vLLM end-to-end request latency histogram
vllm · histogram
Fields
id
metric-exposures/w3m-vllm-vllm-e2e-request-latency-secondsMetric name
vllm:e2e_request_latency_seconds
Metric type
histogram
Unit
seconds
Exposure
prometheus
Endpoint
:8000/metrics
Histogram buckets
0.30.50.811.522.551015203040506012024048096019207680
Labels
model_nameengine
Enabled by default
yes
Phase scoped
yes
notes
vLLM main, PrometheusStatLogger in vllm/v1/metrics/loggers.py, read 2026-10-04. On by default (EngineArgs.disable_log_stats defaults False); served at :8000/metrics. Bucket family name in --custom-histogram-buckets: request_latency. Split from the whole request rather than attributed to the phase, which is what makes it usable for diagnosing where a TTFT-vs-ITL regression came from.
Notes
vLLM main, PrometheusStatLogger in vllm/v1/metrics/loggers.py, read 2026-10-04. On by default (EngineArgs.disable_log_stats defaults False); served at :8000/metrics. Bucket family name in --custom-histogram-buckets: request_latency. Split from the whole request rather than attributed to the phase, which is what makes it usable for diagnosing where a TTFT-vs-ITL regression came from.