Switch site host to WebApplicationBuilder with Kestrel HTTP/2 gRPC server,
add GrpcPort/keepalive config, wire SiteStreamManager as ISiteStreamSubscriber,
expose gRPC ports in docker-compose, add site seed script, update all 10
requirement docs + CLAUDE.md + README.md for the new dual-transport architecture.
After receiving the initial snapshot via ClusterClient, the bridge actor
now opens a gRPC server-streaming subscription via SiteStreamGrpcClient
for ongoing AttributeValueChanged/AlarmStateChanged events. Adds NodeA/
NodeB failover with max 3 retries, retry count reset on successful event,
and IWithTimers-based reconnect scheduling.
- DebugStreamBridgeActor: gRPC stream after snapshot, reconnect state machine
- DebugStreamService: inject SiteStreamGrpcClientFactory, resolve gRPC addresses
- ServiceCollectionExtensions: register SiteStreamGrpcClientFactory singleton
- SiteStreamGrpcClient: make SubscribeAsync/Unsubscribe virtual for testability
- SiteStreamGrpcClientFactory: make GetOrCreate virtual for testability
- New test suite: DebugStreamBridgeActorTests (8 tests)
Per-site gRPC client for central-side streaming subscriptions to site
servers. SiteStreamGrpcClient manages server-streaming calls with
keepalive, converts proto events to domain types, and supports
cancellation via Unsubscribe. SiteStreamGrpcClientFactory caches one
client per site identifier.
Includes InternalsVisibleTo for test access to conversion helpers and
comprehensive unit tests for event mapping, quality/alarm-state
conversion, unsubscribe behavior, and factory caching.
Proto3 definition with SiteStreamService (server streaming), Quality and
AlarmStateEnum enums with UNSPECIFIED=0, google.protobuf.Timestamp for
cross-platform timestamps. Pre-generated C# stubs checked in (no protoc
at build time). 10 roundtrip tests covering serialization, oneof
discrimination, and Timestamp<->DateTimeOffset conversion.
Define the SiteStreamService proto for real-time instance event
streaming (attribute value changes, alarm state changes) from site
nodes to central. Add pre-generated C# stubs following the existing
LmxProxy pattern, gRPC NuGet packages with FrameworkReference for
ASP.NET Core server types, and proto roundtrip tests.
ClusterClient Sender refs are temporary proxies — valid for immediate reply
but not durable for future Tells. Events now flow as DebugStreamEvent through
SiteCommunicationActor → ClusterClient → CentralCommunicationActor → bridge
actor (same pattern as health reports). Also fix DebugStreamHub to use
IHubContext for long-lived callbacks instead of transient hub instance.
The debug view polled every 2s by re-subscribing for full snapshots. Now a
persistent DebugStreamBridgeActor on central subscribes once and receives
incremental Akka stream events from the site, forwarding them to the Blazor
component via callbacks and to the CLI via a new SignalR hub at
/hubs/debug-stream. Adds `debug stream` CLI command with auto-reconnect.
Organize documentation by moving requirements (HighLevelReqs, Component-*,
lmxproxy_protocol) to docs/requirements/ and test infrastructure docs to
docs/test_infra/. Updates all cross-references in README, CLAUDE.md,
infra/README, component docs, and 23 plan files.
Add ActiveNodeHealthCheck that returns 200 only on the Akka.NET cluster
leader, enabling Traefik to route traffic to the active central node and
automatically fail over when the leader changes. Also fixes AkkaClusterHealthCheck
to resolve ActorSystem from AkkaHostedService (was always null via DI).
Replace the CLI's Akka.NET ClusterClient transport with a simple HTTP client
targeting a new POST /management endpoint on the Central Host. The endpoint
handles Basic Auth, LDAP authentication, role resolution, and ManagementActor
dispatch in a single round-trip — eliminating the CLI's Akka, LDAP, and
Security dependencies.
Also fixes DCL ReSubscribeAll losing subscriptions on repeated reconnect by
deriving the tag list from _subscriptionsByInstance instead of _subscriptionIds.
- Add JoeAppEngine folder to OPC UA nodes.json (BTCS, AlarmCntsBySeverity, Scheduler/ScanTime)
- Fix DataConnectionActor: capture Self in PreStart for use from non-actor threads,
preventing Self.Tell failure in Disconnected event handler
- Implement InstanceActor.HandleConnectionQualityChanged to mark attributes Bad on disconnect
- Fix LmxFakeProxy TagMapper to serialize arrays as JSON instead of "System.Int32[]"
- Allow DataType and DataSourceReference updates in TemplateService.UpdateAttributeAsync
- Update test_infra_opcua.md with JoeAppEngine documentation
Defines a gRPC server implementing the scada.ScadaService proto that bridges
to the existing OPC UA test server. Enables end-to-end testing of
RealLmxProxyClient without a Windows LmxProxy deployment.
Replace stub ILmxProxyClient with production proto-generated gRPC client
(RealLmxProxyClient) that connects to LmxProxy servers with x-api-key
metadata header authentication. Includes pre-generated proto stubs for
ARM64 Docker compatibility, updated adapter with proper quality mapping
(Good/Uncertain/Bad), subscription via server-streaming RPC, and 20 unit
tests covering all operations. Updated Component-DataConnectionLayer.md
to reflect the actual implementation.
- InboundScriptExecutor lazy-compiles scripts on first request, solving
the multi-node problem where methods created via CLI/UI were only compiled
on the ManagementActor's node, not the node handling the HTTP request.
- ManagementActor hot-registers API method scripts on create/update/delete
for the local node.
