perf(worker): event hot-path allocation + flush cuts (WRK-06/11/12, IPC-15)

WRK-06: MxStatusProxyConverter caches the four resolved FieldInfo per status
type in a static ConcurrentDictionary (the GetField metadata scan ran 4x per
status per event on the STA path). GetValue+Convert.ToInt32 still run per event
(late-bound RCW). Exceptions byte-identical: missing-field message unchanged
(ResolveField, not cached on throw via GetOrAdd); null-value message unchanged.

WRK-11: MxAccessEventQueue.Enqueue takes ownership of the passed MxEvent -
stamps WorkerSequence/WorkerTimestamp on it in place and enqueues it, no
Clone(). Audited all 3 callers (base/alarm event sinks, provider-mode handler):
each builds a fresh event per Enqueue, none reuse it. MxAccessValueCache.Set now
deep-copies its retained Value/SourceTimestamp/Statuses so the cache snapshot
never aliases the queue-owned (later serialized) event. Net: alarm/other events
clone nothing (was full clone); data-change clones payload-only.

WRK-12: WorkerFrameWriter coalesces the flush across a drained batch - each
frame is written but not flushed individually; one FlushAsync after the batch,
then all written frames complete. Preserves the written+flushed completion
contract; a burst of N events costs 1 flush, not N. On write failure the whole
in-flight batch + queue fail so no caller hangs.

IPC-15 (doc): the multi-event WorkerEnvelope body remains unimplemented (wire
still carries one event per worker_event frame); gateway.md Performance section
now distinguishes the shipped flush-coalescing from that deferred proto change.

net48-safe (no init/records; readonly struct cache entry). Worker builds x86
only - verification on windev. Tests added: converter cache-reuse, queue
ownership-transfer, value-cache snapshot independence, writer batch-flush-once.

Claude-Session: https://claude.ai/code/session_01DMXXvNuPekkkrTEyPNxEkW
This commit is contained in:
Joseph Doherty
2026-07-09 15:52:49 -04:00
parent 73ce824f6c
commit f61c816acf
9 changed files with 303 additions and 28 deletions
@@ -8,6 +8,17 @@ namespace ZB.MOM.WW.MxGateway.Worker.MxAccess;
/// <summary>
/// Thread-safe queue for MxAccess events with capacity overflow and fault tracking.
/// </summary>
/// <remarks>
/// Ownership invariant: <see cref="Enqueue"/> takes ownership of the
/// <see cref="MxEvent"/> passed to it — it stamps the worker sequence and
/// timestamp on that same instance and enqueues it directly, without
/// cloning (WRK-11). Every caller must therefore pass a freshly built
/// <see cref="MxEvent"/> that it does not retain, reuse, or mutate after the
/// call returns. All production callers (MxAccessBaseEventSink,
/// MxAccessAlarmEventSink, AlarmCommandHandler) build a new event per
/// Enqueue via the mapper and satisfy this; the value cache stores its own
/// independent snapshot (see <see cref="MxAccessValueCache.Set"/>).
/// </remarks>
public sealed class MxAccessEventQueue
{
/// <summary>
@@ -110,8 +121,11 @@ public sealed class MxAccessEventQueue
/// <summary>
/// Enqueues an MxAccess event, assigning a sequence number and timestamp.
/// Takes ownership of <paramref name="mxEvent"/>: the sequence and timestamp
/// are stamped on that same instance and it is enqueued without cloning, so
/// the caller must pass a freshly built event it does not reuse afterwards.
/// </summary>
/// <param name="mxEvent">MXAccess event to enqueue.</param>
/// <param name="mxEvent">Freshly built MXAccess event to enqueue; ownership transfers to the queue.</param>
public void Enqueue(MxEvent mxEvent)
{
if (mxEvent is null)
@@ -132,13 +146,17 @@ public sealed class MxAccessEventQueue
throw new MxAccessEventQueueOverflowException(capacity);
}
MxEvent queuedEvent = mxEvent.Clone();
queuedEvent.WorkerSequence = ++lastEventSequence;
queuedEvent.WorkerTimestamp = Timestamp.FromDateTime(DateTime.UtcNow);
// WRK-11: stamp the sequence/timestamp on the caller's own event and
// enqueue that same instance under the lock instead of cloning. See
// the ownership invariant on the class summary — the caller hands the
// event over exclusively, so a defensive Clone() here is pure
// overhead on the hottest path.
mxEvent.WorkerSequence = ++lastEventSequence;
mxEvent.WorkerTimestamp = Timestamp.FromDateTime(DateTime.UtcNow);
WorkerEvent workerEvent = new()
{
Event = queuedEvent,
Event = mxEvent,
};
events.Enqueue(workerEvent);
}
@@ -40,6 +40,20 @@ public sealed class MxAccessValueCache
throw new ArgumentNullException(nameof(mxEvent));
}
// WRK-11: the event sink no longer clones before enqueue, so the passed
// mxEvent is the very instance handed to the outbound queue. Deep-copy
// the value/timestamp/statuses payload we retain here so the cache's
// snapshot stays independent of the enqueued (and later serialized)
// event — the two must never share mutable protobuf sub-messages.
// Value is always set for OnDataChange; SourceTimestamp may be unset when
// the source timestamp could not be parsed, so both are cloned only when
// present. The null-forgiving result matches CachedValue's non-null-
// annotated parameters, which already accepted a runtime-null value or
// timestamp before WRK-11 (the ternary keeps the compiler's null-state
// from poisoning to maybe-null, which a plain null check would do).
MxValue cachedValue = mxEvent.Value is null ? null! : mxEvent.Value.Clone();
Timestamp cachedTimestamp = mxEvent.SourceTimestamp is null ? null! : mxEvent.SourceTimestamp.Clone();
long key = CreateItemKey(serverHandle, itemHandle);
lock (syncRoot)
{
@@ -49,10 +63,10 @@ public sealed class MxAccessValueCache
entries[key] = new CachedValue(
nextVersion,
mxEvent.Value,
cachedValue,
mxEvent.Quality,
mxEvent.SourceTimestamp,
mxEvent.Statuses);
cachedTimestamp,
mxEvent.Statuses.Clone());
}
}