dohertj2/lmxopcua
1
0
Fork 0
Code Issues 43 Pull Requests Actions Packages Projects Releases Wiki Activity

fix(opcua): replay kept scripted conditions after a rebuild; fit window values onto re-shaped nodes (Lane B fix round 3)

Browse Source 
N3 (critical): the node manager carries only NATIVE conditions' state over a
full rebuild; scripted conditions come back fresh and rely on the engine's
re-assert, which since the off-thread plan is handled in the window and lands
on the discarded instance. On a rebuild the condition replay scope is now
AddedConditions plus every enabled scripted condition in the new composition
(native kept conditions stay carried-over, not replayed). The test sink again
models the real rebuild: it clears non-native condition state only.

N4: a surgical in-place shape edit (DataType / IsArray / ArrayLength) resets the
raw node and its UNS mirror to BadWaitingForInitialData without ensuring them,
so the window's value was not replayed and the two NodeIds disagreed. The
applier records re-shaped nodes (MaterialisationRecording.Reshaped); their
deferred value replays only once fitted to the new shape (VariableShape.TryFit:
as is, or converted invariantly; never an old-typed value - WriteValues assigns
the value verbatim with no type check), otherwise it is dropped at Debug and the
node keeps waiting. Status-only entries are not replayed onto a re-shaped node.

Minors: m1/m2 documented on _replayScope (degraded-path only); m3 corrected the
false carry-over premise in the scope doc (the ScriptedAlarmHostActor ordering
remark is outside this lane - F8); m4 the replay Debug log reports alarm-state
and alarm-quality counts again.

