SGLang max-running-requests-under-SLO gauge - DEAD, NO SETTER
sglang · gauge
Fields
id
metric-exposures/w3m-sglang-sglang-max-running-requests-under-sloMetric name
sglang:max_running_requests_under_SLO
Metric type
gauge
Unit
requests
Exposure
prometheus
Endpoint
:30000/metrics
Labels
model_nameengine_typetp_rankpp_rankmoe_ep_rank
Enabled by default
no
notes
The other half of the dead chain. managers/scheduler.py line 1250 constructs the stats object with the comment 'TODO: max_running_requests_under_SLO has no setter - dead chain.' and max_running_requests_under_SLO=None, and the reporter reads it with getattr(..., None). Exported, permanently 0, and it is also what pins sglang:utilization to 0. The metric a capacity planner would most want is the one that does not work.
Notes
The other half of the dead chain. managers/scheduler.py line 1250 constructs the stats object with the comment 'TODO: max_running_requests_under_SLO has no setter - dead chain.' and max_running_requests_under_SLO=None, and the reporter reads it with getattr(..., None). Exported, permanently 0, and it is also what pins sglang:utilization to 0. The metric a capacity planner would most want is the one that does not work.