dohertj2/ScadaBridge
1
0
Fork 0
Code Issues 2 Pull Requests Actions Packages Projects Releases Wiki Activity

refactor(auditlog+sitecallaudit): delete the _injectedRepository test ctors — register row 61

Browse Source 
AuditLogIngestActor and SiteCallAuditActor each exposed constructors taking a scoped
repository and caching it on a field for the actor's lifetime. Safe only because no
production call site picked those overloads — i.e. one call site away from re-introducing
the row-45 captive-dependency bug — which is why CaptiveDependencyGuard had to allowlist
both fields.

Preferred fix applied: the test fixtures fold onto the IServiceProvider overload with a
stub provider, and the constructors are gone.

Product deletions:
  - AuditLogIngestActor: the (IAuditLogRepository, ILogger) ctor, the _injectedRepository
    field and its branch in OnIngestAsync, and OnCachedTelemetryWithoutDualWriteAsync (the
    "this ctor cannot dual-write" fallback handler that only that ctor installed).
    _serviceProvider is now non-nullable and the null-forgiving operators are gone.
  - SiteCallAuditActor: both test ctors (repo-only, and repo + reconciliation
    collaborators), the _injectedRepository field and its four branches, and the private
    SerializedRepository class. _backgroundTimersEnabled went with them: only the deleted
    repo-only ctor ever set it false, so it was a constant and its two guards are removed.
    The reconciliation tick is still gated — on the collaborators being resolvable, which
    is how production reaches the same state.

Test fixtures (12 call sites): two small TestSupport helpers build the stub providers.
SiteCallAuditActor's production wiring resolves a FRESH repository per scope, so the
serializing decorator existed only to protect fixtures that share one DbContext-backed
instance across every scope — it moves into the test project as
SerializingSiteCallAuditRepository, which is where the hazard is actually created.
Fixtures that relied on it keep it. The relay test's ICentralAuditWriter now arrives via
the provider, exactly as production resolves it.

Behaviour is preserved on the paths under test: the deleted audit ctor passed
redactor:null / failureCounter:null, and a provider registering neither resolves both to
null, so the SafeDefaultAuditRedactor fallback and counter-free logging are unchanged. No
fixture on the folded sites sends IngestCachedTelemetryCommand, so swapping to the real
dual-write handler is inert. The SiteCallAudit purge timer now arms in fixtures that
previously had timers off, but every one of them uses default purge settings whose first
tick is 5 minutes out — nothing fires inside a test.

One reconciliation test was renamed and rewritten rather than deleted:
TestCtor_RepositoryOnly_DoesNotStartReconciliationTick became
NoReconciliationCollaboratorsRegistered_DoesNotStartReconciliationTick — the same
guarantee, expressed the way production actually reaches it.

Guard: the allowlist is now EMPTY. The staleness self-test (every entry must still match a
flagged field) is replaced by two stronger pins — Allowlist_IsEmpty asserts nothing is
exempt at all, and FormerlyAllowlistedActors_NowScanClean scans both actors with the
allowlist DISABLED, so the empty allowlist is evidence the fields are gone rather than
evidence the scan stopped looking.

Negative control: re-adding an ISiteCallAuditRepository field to SiteCallAuditActor fails
both FormerlyAllowlistedActors_NowScanClean and the real
Central_NoActorHoldsCaptiveScopedState scan.

Verification: solution builds clean. AuditLog.Tests 3/3 (383), SiteCallAudit.Tests 3/3
(66), Host.Tests 2/2 (505), IntegrationTests 2/2 (115), Communication.Tests 2/2 (717).
This commit is contained in:
Joseph Doherty
2026-08-15 17:49:39 -04:00
parent 7a7d345b89
commit 92fe5acca7
15 changed files with 474 additions and 452 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/ZB.MOM.WW.ScadaBridge.AuditLog/Central/AuditLogIngestActor.cs
+57 -94
View File
@@ -42,13 +42,14 @@ namespace ZB.MOM.WW.ScadaBridge.AuditLog.Central;
/// placeholder.
