Purpose
How the S3 endpoint is performing from the callers’ point of view: request volume,
errors, latency percentiles, and bytes moved.
Panels
- Request Rate by Method (req/s) — per HTTP method (GET/PUT/HEAD…).
- Error Rate (4xx + 5xx) (req/s) — client and server errors together; correlate
spikes with the Audit Log to see which requests failed. - Request Duration p50 / p95 / p99 (seconds) — median vs tail latency. A rising
p99 with a flat p50 = a subset of requests (often large objects or a struggling
provider set) is slow, not everything. - Bytes Transferred by Direction (bytes/s) — in vs out.
Reading it
- Error spike + Audit Log full of AUTH_FAILED → a client has bad credentials, the
service itself is fine. - Sustained 5xx → check Service Health next.
Screenshots
