Platform requirements

Orleans can run on Kubernetes, managed container platforms, virtual machines, and bare-metal hosts. The platform must support Orleans’s process and network model; a platform name alone doesn’t make a deployment production-ready.

Required capabilities

A production platform must provide:

Stable per-instance addresses or explicit advertised endpoint mapping.
Direct bidirectional TCP connectivity from every silo to every silo endpoint.
Reachability from Orleans clients to advertised gateway endpoints.
Multiple concurrently running instances across failure domains.
Graceful termination notification and a configurable termination deadline.
Startup, readiness, and liveness checks with separate semantics.
Secure access to a supported clustering provider and all state providers.
Workload identity or secure secret and certificate delivery.
Central logs, metrics, traces, and deployment metadata.
Resource requests, limits, scaling controls, and disruption policies.

An HTTP-only platform isn’t sufficient unless Orleans silos can also establish direct TCP connections to individual instances. A load-balanced HTTP endpoint doesn’t replace silo-to-silo connectivity.

Platform guidance

Kubernetes provides direct pod networking. Explicit endpoint configuration is recommended; the Orleans hosting package is optional and limited to simple one-Deployment-per-cluster topologies.
Service Fabric uses an application-authored stateless Reliable Service integration with runtime-allocated endpoints and an external Orleans clustering provider.
Azure App Service requires validation of private per-instance address and port mapping on Windows and Linux.
Azure Container Apps can map unique TCP port pairs on an internal environment’s private IP to one replica per silo app; scaling multiple silos as replicas of one app relies on per-replica networking that the platform doesn’t publish as a supported contract.

For another platform, prove the required capabilities with at least three silos under rolling replacement, scale-in, host restart, and network interruption. Record the exact endpoint mapping and shutdown behavior in the deployment runbook.

Shared responsibility

Orleans provides membership, grain activation, placement, and call routing. The platform and application remain responsible for:

Durable state and provider availability.
Correct health endpoints and degraded-mode policy.
Network security and transport encryption.
Compatible releases and state migrations.
Capacity, overload control, backup, restore, and incident response.