/// </para>
/// <para>
/// Two constructors exist for a deliberate reason: the test ctor injects a
/// concrete <see cref="IAuditLogRepository"/> against a per-test MSSQL fixture
/// (the only way to verify the ingest-timestamp stamp + duplicate-key
/// idempotency end to end), while the production host wiring registers the
/// actor as a cluster singleton and must therefore resolve the repository —
/// which is a scoped EF Core service — from a fresh DI scope per message.
/// Mirroring the Notification Outbox actor's pattern.
/// There is exactly ONE constructor, taking an <see cref="IServiceProvider"/>:
/// the actor is a cluster singleton and the repository is a scoped EF Core
/// service, so it must be resolved from a fresh DI scope per message and never
/// held on a field. Mirroring the Notification Outbox actor's pattern. Tests
/// that need a concrete repository (e.g. the MSSQL-backed fixture verifying the
/// ingest-timestamp stamp and duplicate-key idempotency) supply a stub provider
/// that hands one out, rather than a second constructor that would cache it —
/// see register row 61.
/// </para>
/// </remarks>
public class AuditLogIngestActor : ReceiveActor
@@ -100,41 +101,25 @@ public class AuditLogIngestActor : ReceiveActor
/// </remarks>
internal static readonly TimeSpan IngestFallbackBudget = TimeSpan.FromSeconds(5);
private readonly IServiceProvider? _serviceProvider;
private readonly IAuditLogRepository? _injectedRepository;
private readonly IServiceProvider _serviceProvider;
private readonly ILogger<AuditLogIngestActor> _logger;
/// <summary>
/// Test-mode constructor — injects a concrete repository instance whose
/// lifetime exceeds the test, so the actor reuses the same instance across
/// every message. Used by the MSSQL-backed TestKit fixture.
/// </summary>
/// <param name="repository">Audit log repository instance shared across all messages.</param>
/// <param name="logger">Logger for ingest diagnostics.</param>
public AuditLogIngestActor(
IAuditLogRepository repository,
ILogger<AuditLogIngestActor> logger)
{
ArgumentNullException.ThrowIfNull(repository);
ArgumentNullException.ThrowIfNull(logger);
_injectedRepository = repository;
_logger = logger;
ReceiveAsync<IngestAuditEventsCommand>(OnIngestAsync);
// The single-repository test ctor cannot service the dual-write —
// it has no SiteCalls repo and no DbContext. The handler still
// registers (so callers don't dead-letter) but replies empty so the
// test surface stays explicit about what this ctor supports.
ReceiveAsync<IngestCachedTelemetryCommand>(OnCachedTelemetryWithoutDualWriteAsync);
}
/// <summary>
/// Production constructor — resolves <see cref="IAuditLogRepository"/> from
/// The only constructor — resolves <see cref="IAuditLogRepository"/> from
/// a fresh DI scope per message because the repository is a scoped EF Core
/// service registered by <c>AddConfigurationDatabase</c>. The actor itself
/// is a long-lived cluster singleton, so it cannot hold a scope across
/// messages.
///
/// <para>
/// Register row 61: a second, repository-taking "test-mode" constructor used
/// to cache a scoped repository on a field for the actor's whole lifetime.
/// Nothing in production selected it, but it left the captive-dependency bug
/// one call site away and had to be allowlisted in the Host tests'
/// <c>CaptiveDependencyGuard</c>. It is gone; test fixtures build a stub
/// <see cref="IServiceProvider"/> instead, which keeps the actor's own
/// contract "resolve per scope, hold nothing".
/// </para>
/// </summary>
/// <param name="serviceProvider">Root service provider used to open a fresh scope per message.</param>
/// <param name="logger">Logger for ingest diagnostics.</param>
@@ -169,59 +154,51 @@ public class AuditLogIngestActor : ReceiveActor
var accepted = new List<Guid>(cmd.Events.Count);
// Resolve the repository for the whole batch — one DbContext per
// message, mirroring NotificationOutboxActor. The injected-repository
// mode (test ctor) skips the scope entirely.
