perf(misc): cached hot-path lookups, bounded observer queue, alarm-priority stream path

WP2.6 (arch-review remediation, cross-cutting misc):
- SiteExternalSystemRepository: name/ID-indexed ExternalSystemDefinitionCache replaces
  the fetch-all + reverse-map scan on every by-ID/method lookup; loaded once per
  redeploy, invalidated by DeploymentManagerActor after HandleDeployArtifacts applies
  external-system changes. Static JsonSerializerOptions for method-list parsing.
- Inbound API: short-TTL ApiMethodCache fronts the per-request ApiMethod repository
  fetch; invalidated by name via the existing ScriptArtifactChangeSubscriber/
  IScriptArtifactChangeBus pipeline, self-healing via TTL for changes the bus
  doesn't cover (e.g. Management API edits).
- StoreAndForward: the cached-call audit-observer queue — the one unbounded channel
  left in the system — is now bounded (ObserverQueueCapacity, default 10,000) with
  DropOldest overflow and a dropped-notification counter.
- SiteStreamManager: alarm state changes now travel a dedicated publish
  source/broadcast hub, isolated from the (far higher-volume) attribute path, so an
  attribute storm can no longer evict a pending alarm transition; the alarm hand-off
  queue is bounded with a drop counter surfaced on the site health report
  (SiteStreamAlarmDropCount via the new SiteStreamAlarmDropReporter), and publishing
  is skipped entirely at zero subscribers on either path.
- CLI ManagementHttpClient: explicit 30s HttpClient.Timeout on the shared
  construction (was the 100s framework default), overridable via
  SCADABRIDGE_HTTP_TIMEOUT_SECONDS.

Deviation: the failback-probe heartbeat item is NOT included — its only viable
surface (CentralChannelProvider.cs / heartbeat consumers) lives entirely in the
Communication project, explicitly off-limits to this work package this phase.

Tests: SiteRuntime.Tests (550), InboundAPI.Tests (278), StoreAndForward.Tests (133),
CLI.Tests (390), HealthMonitoring.Tests (97) — all green after full solution build.
This commit is contained in:
Joseph Doherty
2026-08-14 20:59:43 -04:00
parent ee193cd2bb
commit a212283104
32 changed files with 1361 additions and 111 deletions
@@ -159,4 +159,86 @@ public class SiteStreamManagerTests : TestKit, IDisposable
_streamManager.RemoveSubscriber(probe.Ref);
Assert.Equal(0, _streamManager.SubscriptionCount);
}
// ── WP2.6d: separate alarm publish path, drop counter, skip-at-zero-subscribers ──
/// <summary>
/// WP2.6d: alarm state changes now travel a dedicated publish source, isolated from
/// the (far higher-volume) attribute path — a burst of attribute events for OTHER
/// instances, interleaved with an alarm-only subscriber's events, must not cause any
/// alarm to be lost.
/// </summary>
[Fact]
public void PublishAlarmStateChanged_SurvivesConcurrentAttributeStorm_ForUnrelatedInstances()
{
var alarmProbe = CreateTestProbe();
_streamManager.SubscribeSiteAlarms(alarmProbe.Ref);
// A storm of attribute events for a DIFFERENT instance — none of which the
// alarm subscriber is even listening to — interleaved with alarm events. Before
// WP2.6d these shared one upstream buffer; now they are fully separate sources.
for (var i = 0; i < 500; i++)
{
_streamManager.PublishAttributeValueChanged(new AttributeValueChanged(
"NoisyPump", "Temperature", "Temperature", i.ToString(), "Good", DateTimeOffset.UtcNow));
}
_streamManager.PublishAlarmStateChanged(new AlarmStateChanged(
"Pump1", "HighTemp", AlarmState.Active, 1, DateTimeOffset.UtcNow));
var received = alarmProbe.ExpectMsg<AlarmStateChanged>(TimeSpan.FromSeconds(3));
Assert.Equal("Pump1", received.InstanceUniqueName);
}
/// <summary>
/// WP2.6d: with zero subscribers of any kind, PublishAlarmStateChanged/
/// PublishAttributeValueChanged must be no-ops — no exception, and (for alarms) the
/// event never reaches the bounded hand-off queue, so
/// <see cref="SiteStreamManager.AlarmPublishDroppedCount"/> stays at zero rather than
/// counting events nobody could ever have received anyway.
/// </summary>
[Fact]
public void Publish_WithNoSubscribers_IsNoOp_AndDoesNotCountAsDropped()
{
_streamManager.PublishAlarmStateChanged(new AlarmStateChanged(
"Pump1", "HighTemp", AlarmState.Active, 1, DateTimeOffset.UtcNow));
_streamManager.PublishAttributeValueChanged(new AttributeValueChanged(
"Pump1", "Temperature", "Temperature", "1", "Good", DateTimeOffset.UtcNow));
Assert.Equal(0, _streamManager.AlarmPublishDroppedCount);
// Publishing resumes working normally once a subscriber exists.
var probe = CreateTestProbe();
_streamManager.SubscribeSiteAlarms(probe.Ref);
_streamManager.PublishAlarmStateChanged(new AlarmStateChanged(
"Pump1", "HighTemp", AlarmState.Active, 2, DateTimeOffset.UtcNow));
probe.ExpectMsg<AlarmStateChanged>(TimeSpan.FromSeconds(3));
}
/// <summary>
/// WP2.6d: once the bounded alarm hand-off queue (the exact factory
/// <see cref="SiteStreamManager.Initialize"/> wires into the drop-count callback) is
/// full, a further write evicts the oldest queued item and fires the drop callback.
/// Exercised directly against the queue factory — deterministic, no actor
/// system/async pump race involved (see the factory's own doc comment).
/// </summary>
[Fact]
public void CreateAlarmPublishQueue_OverCapacity_DropsOldestAndInvokesCallback()
{
var dropped = 0;
var queue = SiteStreamManager.CreateAlarmPublishQueue(capacity: 2, () => Interlocked.Increment(ref dropped));
// Nothing reads from this queue, so all five writes stay purely upstream —
// capacity 2 means the 3rd/4th/5th writes must each evict the oldest.
for (var i = 0; i < 5; i++)
{
var wrote = queue.Writer.TryWrite(new AlarmStateChanged(
"Pump1", "HighTemp", AlarmState.Active, i, DateTimeOffset.UtcNow));
Assert.True(wrote); // DropOldest TryWrite always succeeds once capacity > 0
}
Assert.Equal(3, dropped);
Assert.Equal(2, queue.Reader.Count);
}
}