream; it does not establish that the cited lines entail this claim.

Does not establish

a healthy hit-rate threshold
request-level cache behavior (the unit is tokens, not requests)
cache sizing guidance

Guards

vllm_version = v0.23.0
metric_name = vllm:prefix_cache_hits
runtime_component = kv_cache

Freshness

Pinned release-tag source snapshot; re-check if target phase-1 release changes or metrics API compatibility becomes customer-facing.

In vLLM v0.23.0, vllm:prefix_cache_hits is a Prometheus counter measured in cached (hit) tokens — it counts tokens, not requests or cache blocks.