// message, mirroring NotificationOutboxActor.
// The IAuditRedactor is also resolved from the per-message scope when
// one is available so the row is truncated + redacted before
// InsertIfNotExistsAsync. The single-repository test ctor has no
// service provider — it falls through with no redactor, which preserves
// the small-payload assumptions baked into the existing fixtures.
// InsertIfNotExistsAsync. A composition root that registers no redactor
// (test fixtures) falls back to the SafeDefault inside
// IngestWithRepositoryAsync rather than to pass-through.
// Use CreateAsyncScope + await using so scoped EF Core
// services (IAsyncDisposable DbContexts) dispose asynchronously
// without blocking on sync Dispose() of pending connection cleanup.
if (_injectedRepository is not null)
//
// Guard scope-creation + repository resolution in a
// try/catch, mirroring OnCachedTelemetryAsync. A transient DI /
// DbContext-factory fault (pooled-context init, SQL-connection
// exhaustion, a resolution race during host churn) would otherwise
// propagate out of the ReceiveAsync handler, trip the parent's
// supervision, and RESTART this central singleton over a transient
// fault — dropping the captured reply so the site's Ask times out.
// Best-effort audit must never wedge the singleton: log, optionally
// bump the failure counter, and still reply with whatever was
// accepted (empty on an up-front scope-resolution throw) so the
// site keeps its rows Pending and retries on the next drain.
try
{
await IngestWithRepositoryAsync(_injectedRepository, redactor: null, failureCounter: null, cmd, nowUtc, accepted)
await using var scope = _serviceProvider.CreateAsyncScope();
var repository = scope.ServiceProvider.GetRequiredService<IAuditLogRepository>();
var redactor = scope.ServiceProvider.GetService<IAuditRedactor>();
// Central health counter is best-effort —
// unregistered (test composition roots) means the per-row catch
// simply logs without surfacing on the health dashboard.
var failureCounter = scope.ServiceProvider.GetService<ICentralAuditWriteFailureCounter>();
await IngestWithRepositoryAsync(repository, redactor, failureCounter, cmd, nowUtc, accepted)
.ConfigureAwait(false);
}
else
catch (Exception ex)
{
// Guard scope-creation + repository resolution in a
// try/catch, mirroring OnCachedTelemetryAsync. A transient DI /
// DbContext-factory fault (pooled-context init, SQL-connection
// exhaustion, a resolution race during host churn) would otherwise
// propagate out of the ReceiveAsync handler, trip the parent's
// supervision, and RESTART this central singleton over a transient
// fault — dropping the captured reply so the site's Ask times out.
// Best-effort audit must never wedge the singleton: log, optionally
// bump the failure counter, and still reply with whatever was
// accepted (empty on an up-front scope-resolution throw) so the
// site keeps its rows Pending and retries on the next drain.
try
{
await using var scope = _serviceProvider!.CreateAsyncScope();
var repository = scope.ServiceProvider.GetRequiredService<IAuditLogRepository>();
var redactor = scope.ServiceProvider.GetService<IAuditRedactor>();
// Central health counter is best-effort —
// unregistered (test composition roots) means the per-row catch
// simply logs without surfacing on the health dashboard.
var failureCounter = scope.ServiceProvider.GetService<ICentralAuditWriteFailureCounter>();
await IngestWithRepositoryAsync(repository, redactor, failureCounter, cmd, nowUtc, accepted)
.ConfigureAwait(false);
}
catch (Exception ex)
{
// Scope creation or a required-service resolution threw before
// (or while) processing the batch. Surface a sustained fault on
// the dashboard if the counter is registered, but never let the
// throw escape the handler and restart the singleton.
try { _serviceProvider!.GetService<ICentralAuditWriteFailureCounter>()?.Increment(); }
catch { /* counter must never throw — defence in depth */ }
_logger.LogError(
ex,
"Audit event batch ingest failed before/while resolving the repository scope; replying with {Accepted} accepted row(s). The site keeps unaccepted rows Pending and retries on the next drain.",
accepted.Count);
}
// Scope creation or a required-service resolution threw before
// (or while) processing the batch. Surface a sustained fault on
// the dashboard if the counter is registered, but never let the
// throw escape the handler and restart the singleton.