- FlatteningService prefixes the "attribute" field in composed alarm trigger
configs with the composition instance name so alarms evaluate against the
correct path-qualified attribute (e.g. CoolingTank.Level not Level).
NotificationRepository.GetAllNotificationListsAsync() was missing
.Include(Recipients), causing artifact deployments to push empty recipient
lists to sites. Also load shared scripts from SQLite on DeploymentManager
startup so they're available before Instance Actors compile their scripts.
Completes the Inbound API → site script call chain by adding RouteToCallRequest
handlers in SiteCommunicationActor and DeploymentManagerActor. Also replaces the
placeholder dispatch table in InboundScriptExecutor with Roslyn compilation of
API method scripts at startup, enabling user-defined inbound API methods to call
instance scripts across the cluster.
Add SiteReplicationActor (runs on every site node) to replicate deployed
configs and store-and-forward buffer operations to the standby peer via
cluster member discovery and fire-and-forget Tell. Wire ReplicationService
handler and pass replication actor to DeploymentManagerActor singleton.
Fix 5 pre-existing ConfigurationDatabase test failures: RowVersion NOT NULL
on SQLite, stale migration name assertion, and seed data count mismatch.
Three runtime bugs fixed:
- DataConnectionActor: TagValueReceived/TagResolutionSucceeded/Failed not
handled in any Become state — OPC UA values went to dead letters. Added
initial read after subscribe to seed current values immediately.
- AlarmActor: ParseEvalConfig expected "attributeName"/"matchValue"/"min"/
"max" keys but seed data uses "attribute"/"value"/"high"/"low". Added
support for both conventions and !=prefix for not-equal matching.
- InstanceActor: snapshots reported all alarms (including unevaluated) with
correct priorities and source timestamps instead of current UTC. Removed
bogus Vibration template attribute that shadowed Speed's tag mapping.
Adds `debug snapshot --id <int>` to query a running instance's current
attribute values and alarm states without the subscribe/stream overhead
of the debug view. Routes through ManagementActor → CommunicationService
→ site DeploymentManager → InstanceActor using the existing remote query
pattern.
IServiceProvider now flows through the actor chain (DeploymentManagerActor
→ InstanceActor → ScriptActor → ScriptExecutionActor) so scripts can
resolve IExternalSystemClient, IDatabaseGateway, and
INotificationDeliveryService from DI. ScriptGlobals exposes ExternalSystem,
Database, Notify, and Scripts as top-level properties so scripts can use
them without the Instance. prefix.
Both nodes of a site cluster were sending health reports. The standby
node (without the DeploymentManager singleton) reported 0 instances and
no connections, overwriting the active node's data in the aggregator.
Added IsActiveNode flag to ISiteHealthCollector, set by
DeploymentManagerActor on PreStart/PostStop. HealthReportSender skips
sending when the node is not active. Also ensured EnsureDclConnections
is called during startup batch creation so data connections survive
container restarts.
UpdateInstanceCounts() was only called before Instance Actors were
created (in HandleStartupConfigsLoaded), showing 0 enabled on the
health dashboard. Now also called after each batch in
HandleStartNextBatch to reflect actual running actor count.
Add 33 new management message records, ManagementActor handlers, and CLI
commands to close all functionality gaps between the Central UI and the
Management CLI. New capabilities include:
- Template member CRUD (attributes, alarms, scripts, compositions)
- Shared script CRUD
- Database connection definition CRUD
- Inbound API method CRUD
- LDAP scope rule management
- API key enable/disable
- Area update
- Remote event log and parked message queries
- Missing get/update commands for templates, sites, instances, data
connections, external systems, notifications, and SMTP config
Includes 12 new ManagementActor unit tests covering authorization,
happy-path queries, and error handling. Updates CLI README and component
design documents (Component-CLI.md, Component-ManagementService.md).
Wired ISiteHealthCollector calls for script errors (ScriptExecutionActor),
alarm eval errors (AlarmActor), dead letters (DeadLetterMonitorActor), and
S&F buffer depth placeholder. Added instance count tracking (deployed/
enabled/disabled) to SiteHealthReport via DeploymentManagerActor. Updated
Health Dashboard UI to show instance counts per site. All metrics flow
through the existing health report pipeline via ClusterClient.
DeploymentManagerActor deserialized connection config JSON as
Dictionary<string, string>, which silently failed on non-string values
like {"publishInterval":1000}. The OPC UA adapter then fell back to
localhost:4840 (unreachable in Docker). Now uses JsonDocument to handle
any JSON value type. OPC PLC Simulator connects successfully.
Areas management is a design concern, not admin. Moved Areas page
authorization from RequireAdmin to RequireDesign, moved nav link from
Admin to Design section, updated ManagementActor role check. Added
GET /logout endpoint (was 404, now redirects to login). Improved Sign
Out button visibility in sidebar next to username.
DataConnectionActor now calls UpdateConnectionHealth() on state
transitions (Connecting/Connected/Reconnecting) and UpdateTagResolution()
on connection establishment. DataConnectionManagerActor calls
RemoveConnection() on actor removal. Health reports now include
data connection statuses when instances are deployed with bindings.
Central and site clusters now communicate via ClusterClient/
ClusterClientReceptionist instead of direct ActorSelection. Both
CentralCommunicationActor and SiteCommunicationActor are registered
with their cluster's receptionist. Central creates one ClusterClient
per site using NodeA/NodeB contact points from the DB. Sites configure
multiple CentralContactPoints for automatic failover between central
nodes. ISiteClientFactory enables test injection.
