Deployment
Purpose
Helm chart and deployment scripts for deploying the Virtufin platform (API Gateway, WorkManager, WorkManager Runtime) on Kubernetes with Dapr sidecars, configurable per-service versioning, Dapr Components, health probes, and private registry support.
Requirements
Requirement: Helm Chart
The platform SHALL be deployable via a single Helm v2 chart that covers the API Gateway, WorkManager, and WorkManager Runtime with Dapr sidecars. Each service SHALL be independently enableable. WebSocket connections to venues are WebSocketFeed workers hosted by the Runtime, not a separate service (the WebSocketManager was retired, workmanager#103).
Scenario: Full stack install
- WHEN
helm install virtufin .is run - THEN all three services, their Dapr sidecars, Dapr Components, and ConfigMaps SHALL be created in the target namespace
Scenario: Single-service install
- WHEN the chart is installed with all but one service disabled
- THEN only the enabled service and its Dapr sidecar SHALL be deployed
Requirement: Per-Service Versioning
Each service's Docker image tag SHALL be independently configurable. Services SHALL support replicas, resource limits, and environment variables.
Scenario: Canary deployment
- WHEN a specific service's image tag is overridden
- THEN only that deployment SHALL roll out the new image; other services SHALL remain unchanged
Requirement: Dapr Component Management
The Helm chart SHALL create dapr.io/v1alpha1 Component resources for the state store and pub/sub broker. These SHALL reference Redis/Valkey with configurable host, port, password, redelivery count, and dead-letter topic.
Scenario: Production Redis
- WHEN deployed with a custom state store host and password
- THEN the Dapr state store component SHALL connect to the specified Redis with authentication
Scenario: Pub/Sub redelivery
- WHEN a message cannot be processed
- THEN the pubsub component SHALL redeliver up to the configured maximum count, with a configurable redelivery interval, before routing to the dead-letter topic
Requirement: Dapr Configuration
The Helm chart SHALL create a dapr.io/v1alpha1 Configuration resource with tracing, metrics, and mTLS settings.
Scenario: mTLS enforcement
- WHEN mTLS is enabled in Dapr config
- THEN all inter-service communication SHALL be encrypted and authenticated
Scenario: Tracing sampling
- WHEN tracing is configured with a sampling rate
- THEN the Dapr sidecar SHALL sample traces at that rate
Requirement: Service Discovery Configuration
The API Gateway's services.json SHALL be generated as a ConfigMap from Helm values, listing each enabled backend service with its gRPC host, port, Dapr app ID, and pubsub/state component names.
Scenario: Service registration
- WHEN a new backend service is added via Helm values
- THEN it SHALL appear in the generated
services.jsonConfigMap without manual intervention
Requirement: Health Probes
Every deployment SHALL configure Kubernetes liveness and readiness probes on the service's HTTP health endpoint. Dapr sidecars SHALL use the service health check path.
Scenario: Readiness gate
- WHEN a service is starting up and has not completed recovery
- THEN the readiness probe SHALL fail, preventing the Service from routing traffic to that pod
Requirement: Private Registry
All container images SHALL be pulled from a private Harbor registry using a Kubernetes image pull secret.
Scenario: Image pull
- WHEN Kubernetes schedules a pod
- THEN it SHALL authenticate to the private registry using the configured image pull secret
Requirement: Single-Service Deploy/Undeploy
Deployment scripts SHALL support deploying or undeploying individual services without affecting the rest of the release.
Scenario: Remove a single service
- WHEN the undeploy script is run targeting a specific service
- THEN only that deployment and service SHALL be removed while other services continue running
Scenario: Deploy a single service
- WHEN the deploy script is run targeting a specific service
- THEN only that deployment SHALL be installed or upgraded without redeploying other services