try { _serviceProvider.GetService<ICentralAuditWriteFailureCounter>()?.Increment(); }
catch { /* counter must never throw — defence in depth */ }
_logger.LogError(
ex,
"Audit event batch ingest failed before/while resolving the repository scope; replying with {Accepted} accepted row(s). The site keeps unaccepted rows Pending and retries on the next drain.",
accepted.Count);
}
replyTo.Tell(new IngestAuditEventsReply(accepted));
@@ -357,7 +334,7 @@ public class AuditLogIngestActor : ReceiveActor
try
{
await using var scope = _serviceProvider!.CreateAsyncScope();
await using var scope = _serviceProvider.CreateAsyncScope();
var auditRepo = scope.ServiceProvider.GetRequiredService<IAuditLogRepository>();
var siteCallRepo = scope.ServiceProvider.GetRequiredService<ISiteCallAuditRepository>();
var dbContext = scope.ServiceProvider.GetRequiredService<ScadaBridgeDbContext>();
@@ -544,18 +521,4 @@ public class AuditLogIngestActor : ReceiveActor
return true;
}
/// <summary>
/// Fallback handler installed on the single-repository test ctor — that
/// ctor has no DbContext and no <see cref="ISiteCallAuditRepository"/>, so
/// it cannot service the dual-write. Logs a warning and replies with an
/// empty ack so callers fall through to their retry path.
/// </summary>
private Task OnCachedTelemetryWithoutDualWriteAsync(IngestCachedTelemetryCommand cmd)
{
_logger.LogWarning(
"AuditLogIngestActor received IngestCachedTelemetryCommand on the single-repository ctor; dual-write requires the IServiceProvider ctor. Replying with empty ack ({Count} entries).",
cmd.Entries.Count);
Sender.Tell(new IngestCachedTelemetryReply(Array.Empty<Guid>()));
return Task.CompletedTask;
}
}
src/ZB.MOM.WW.ScadaBridge.SiteCallAudit/SiteCallAuditActor.cs
+37 -256
View File
@@ -87,8 +87,7 @@ public class SiteCallAuditActor : ReceiveActor
/// </summary>
private const int MaxReconciliationPagesPerTick = 50;
private readonly IServiceProvider? _serviceProvider;
private readonly ISiteCallAuditRepository? _injectedRepository;
private readonly IServiceProvider _serviceProvider;
private readonly SiteCallAuditOptions _options;
private readonly ILogger<SiteCallAuditActor> _logger;
@@ -108,28 +107,17 @@ public class SiteCallAuditActor : ReceiveActor
/// (<see cref="IPullSiteCallsClient"/>) and the site list
/// (<see cref="ISiteEnumerator"/>). On the production path these are
/// resolved once from the root <see cref="IServiceProvider"/> (central
/// singletons registered by <c>AddAuditLogCentralReconciliationClient</c>);
/// in the test path they are injected directly. They are <c>null</c> when
/// the actor was built via the repo-only test ctor — in that case the
/// reconciliation tick is NOT started (see <see cref="StartReconciliationTimer"/>).
/// The purge tick, by contrast, does NOT depend on these collaborators:
/// it needs only the repository, so it is armed by
/// <see cref="_backgroundTimersEnabled"/> alone (see <see cref="StartPurgeTimer"/>).
/// singletons registered by <c>AddAuditLogCentralReconciliationClient</c>).
/// They are <c>null</c> when the composition root did not register them (a
/// minimal host, or a test fixture that deliberately omits them) — in that
/// case the reconciliation tick is NOT started (see
/// <see cref="StartReconciliationTimer"/>). The purge tick, by contrast, does
/// NOT depend on these collaborators: it needs only the repository, so it is
/// always armed (see <see cref="StartPurgeTimer"/>).
