In vLLM v0.23.0, `vllm:iteration_tokens_total` describes batch/step work distribution rather than a cumulative token counter despite total appearing in its declared name.

MetricvLLM v0.23.0confidence: highverified 10 hr agoverified-code

PrometheusStatLoggerV1 metrics

vLLM v0.23.0 v1 metrics loggers.py(vllm/v1/metrics/loggers.py:1152-1177)

definesprimarysource_codecode_introspection

Pinned release-tag semantics; do not carry to another version without continuity evidence.

In vLLM v0.23.0, vllm:iteration_tokens_total describes batch/step work distribution rather than a cumulative token counter despite total appearing in its declared name.