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

fix(site): DeployedSecretsMirror — confirm-then-ledger LWW apply, per-row isolation, cache invalidation

Browse Source 
Fix round 1 (review task-9-review.md), all rulings:

1. (Important 1) ApplyReplicatedAsync is no longer trusted blindly. Every apply is
   confirmed by a store.GetAsync read-back before ledgering: a tombstoned local row
   (routine remove-then-re-add) is safely re-stamped past the tombstone and re-applied
   (confirmed neither UpdatedUtc nor Revision is AEAD-bound in AesGcmEnvelopeCipher, so
   the bump cannot break decryption); a live local row with different ciphertext (a
   strictly-newer out-of-band secret) is left untouched, never ledgered, never verified
   against, and reported with a distinct Failed message so a later central tombstone can
   never delete it via this mirror.
2. (Important 2) Per-row try/catch (rethrow only on ct.IsCancellationRequested) so one
   poisoned row (e.g. an unrecognised ContentType from a newer central) no longer blocks
   every row after it.
3. (Important 3) Optional ISecretCacheInvalidator, invalidated after every successful
   apply/delete and before verify, so a rotated/removed secret's plaintext cannot be
   served stale for the remainder of the resolver's TTL.
4. (Minor 4) Loop and VerifyAsync OperationCanceledException handling now filters on
   ct/stoppingToken.IsCancellationRequested specifically, rather than swallowing every OCE.
5. (Minor 5) Class remarks note the mirror assumes a node-local ISecretStore; Task 11
   will reject SqlServer secrets replication on the Site role.
6. (Minor 6/7) Added cancellation + concurrent-callers tests, dropped the redundant
   FailFor (tests now use ThrowFor/ReturnNullFor to exercise distinct decrypt-failure
   paths), and Dispose() now disposes the gate SemaphoreSlim.

