Operate#

Day-2 operator playbook for running Strata. Each page is a workflow: the steps an operator takes from the LB or the admin API to keep the cluster healthy, drain a member, plan capacity, or recover from a backup. Implementation details and tuning rationale live in Best Practices and Architecture; this section answers “what do I run, in what order, and what do I watch?”

Drain a cluster
Step-by-step decommission workflow — /admin/v1/clusters/{id}/drain, watch live ETA + bandwidth chips, wait for deregister-ready, remove from STRATA_RADOS_CLUSTERS.
Monitoring
Prometheus scrape, key metrics, Grafana dashboard, OTel collector + Jaeger wire-up, in-process trace browser, audit log retention.
Scaling
CPU / RAM per replica, metadata-tier sizing pointers, when to scale out vs scale up the gateway tier or the meta tier.

Image verification
Verify the cosign signature, SLSA provenance, SBOM, and license inventory of a released image before deploy — plus Kyverno + Sigstore admission-control manifests and a Rekor forensic query.

See also#

Get Started — first install and smoke run.
Deploy — deployment shapes (single node, docker compose, multi-replica, Kubernetes).
Best Practices — tuning knobs and runbooks for placement, tracing, GC, quotas, multi-cluster.
Architecture — implementation rationale for every operate-facing knob.