Templates page is now a tree-only browser; editing happens on a dedicated
TemplateEdit page. Drag-drop is replaced by context-menu Move-to-Folder.
TreeView gains Bootstrap Icons (chevron + per-kind glyphs), ancestor guide
lines, defined hover/selected/focus tokens, and Escape-dismisses-menu per
the new Visual Design Guide (V1-V7) in Component-TreeView.md.
The previous fix tried to defer page-side RevealNode to the second
render so TreeView's async sessionStorage load could finish first. In
practice Blazor Server didn't always fire a second OnAfterRenderAsync
on the page after the deep-link load, so the reveal never ran.
Real fix: change TreeView's storage-load to UNION the restored keys
with whatever's already in _expandedKeys, instead of REPLACING. That
way the page can call RevealNode whenever it wants and the storage
restore can't clobber the reveal regardless of completion order. The
page-side guard simplifies back to a one-shot reveal on first render.
Semantic note: if a deep-link reveal expands an ancestor that the user
had previously collapsed, the deep link wins. Intentional — the URL
expresses the navigation intent.
Both page.OnAfterRenderAsync(firstRender=true) and
TreeView.OnAfterRenderAsync(firstRender=true) ran concurrently:
- Page called RevealNode → added ancestor keys to _expandedKeys
- TreeView awaited treeviewStorage.load → replaced _expandedKeys with
the persisted set (often empty if user collapsed before navigating)
Whichever JS interop completed second won. When TreeView won, the deep-link
reveal silently lost. Gate the reveal on firstRender==false so it runs
strictly after TreeView's restore is done.
Without stopPropagation, dropping a template onto a folder fires both
OnDrop(folder) and OnDropOnRoot via event bubbling. The two async handlers
race on the same scoped DbContext, which is not thread-safe — the second
throws ObjectDisposedException and tears down the Blazor circuit. Surfaced
during browser smoke testing via JS-dispatched DragEvent sequence.
Smoke testing revealed two issues introduced by the modal extraction commit:
1. ErrorMessage / InitialName / TemplateName parameters on the dialog
components were passed as bare strings (e.g. ErrorMessage="_newFolderError")
instead of dereferenced C# expressions (ErrorMessage="@_newFolderError").
Razor treats unquoted-but-not-@-prefixed values to string parameters as
string literals — so the error block rendered the literal field name in
red whenever the modal opened. Non-string parameters (int/IEnumerable)
were fine since Razor treats those as C# expressions by default.
2. The Templates header + 4-button toolbar shared one flex row, but at
col-md-4 / col-lg-3 width the buttons overflowed into the right-column
empty-state area. Stack title above a full-width btn-group instead.
Captures the pre-existing entity drift from commit 04af039 (rename
Configuration to PrimaryConfiguration, add BackupConfiguration and
FailoverRetryCount), which was committed without a corresponding
migration. Generating this here unblocks the upcoming AddTemplateFolders
migration on the templates-folder-hierarchy branch.
Replaces the current /design/templates list view with a Wonderware-style
template toolbox: nested TemplateFolder entity, FolderId on Template,
composition children as inline tree leaves, persistent split-pane with
editor on the right, context menus + drag-drop reorg.
The lmxproxy workspace was relocated to deprecated/ in 9dccf8e but the
Dockerfile still tried to COPY lmxproxy/src/ZB.MOM.WW.LmxProxy.Client/,
breaking docker build. Remove the two stale COPY lines.
Restore inside the docker build was failing because TreatWarningsAsErrors
promotes NU1902/NU1903/NU1904 (vulnerable package warnings) to errors.
Bump the flagged packages to advisory-free versions:
- MailKit 4.15.1 -> 4.16.0 (GHSA-9j88-vvj5-vhgr)
- Microsoft.AspNetCore.DataProtection.EFCore 10.0.5 -> 10.0.7 (GHSA-9mv3-2cwr-p262, transitively pulls fixed System.Security.Cryptography.Xml — GHSA-37gx-xxp4-5rgx, GHSA-w3x6-4m5h-cxqf)
- OpenTelemetry.Api (transitive via Akka.Hosting) 1.9.0 -> 1.15.3 (GHSA-g94r-2vxg-569j, GHSA-8785-wc3w-h8q6) — added as a direct PackageReference in ScadaLink.Host to override the Akka.Hosting pin
To resolve the NU1605 downgrade chain triggered by DataProtection.EFCore
10.0.7 (which transitively requires Microsoft.EntityFrameworkCore >= 10.0.7
and friends), bump every Microsoft.* 10.0.5 reference across src/ and
tests/ to 10.0.7 in lockstep.