Refs #732
Claude-Session: https://claude.ai/code/session_01Ub9hGmYLQvn9XETzw7CMsT
This commit is contained in:
Joseph Doherty
2026-10-22 15:46:29 -04:00
parent 9c8e70c5b7
commit ff4e210a27
4 changed files with 457 additions and 46 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/Server/ZB.MOM.WW.OtOpcUa.Runtime/OpcUa/OpcUaPublishActor.cs
+62 -11
View File
@@ -187,8 +187,21 @@ public sealed class OpcUaPublishActor : ReceiveActor, IWithTimers
/// <list type="bullet">
/// <item>Variables: after a FULL rebuild every one (each was recreated, its live write landed on the
/// discarded node); otherwise exactly those the apply asked the sink to ensure (the applier records them).</item>
/// <item>Conditions: those in the applied composition but not the previous one — on a rebuild too, because the
/// node manager carries a kept condition's state over to its re-materialised instance.</item>
/// <item>Variables an in-place attribute update RE-SHAPED (review N4): the node manager reset their value to
/// BadWaitingForInitialData, so the window's value is replayed — but only once fitted to the new shape
/// (<see cref="VariableShape.TryFit"/>); a value that does not fit is dropped and the node left waiting.
/// The raw node and its UNS mirror are re-shaped by the same edit, so both get the same treatment.</item>
/// <item>Conditions: those in the applied composition but not the previous one; after a full rebuild ALSO
/// every kept SCRIPTED condition (review N3) — the node manager carries only NATIVE conditions' state over
/// a rebuild, so a scripted one is recreated fresh and needs the window's state. A kept native condition
/// is restored from the carry-over (which already holds the live write and any SDK ack), so it is not
/// replayed. Accepted cost on the scripted path: an SDK ack on the discarded instance in the window is
/// lost with it; the engine's own ack emission (a scripted ack routes to the engine) restores it.</item>
/// <item>Known gaps (degraded path only, review m1 / m2): a degraded non-rebuild apply does not advance
/// <c>_lastApplied</c>, so the next apply counts its conditions as added again and may replay onto them;
/// and the empty-plan repair of a degraded composition scopes no condition (the composition is
/// unchanged), so a condition the repair newly places gets no window state — writes to it between the
/// two applies were already dropped before this replay existed.</item>
/// </list>
/// (Precise "did the live write land" would need the sink to report it; the production sink chain runs through
/// Commons' DeferredAddressSpaceSink, outside this lane — follow-up F7.)
@@ -198,10 +211,12 @@ public sealed class OpcUaPublishActor : ReceiveActor, IWithTimers
private sealed record ReplayScope(
bool AllVariables,
IReadOnlySet<(AddressSpaceRealm, string)> Variables,
IReadOnlyDictionary<(AddressSpaceRealm, string), VariableShape> Reshaped,
IReadOnlySet<(AddressSpaceRealm, string)> Conditions)
{
private static readonly IReadOnlySet<(AddressSpaceRealm, string)> None = new HashSet<(AddressSpaceRealm, string)>();
public static readonly ReplayScope Nothing = new(false, None, None);
public static readonly ReplayScope Nothing =
new(false, None, new Dictionary<(AddressSpaceRealm, string), VariableShape>(), None);
}
/// <summary>True while a rebuild's parse / plan step is running off-thread. Rebuilds are applied strictly in
@@ -851,8 +866,28 @@ public sealed class OpcUaPublishActor : ReceiveActor, IWithTimers
private void ReplayDeferredWrites(ReplayScope scope)
{
if (_deferred.IsEmpty) return;
var (values, statuses) = _deferred.VariableWrites(
key => scope.AllVariables || scope.Variables.Contains(key));
var (allValues, allStatuses) = _deferred.VariableWrites(
key => scope.AllVariables || scope.Variables.Contains(key) || scope.Reshaped.ContainsKey(key));
var values = new List<ValueWrite>(allValues.Count);
var statuses = new List<StatusWrite>(allStatuses.Count);
var unfit = 0;
foreach (var v in allValues)
{
if (!scope.Reshaped.TryGetValue((v.Realm, v.NodeId), out var shape)) { values.Add(v); continue; }
if (shape.TryFit(v.Value, out var fitted)) { values.Add(v with { Value = fitted }); continue; }
unfit++;
_log.Debug(
"OpcUaPublish: a value deferred for re-shaped {Realm} node {Node} does not fit its new shape ({DataType}, array={IsArray}) — not replayed; the node waits for the driver's next value",
v.Realm, v.NodeId, shape.DataType, shape.IsArray);
}
foreach (var st in allStatuses)
{
// A status-only annotation onto a re-shaped (value-less) node would serve a null value under it; the
// node's honest state is BadWaitingForInitialData until a value of the new shape arrives.
if (scope.Reshaped.ContainsKey((st.Realm, st.NodeId))) { unfit++; continue; }
statuses.Add(st);
}
if (values.Count > 0)
{
try { _sink.WriteValues(values); }
@@ -865,22 +900,27 @@ public sealed class OpcUaPublishActor : ReceiveActor, IWithTimers
catch (Exception ex) { _log.Warning(ex, "OpcUaPublish: replaying {Count} deferred status writes threw", statuses.Count); }
}
var alarms = 0;
foreach (var alarm in _deferred.Alarms)
{
if (!scope.Conditions.Contains((alarm.Realm, alarm.AlarmNodeId))) continue;
alarms++;
try { _sink.WriteAlarmCondition(alarm.AlarmNodeId, alarm.State, alarm.TimestampUtc, alarm.Realm); }
catch (Exception ex) { _log.Warning(ex, "OpcUaPublish: replaying a deferred alarm write threw for {Node}", alarm.AlarmNodeId); }
}
var qualities = 0;
foreach (var quality in _deferred.AlarmQualities)
{
if (!scope.Conditions.Contains((quality.Realm, quality.AlarmNodeId))) continue;
qualities++;
try { _sink.WriteAlarmQuality(quality.AlarmNodeId, quality.StatusCode, quality.TimestampUtc, quality.Realm); }
catch (Exception ex) { _log.Warning(ex, "OpcUaPublish: replaying a deferred alarm-quality write threw for {Node}", quality.AlarmNodeId); }
}
_log.Debug("OpcUaPublish: replayed {Values} value and {Statuses} status writes (of {Held} held) deferred during a rebuild plan",
values.Count, statuses.Count, _deferred.Count);
_log.Debug(
"OpcUaPublish: replayed {Values} value, {Statuses} status, {Alarms} alarm-state and {Qualities} alarm-quality writes (of {Held} held; {Unfit} not fitting a re-shaped node) deferred during a rebuild plan",
values.Count, statuses.Count, alarms, qualities, _deferred.Count, unfit);
}
/// <summary>
@@ -1019,17 +1059,28 @@ public sealed class OpcUaPublishActor : ReceiveActor, IWithTimers
private (bool Degraded, string? Reason) ApplyPlan(PlanReady ready)
{
var previous = _lastApplied;
var ensured = _applier!.BeginRecordingEnsuredVariables();
var recording = _applier!.BeginRecording();
try
{
var result = ApplyPlanCore(ready, out var rebuilt);
// What this apply created (see _replayScope). A throwing pass still scopes to what it did create.
_replayScope = new ReplayScope(rebuilt, ensured, AddedConditions(previous, ready.Composition));
// What this apply created or reset (see _replayScope). A throwing pass still scopes to what it did.
var conditions = AddedConditions(previous, ready.Composition);
if (rebuilt)
{
// Review N3: a full rebuild recreates every SCRIPTED condition fresh — the node manager carries only
// NATIVE conditions' state over (OtOpcUaNodeManager.RebuildAddressSpace). The engine's re-assert
// (or any transition) handled in the window landed on the discarded instance, so every kept scripted
// condition must be replayed onto its new one. Native kept conditions stay "added only".
foreach (var a in ready.Composition.EquipmentScriptedAlarms)
if (a.Enabled) conditions.Add((AddressSpaceRealm.Uns, a.ScriptedAlarmId));
}
_replayScope = new ReplayScope(rebuilt, recording.Ensured, recording.Reshaped, conditions);
return result;
}
finally
{
_applier.EndRecordingEnsuredVariables();
_applier.EndRecording();
}
}