/// </summary>
private readonly IPullSiteCallsClient? _pullClient;
private readonly ISiteEnumerator? _siteEnumerator;
/// <summary>
/// Master switch for the two background schedulers (reconciliation + purge),
/// set <c>true</c> by the production and reconciliation ctors and <c>false</c>
/// by the repo-only MSSQL test ctor. The purge timer is
/// gated on THIS flag rather than on the reconciliation collaborators, so a
/// host that omits the reconciliation client still purges (no unbounded
/// central <c>SiteCalls</c> growth) while the MSSQL read/upsert tests stay
/// free of any scheduled side effects.
/// </summary>
private readonly bool _backgroundTimersEnabled;
/// <summary>
/// Per-site reconciliation watermark — the composite
/// <c>(UpdatedAtUtc, TrackedOperationId)</c> keyset of the highest row seen
@@ -195,98 +183,23 @@ public class SiteCallAuditActor : ReceiveActor
private IActorRef? _centralCommunication;
/// <summary>
/// Test-mode constructor — injects a concrete repository instance whose
/// lifetime exceeds the test, so the actor reuses the same instance
/// across every message. Used by the MSSQL-backed TestKit fixture.
/// An optional <paramref name="options"/> lets a test pin the stuck/KPI
/// windows; when omitted the production defaults apply.
/// <para>
/// This ctor disables BOTH background timers (sets
/// <see cref="_backgroundTimersEnabled"/> to <c>false</c>) and injects no
/// reconciliation client/enumerator, so neither the reconciliation tick nor
/// the purge tick fires — the MSSQL-backed read/upsert tests must see no
/// scheduled side effects (no phantom pulls, no background purge).
/// </para>
/// </summary>
/// <param name="repository">Concrete repository instance to use for all messages.</param>
/// <param name="logger">Logger for diagnostics and error reporting.</param>
/// <param name="options">Optional configuration overrides; production defaults apply when null.</param>
/// <param name="auditWriter">Optional central direct-write audit sink for operator relay actions; null disables the relay audit row.</param>
public SiteCallAuditActor(
ISiteCallAuditRepository repository,
ILogger<SiteCallAuditActor> logger,
SiteCallAuditOptions? options = null,
ICentralAuditWriter? auditWriter = null)
{
ArgumentNullException.ThrowIfNull(repository);
ArgumentNullException.ThrowIfNull(logger);
_injectedRepository = new SerializedRepository(repository);
_logger = logger;
_options = options ?? new SiteCallAuditOptions();
_auditWriter = auditWriter;
// Repo-only MSSQL test ctor: keep BOTH background timers off so the
// read/upsert tests see no scheduled side effects.
_backgroundTimersEnabled = false;
RegisterHandlers();
}
/// <summary>
/// Test-mode constructor for the reconciliation tick (Piece A) — injects a
/// concrete repository PLUS the two reconciliation collaborators directly,
/// so the per-site self-heal pull is unit-testable in-memory without a DI
/// container or a live gRPC channel. Background timers are enabled, so the
/// purge tick starts (it needs only the repository) and the reconciliation
/// tick starts too because the client + enumerator are present — see
/// <see cref="StartReconciliationTimer"/> / <see cref="StartPurgeTimer"/>.
/// </summary>
/// <param name="repository">Concrete repository instance used for upserts and purges.</param>
/// <param name="siteEnumerator">Enumerates the sites to reconcile each tick.</param>
/// <param name="pullClient">Pull client used to fetch changed rows from each site.</param>
/// <param name="logger">Logger for diagnostics and error reporting.</param>
/// <param name="options">Optional configuration overrides; production defaults apply when null.</param>
/// <remarks>
/// Public (not internal) because Akka's default <c>ActivatorProducer</c>
/// instantiates the actor via reflection with public-only binding flags —
/// an internal ctor yields a <c>MissingMethodException</c> at actor
/// creation. Distinguished from the production <see cref="IServiceProvider"/>
/// ctor by its concrete-collaborator parameter list; only the test project
/// (or a host that hand-resolves the collaborators) constructs it this way.