Central nodes crashed at startup with `CREATE DATABASE permission denied`
when MSSQL accepted connections before recovering user databases —
DB_ID(@db) returned null, so EF Core's MigrateAsync fell through to
SqlServerDatabaseCreator.CreateAsync. The non-privileged app login then
failed CREATE DATABASE and the host terminated with FTL, leaving Traefik's
/health/active probe unable to find an upstream ("no available server" at
localhost:9000).
Add MigrationHelper.WaitForDatabaseReadyAsync that polls
Database.CanConnectAsync() for up to 60s before invoking MigrateAsync, and
thread an ILogger through so retry attempts surface in normal logs. This
removes the startup race without requiring depends_on across compose stacks
or granting dbcreator to the app login.
LmxProxy is no longer needed. Moved the entire lmxproxy/ workspace, DCL
adapter files, and related docs to deprecated/. Removed LmxProxy registration
from DataConnectionFactory, project reference from DCL, protocol option from
UI, and cleaned up all requirement docs.
HandleConnectionQualityChanged now publishes AttributeValueChanged events
to the SiteStreamManager for all affected attributes. This ensures the
central UI debug view updates in real-time when a data connection
disconnects and attributes go bad quality.
Only publishes to the stream — does NOT notify script or alarm actors,
since the value hasn't changed and firing scripts/alarms on quality-only
changes would cause spurious evaluations.
Key subscriptions by unique subscriptionId instead of sessionId to prevent
overwrites when the same session calls Subscribe multiple times (e.g. DCL
StaleTagMonitor). Add session-to-subscription reverse lookup for cleanup.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Previously, failover only triggered when ConnectAsync failed consecutively.
If a connection succeeded but went stale quickly (e.g., heartbeat timeout),
the failure counter reset on each successful connect and failover never
triggered.
Added a separate _consecutiveUnstableDisconnects counter that increments
when a connection lasts less than StableConnectionThreshold (60s) before
disconnecting. When this counter reaches failoverRetryCount, the actor
fails over to the backup endpoint. Stable connections (lasting >60s)
reset this counter.
The original connection-failure failover path is unchanged.
Existing site databases created before the primary/backup data connections
feature lack the backup_configuration and failover_retry_count columns.
Added TryAddColumnAsync migration that runs on startup after table creation.
Composable StaleTagMonitor class in Commons fires a Stale event when no
value is received within a configurable max silence period. Integrated
into both LmxProxyDataConnection and OpcUaDataConnection adapters via
optional HeartbeatTagPath/HeartbeatMaxSilence connection config keys.
When stale, the adapter fires Disconnected triggering the standard
reconnect cycle. 10 unit tests cover timer behavior.
Add NodeStatus record, IClusterNodeProvider interface, and AkkaClusterNodeProvider
that queries Akka cluster membership for all site-role nodes. HealthReportSender
populates ClusterNodes before each report. UI shows a row per node with
hostname, Online/Offline badge, and Primary/Standby badge. Falls back to
single-node display if ClusterNodes is not populated.
- AkkaHostedService: SetNodeHostname from NodeOptions
- DataConnectionActor: UpdateConnectionEndpoint on state transitions,
track per-tag quality counts and UpdateTagQuality on value changes
- HealthReportSender: query StoreAndForwardStorage for parked message count
- StoreAndForwardStorage: add GetParkedMessageCountAsync()
New fields in SiteHealthReport: NodeHostname, DataConnectionEndpoints
(primary/secondary), DataConnectionTagQuality (good/bad/uncertain),
ParkedMessageCount. New collector methods to populate them.
Health dashboard redesigned to match mockup: Nodes | Data Connections
(with per-connection tag quality) | Instances + S&F Buffers | Error
Counts + Parked Messages. Site names resolved from repository.
