starlette.requests.total

Total HTTP requests received

Dimensions:None

Technical Annotations (21)

Configuration Parameters (8)

metrics[].typerecommended: Pods

HPA metric type for custom per-pod QPS

metrics[].pods.metric.namerecommended: http_requests_per_second

Custom metric name exposed via Prometheus Adapter

metrics[].pods.target.averageValuerecommended: 10

Example: 10 RPS/pod target (adjust to 70-80% of observed max)

db_pool_sizerecommended: 32

Increased from 10 to handle async I/O concurrency (2-4× workers)

maxReplicasrecommended: 40

HPA max for burst capacity

spec.minAvailablerecommended: 6

PodDisruptionBudget to maintain availability during disruptions

scrape_intervalrecommended: 1s

provides sufficient samples for rate() queries on short time windows

groups.intervalrecommended: 5s

recording rule evaluation interval for requests per second calculation

Technical References (13)

APIRoutecomponentBaseHTTPMiddlewarecomponentlimit_reqcomponentPrometheus AdaptercomponentHPA v2componentstarlette.responses.FileResponsecomponentstarlette.staticfiles.StaticFilescomponent_parse_range_header()componentRangeprotocol_RANGE_PATTERNcomponentrate()conceptrecording ruleconceptstarlette_requests_totalcomponent

Related Insights (5)

Missing request timeouts and rate limits allow resource exhaustionwarning

▸

CPU-only HPA fails to scale I/O-bound async FastAPI under loadwarning

▸

Payments API p95 latency at 420ms during bursts with idle CPUwarning

▸

Range header causes O(n^2) CPU exhaustion in FileResponsecritical

▸

Prometheus rate() queries return no data due to insufficient scrape sampleswarning

▸