/// </remarks>
public SiteCallAuditActor(
ISiteCallAuditRepository repository,
ISiteEnumerator siteEnumerator,
IPullSiteCallsClient pullClient,
ILogger<SiteCallAuditActor> logger,
SiteCallAuditOptions? options = null)
{
ArgumentNullException.ThrowIfNull(repository);
ArgumentNullException.ThrowIfNull(siteEnumerator);
ArgumentNullException.ThrowIfNull(pullClient);
ArgumentNullException.ThrowIfNull(logger);
_injectedRepository = new SerializedRepository(repository);
_siteEnumerator = siteEnumerator;
_pullClient = pullClient;
_logger = logger;
_options = options ?? new SiteCallAuditOptions();
// Reconciliation test ctor: collaborators present, so both timers arm
// (the reconciliation tick uses the collaborators; the purge tick needs
// only the repository).
_backgroundTimersEnabled = true;
RegisterHandlers();
}
/// <summary>
/// Production constructor — resolves <see cref="ISiteCallAuditRepository"/>
/// The only constructor — resolves <see cref="ISiteCallAuditRepository"/>
/// from a fresh DI scope per message because the repository is a scoped EF
/// Core service registered by <c>AddConfigurationDatabase</c>. The actor
/// itself is a long-lived cluster singleton, so it cannot hold a scope
/// across messages.
///
/// <para>
/// Register row 61: two further "test-mode" constructors used to take a
/// concrete repository and cache it (wrapped in a serializing decorator) on
/// a field for the actor's whole lifetime. Nothing in production selected
/// them, but they left the captive-dependency bug one call site away and had
/// to be allowlisted in the Host tests' <c>CaptiveDependencyGuard</c>. They
/// are gone; test fixtures build a stub <see cref="IServiceProvider"/>
/// instead — and own the serializing decorator themselves, since it exists
/// only to compensate for a fixture handing the SAME repository instance to
/// every scope. Production never needed it: each scope gets its own.
/// </para>
/// </summary>
/// <param name="serviceProvider">DI service provider used to create a scoped repository per message.</param>
/// <param name="options">Actor configuration (stuck threshold, KPI interval, relay timeout).</param>
@@ -321,12 +234,6 @@ public class SiteCallAuditActor : ReceiveActor
_pullClient = serviceProvider.GetService<IPullSiteCallsClient>();
_siteEnumerator = serviceProvider.GetService<ISiteEnumerator>();
// Production path: background timers run. The purge tick is armed
// unconditionally here (it needs only the repository); the reconciliation
// tick additionally requires its collaborators and logs a Warning if
// they were not registered.
_backgroundTimersEnabled = true;
RegisterHandlers();
}
@@ -462,11 +369,6 @@ public class SiteCallAuditActor : ReceiveActor
/// </summary>
private void StartReconciliationTimer()
{
if (!_backgroundTimersEnabled)
{
return;
}
if (_pullClient is null || _siteEnumerator is null)
{
_logger.LogWarning(
@@ -488,23 +390,16 @@ public class SiteCallAuditActor : ReceiveActor
/// <summary>
/// Starts the daily purge tick. The purge needs ONLY the
/// repository — never the reconciliation collaborators — so it is gated on
/// <see cref="_backgroundTimersEnabled"/> alone, NOT on
/// repository — never the reconciliation collaborators — so it is armed
/// unconditionally, NOT gated on
/// <see cref="_pullClient"/> / <see cref="_siteEnumerator"/>. This decouples
/// the daily terminal-row purge from the reconciliation client: a host that
/// registers Site Call Audit without the reconciliation client still purges,
/// so the central <c>SiteCalls</c> table can never grow unbounded just
/// because the self-heal puller is absent. Only the repo-only MSSQL test ctor
/// (background timers off) skips it, keeping the read/upsert tests free of
/// scheduled side effects.
/// because the self-heal puller is absent.
