Skip to content

Deployment

Kubernetes

The WorkManager is deployed as a Kubernetes Deployment with a Dapr sidecar.

Prerequisites

  • Kubernetes cluster
  • Dapr installed on the cluster
  • Valkey (Redis-compatible) for state store and pub/sub

Quick Deploy

helm install virtufin-workmanager ../helm \
  --namespace virtufin --create-namespace \
  --set workmanager.enabled=true

Configuration

Env Variable Default Description
HttpPort 5001 HTTP port
GrpcPort 5002 gRPC port
Logging__LogLevel__Default Information Minimum log level

Health Checks

  • /health, /healthz — liveness
  • /ready — readiness (the Runtime reports not ready until its DotNetDll bridge warm-up finished)

Rolling Updates

The WorkManager keeps no worker state in memory, so its replicas roll freely. When a Runtime pod is replaced, its workers' actors are placed on the remaining pods and self-heal on their next invocation; RuntimeWorkerPokeService bounds that for quiet workers (see Worker Lifecycle: Recovery).

Known Limitations

Dapr gRPC App Health Checks Not Supported

When using Dapr with app-protocol: grpc, the Dapr sidecar probes the application via the dapr.proto.runtime.v1.AppCallbackHealthCheck gRPC service. The Dapr .NET SDK does not currently provide a server-side implementation of this service. As a result, Dapr app health checks are disabled (dapr.io/enable-app-health-check: false).

Kubernetes liveness and readiness probes (/health endpoint) continue to handle pod health monitoring. Future improvement: implement the Dapr AppCallback gRPC service to enable Dapr's soft health check mechanism for pausing work delivery during unhealthy states.