Nais production knowledge check

Nais provides runtime and integrated building blocks through higher-level contracts. Cloud environments use Kubernetes on GKE and managed services underneath.

Platform engineering centralizes common complexity without transferring product and domain responsibility from the team.

Naisjob represents finite work and supports scheduled execution; Application represents a long-running workload.

Readiness controls whether the instance is an eligible traffic endpoint. Liveness asks whether it should be restarted.

Autoscaling multiplies downstream use. Connection pools, rolling-deployment overlap and database capacity must be planned together.

Repository authorization is part of the trusted deployment path and limits which GitHub source can deploy for the team.

TokenX supports internal applications acting on behalf of citizens authenticated through ID-porten.

Workload identity represents the running application. User/delegated context and domain authorization must be handled separately.

Service discovery avoids external exposure and unnecessary hops; access policies declare the permitted service edge.

Connectivity, token authorization and domain authorization are separate controls. An allowed network edge is not blanket business permission.

Nais distinguishes durability/availability from backup protection against mistaken deletion. The team owns recovery requirements and configuration.

Kafka is durable, but Nais recommends being able to restore data from another system. Replay, schemas and duplicate handling need design.

High-cardinality identifiers harm metric systems. Use bounded labels for aggregation and safe request-specific context in logs or traces.

Infrastructure health does not prove product correctness. SLOs and business signals must represent the service outcome.

Production readiness combines workload design, security, data, observability, change, recovery and explicit ownership. Features alone are not evidence.