/// </summary>
private void StartPurgeTimer()
{
if (!_backgroundTimersEnabled)
{
return;
}
var interval = _options.ResolvedPurgeInterval;
_purgeTimer = Context.System.Scheduler.ScheduleTellRepeatedlyCancelable(
// Short first tick (min(interval, 5min)): purges are idempotent, and a full-interval
@@ -532,20 +427,11 @@ public class SiteCallAuditActor : ReceiveActor
var replyTo = Sender;
var id = cmd.SiteCall.TrackedOperationId;
// Scope-per-message mirrors AuditLogIngestActor — production EF
// repository is scoped; the injected-repository mode (tests) skips
// the scope entirely.
IServiceScope? scope = null;
ISiteCallAuditRepository repository;
if (_injectedRepository is not null)
{
repository = _injectedRepository;
}
else
{
scope = _serviceProvider!.CreateScope();
repository = scope.ServiceProvider.GetRequiredService<ISiteCallAuditRepository>();
}
// Scope-per-message mirrors AuditLogIngestActor — the EF repository is
// a scoped service, so it is resolved fresh here and never held.
IServiceScope? scope = _serviceProvider.CreateScope();
ISiteCallAuditRepository repository =
scope.ServiceProvider.GetRequiredService<ISiteCallAuditRepository>();
try
{
@@ -624,15 +510,8 @@ public class SiteCallAuditActor : ReceiveActor
// thread on a synchronous Dispose() of pending connection cleanup — the tick
// holds the scope across many awaited UpsertAsync calls. Mirrors the sibling
// SiteAuditReconciliationActor.OnTickAsync. ResolveRepository() (sync Dispose)
// is retained for the synchronous message-handler paths. In the injected-
// repository test path there is no scope to open and the test repo is reused.
if (_injectedRepository is not null)
{
await ReconcileSitesAsync(sites, client, _injectedRepository).ConfigureAwait(false);
return;
}
await using var scope = _serviceProvider!.CreateAsyncScope();
// is retained for the synchronous message-handler paths.
await using var scope = _serviceProvider.CreateAsyncScope();
var repository = scope.ServiceProvider.GetRequiredService<ISiteCallAuditRepository>();
await ReconcileSitesAsync(sites, client, repository).ConfigureAwait(false);
}
@@ -860,15 +739,8 @@ public class SiteCallAuditActor : ReceiveActor
// Open the scope INLINE with CreateAsyncScope + await using
// so the scoped EF Core repository (an IAsyncDisposable DbContext) disposes
// asynchronously rather than blocking the Akka dispatcher thread on a
// synchronous Dispose(). Mirrors SiteAuditReconciliationActor; the
// injected-repository test path reuses the test repo with no scope.
if (_injectedRepository is not null)
{
await PurgeWithRepositoryAsync(_injectedRepository, threshold).ConfigureAwait(false);
return;
}
await using var scope = _serviceProvider!.CreateAsyncScope();
// synchronous Dispose(). Mirrors SiteAuditReconciliationActor.
await using var scope = _serviceProvider.CreateAsyncScope();
var repository = scope.ServiceProvider.GetRequiredService<ISiteCallAuditRepository>();
await PurgeWithRepositoryAsync(repository, threshold).ConfigureAwait(false);
}
@@ -1454,19 +1326,13 @@ public class SiteCallAuditActor : ReceiveActor
/// <summary>
/// Resolves an <see cref="ISiteCallAuditRepository"/> for one read message.
/// In test mode the injected instance is returned with a null scope; in
/// production a fresh DI scope is created and returned so the caller can
/// dispose it once the read completes — the same scope-per-message pattern
/// as <see cref="OnUpsertAsync"/>.
/// A fresh DI scope is created and returned so the caller can dispose it
/// once the read completes — the same scope-per-message pattern as
/// <see cref="OnUpsertAsync"/>.
/// </summary>
private (IServiceScope? Scope, ISiteCallAuditRepository Repository) ResolveRepository()
{
if (_injectedRepository is not null)
{
return (null, _injectedRepository);
}
var scope = _serviceProvider!.CreateScope();
var scope = _serviceProvider.CreateScope();
return (scope, scope.ServiceProvider.GetRequiredService<ISiteCallAuditRepository>());
}
@@ -1565,91 +1431,6 @@ public class SiteCallAuditActor : ReceiveActor
private PurgeComplete() { }
}
/// <summary>
/// Serializing wrapper applied to a repository handed in through the
/// test constructors. One call at a time, in arrival order.
