How MCPG replicas share state. The cluster coordinator selection (single_node, redis, nats, consul, etcd), the real per-backend config keys, and the primitive-inheritance model that makes single-node work out of the box.

The four shapes a production MCPG fleet lands on — single-node, single-instance with shared state, multi-replica HA via Redis or NATS, and air-gapped — with the config that distinguishes each.

Deploy and run MCPG on Kubernetes with the operator and Helm chart — the eight mcpg.dev CRDs, the validating admission webhook, cert-manager TLS, metrics and probe endpoints, and the air-gap path.

The signal triad (logs, metrics, traces), the audit ledger, and the two health probers that keep a fleet honest — what each sink kind is, where the master switches live, and how to wire scraping and tracing.