vLLM waiting-request gauge split by cause
vllm · gauge
Fields
id
metric-exposures/w3m-vllm-vllm-num-requests-waiting-by-reasonMetric name
vllm:num_requests_waiting_by_reason
Metric type
gauge
Unit
requests
Exposure
prometheus
Endpoint
:8000/metrics
Labels
model_nameenginereason
Enabled by default
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: no bucket family. reason in {capacity, deferred}. Documented invariant: 'Sum of all reasons equals vllm:num_requests_waiting'. Unique in this fleet: it separates 'no capacity' from 'deferred by LoRA budget, KV transfer or blocked status', which are different problems with different fixes.
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: no bucket family. reason in {capacity, deferred}. Documented invariant: 'Sum of all reasons equals vllm:num_requests_waiting'. Unique in this fleet: it separates 'no capacity' from 'deferred by LoRA budget, KV transfer or blocked status', which are different problems with different fixes.