/// </summary>
/// <remarks>
/// <para>
/// In production every consumer of this actor's repository gets its OWN DI
/// scope — one per message, one per background pass — so the off-mailbox
/// reconciliation and purge passes can safely run alongside mailbox handlers:
/// separate scopes mean separate <c>DbContext</c>s. The injected-repository
/// constructors break that assumption: ONE instance (typically wrapping one
/// <c>DbContext</c>) is shared by the mailbox handlers AND the background
/// passes, and <c>DbContext</c> forbids concurrent operations — "A second
/// operation was started on this context instance" is a hard fault, and the
/// interleaving is nondeterministic, so it surfaces as a flaky test rather
/// than a reliable one.
/// </para>
/// <para>
/// Serializing at the CALL, not around the whole pass, is deliberate: it
/// removes the concurrency hazard while preserving the property the
/// off-mailbox passes exist for — a long drain (blocked in a network pull,
/// holding no repository call) still lets ingest, query and KPI messages be
/// answered. Suspending the mailbox for the duration of a pass would have
/// been simpler and would have invalidated exactly those regression tests.
/// Production is untouched: it never constructs this type.
/// </para>
/// </remarks>
private sealed class SerializedRepository : ISiteCallAuditRepository
{
private readonly ISiteCallAuditRepository _inner;
private readonly SemaphoreSlim _gate = new(1, 1);
public SerializedRepository(ISiteCallAuditRepository inner) => _inner = inner;
public Task UpsertAsync(SiteCall siteCall, CancellationToken ct = default) =>
RunAsync(() => _inner.UpsertAsync(siteCall, ct));
public Task<SiteCall?> GetAsync(TrackedOperationId id, CancellationToken ct = default) =>
RunAsync(() => _inner.GetAsync(id, ct));
public Task<IReadOnlyList<SiteCall>> QueryAsync(
SiteCallQueryFilter filter, SiteCallPaging paging, CancellationToken ct = default) =>
RunAsync(() => _inner.QueryAsync(filter, paging, ct));
public Task<int> PurgeTerminalAsync(DateTime olderThanUtc, CancellationToken ct = default) =>
RunAsync(() => _inner.PurgeTerminalAsync(olderThanUtc, ct));
public Task<SiteCallKpiSnapshot> ComputeKpisAsync(
DateTime stuckCutoffUtc, DateTime deliveredSinceUtc, CancellationToken ct = default) =>
RunAsync(() => _inner.ComputeKpisAsync(stuckCutoffUtc, deliveredSinceUtc, ct));
public Task<IReadOnlyList<SiteCallSiteKpiSnapshot>> ComputePerSiteKpisAsync(
DateTime stuckCutoffUtc, DateTime deliveredSinceUtc, CancellationToken ct = default) =>
RunAsync(() => _inner.ComputePerSiteKpisAsync(stuckCutoffUtc, deliveredSinceUtc, ct));
public Task<IReadOnlyList<SiteCallNodeKpiSnapshot>> ComputePerNodeKpisAsync(
DateTime stuckCutoffUtc, DateTime deliveredSinceUtc, CancellationToken ct = default) =>
RunAsync(() => _inner.ComputePerNodeKpisAsync(stuckCutoffUtc, deliveredSinceUtc, ct));
private async Task RunAsync(Func<Task> call)
{
await _gate.WaitAsync().ConfigureAwait(false);
try
{
await call().ConfigureAwait(false);
}
finally
{
_gate.Release();
}
}
private async Task<T> RunAsync<T>(Func<Task<T>> call)
{
await _gate.WaitAsync().ConfigureAwait(false);
try
{
return await call().ConfigureAwait(false);
}
finally
{
_gate.Release();
}
}
}
}
/// <summary>