Best Practices#

Tuning guides for running Strata in production. Each page covers one operational concern end-to-end with the env knobs, metrics, and runbook shape an on-call operator needs. Day-2 workflows (drain a cluster, monitor, scale, back up, plan capacity) live under Operate; the pages below cover the knobs the workflows reference.

Placement + rebalance — Per-bucket Placement policy, STRATA_REBALANCE_* knobs, drain sentinel + safety rails, register → drain → deregister runbook.

Tracing — OpenTelemetry coverage matrix, span name conventions, strata.component=gateway|worker filters, sampling.

Web UI (Strata Console) — Embedded operator console: pages, env vars, end-to-end tests.

GC + lifecycle tuning — Tuning STRATA_GC_CONCURRENCY / STRATA_LIFECYCLE_CONCURRENCY and STRATA_GC_SHARDS, plus the dual-write cutover playbook.

Quotas + billing — BucketQuota / UserQuota shape, QuotaExceeded 403, the bucket-usage counter, the reconcile + usage-rollup workers.

S3 multi-cluster routing — STRATA_S3_CLUSTERS / STRATA_S3_CLASSES env shape, credentials_ref discriminator, per-class (cluster, bucket) routing, rolling-restart workflow.

Compliance — S3 Object Lock COMPLIANCE workflow, retention modes, legal hold, and the three objectlock:* audit verbs.

Billing — Byte-seconds trapezoid math, intra-day sampling, and the usage_aggregates feed external invoice generators consume.

Production hardening — 12-line checklist for prod readiness: HTTP timeouts, built-in TLS, backend mTLS, trusted proxies, split admin listener, ingress rate limit.

See also#

Operate for day-2 ops workflows — drain a cluster, monitor, scale, back up, plan capacity.
Architecture deep dive for the implementation rationale behind each knob.
Deploy for end-to-end deployment guides (single-node, Docker Compose, multi-replica, Kubernetes).
S3 Compatibility for the supported / unsupported S3 surface.