Verify: dotnet test tests/ZB.MOM.WW.ScadaBridge.SiteRuntime.Tests --filter FullyQualifiedName~DeployedSecretsMirrorTests (24/24); full SiteRuntime.Tests suite 876/876.
This commit is contained in:
Joseph Doherty
2026-09-28 19:07:48 -04:00
parent 25c25ccef9
commit 4d7d6456e3
2 changed files with 626 additions and 145 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/ZB.MOM.WW.ScadaBridge.SiteRuntime/Secrets/DeployedSecretsMirror.cs
+283 -74
View File
@@ -11,11 +11,12 @@ namespace ZB.MOM.WW.ScadaBridge.SiteRuntime.Secrets;
/// The outcome of one <see cref="DeployedSecretsMirror.ReconcileAsync"/> pass: the secret
/// names applied (inserted or updated in this node's own secret store), the names removed
/// (deleted because central tombstoned them and this node had previously mirrored them), and
/// a human-readable failure message per live secret that did not decrypt on this node.
/// a human-readable failure message per live secret that could not be applied and/or verified
/// on this node.
/// </summary>
/// <param name="Applied">Names applied (inserted or updated) this pass.</param>
/// <param name="Removed">Names deleted this pass because they were previously mirrored and are now tombstoned upstream.</param>
/// <param name="Failed">One message per live secret that failed to decrypt on this node — never includes secret values.</param>
/// <param name="Failed">One message per live secret that failed to apply and/or decrypt on this node — never includes secret values.</param>
public sealed record MirrorResult(
IReadOnlyList<string> Applied, IReadOnlyList<string> Removed, IReadOnlyList<string> Failed)
{
@@ -39,71 +40,132 @@ public sealed record MirrorResult(
/// master key.
/// </para>
/// <para>
/// <b>Apply semantics (<see cref="ISecretStore.ApplyReplicatedAsync"/>).</b> The store applies
/// last-writer-wins on (<c>UpdatedUtc</c>, <c>Revision</c>) — see
/// <c>ZB.MOM.WW.Secrets.Abstractions.SecretLastWriterWins</c> — and a row that is not
/// <em>strictly</em> newer than the local row is silently ignored. Consequence worth knowing:
/// if an operator ever creates an out-of-band secret directly in a site node's local store
/// under the SAME name as a deployed secret, and stamps it with a later <c>UpdatedUtc</c> (or
/// an equal <c>UpdatedUtc</c> and a higher <c>Revision</c>) than the deployed row central last
/// shipped, that out-of-band secret WINS locally — every subsequent reconcile pass's
/// <c>ApplyReplicatedAsync</c> call for that name is a silent no-op until central ships a
/// still-newer revision. This mirror does not attempt to detect or warn about that case (it
/// has no way to distinguish "already applied, unchanged" from "out-of-band override in
/// place" — both look like "nothing to do" from here); deliberately not adding that
/// complexity per the design's scope.
/// <b>Apply semantics and the confirm-then-ledger fix (fix round 1, review Important #1).</b>
/// <see cref="ISecretStore.ApplyReplicatedAsync"/> applies last-writer-wins strictly on
/// (<c>UpdatedUtc</c>, <c>Revision</c>) — see
/// <see cref="ZB.MOM.WW.Secrets.Abstractions.SecretLastWriterWins"/> — and a row that is not
/// strictly newer than the local row is silently ignored, so this mirror never trusts the call
/// blindly: every apply is followed by an <see cref="ISecretStore.GetAsync"/> read-back, and
/// <c>mirrored_secrets</c> is ledgered only when the local row is confirmed LIVE and its
/// ciphertext (<c>Ciphertext</c> + <c>WrappedDek</c>) matches what was just shipped. Two
/// non-ledgered outcomes are possible, both reported as a <see cref="MirrorResult.Failed"/>
/// entry (never as the decrypt-failure message, which would misdiagnose either case) and never
/// overwritten a second way:
/// <list type="bullet">
/// <item>
/// <b>The local row is a tombstone</b> (this node previously deleted it — e.g. the site
/// stopped referencing it, then referenced it again unchanged at central, so the shipped row's
/// <c>UpdatedUtc</c> is older than the tombstone's). Re-stamping is safe here: neither
/// <c>UpdatedUtc</c> nor <c>Revision</c> is bound into the AEAD associated data — confirmed
/// against <c>ZB.MOM.WW.Secrets.Crypto.AesGcmEnvelopeCipher</c>, whose body layer binds only
/// <c>SecretName.Value</c> and whose wrap layer binds only <c>KekId</c> — so the mirror
/// re-applies the exact shipped ciphertext with <c>UpdatedUtc = max(now, tombstone.UpdatedUtc +
/// 1 tick)</c> and <c>Revision = tombstone.Revision + 1</c>, which beats the tombstone under LWW
/// without touching anything the cipher verifies. If that confirmed re-apply still does not
/// read back live and matching (a concurrent conflict), the row is reported failed instead of
/// silently left broken.
/// </item>
/// <item>
/// <b>The local row is LIVE with different ciphertext</b> — an out-of-band secret created
/// directly on this node under the same name, stamped strictly newer than what central shipped.
/// It wins LWW legitimately. This mirror does NOT overwrite it, does NOT ledger it (so a later
/// central tombstone of the name can never delete someone's out-of-band secret via this
/// mirror), and does NOT run the decrypt-verify against it (verifying would trivially "succeed"
/// against the operator's own value and mask that the deployed revision never took effect).
/// </item>
/// </list>
/// Ledgering always uses the SHIPPED artifact's own revision (not any bumped local revision)
/// so a later pass's "is this ledgered at the shipped revision" check stays correct.
/// </para>
/// <para>
/// <b>Per-row isolation (review Important #2).</b> Each row's apply/verify body runs in its own
/// try/catch; an exception — an unrecognised <c>ContentType</c> from a newer central, a
/// per-row store fault, a malformed name — is caught, reported as a <see cref="MirrorResult.Failed"/>
/// entry naming the secret and the exception type/message (never a value), logged with the same
/// de-spam rule, and does not stop the rest of the pass. Only an exception attributable to the
/// caller's own cancellation (<c>ct.IsCancellationRequested</c>) propagates.
/// </para>
/// <para>
/// <b>Cache invalidation (review Important #3).</b> The concrete resolver
/// (<c>DefaultSecretResolver</c>) caches decrypted plaintext for a TTL and also implements
/// <see cref="ISecretCacheInvalidator"/>; every ZB.MOM.WW.Secrets write path invalidates after a
/// write for exactly this reason. This mirror does the same: after every successful apply
/// (including a tombstone re-materialization) or delete, and before the verify step, it calls
/// <see cref="ISecretCacheInvalidator.Invalidate"/> (injected optionally — <c>null</c> if the
/// host never registered one) so neither the verify nor a script running immediately afterward
/// can observe a stale pre-rotation plaintext for up to the cache TTL.
/// </para>
/// <para>
/// <b>Node-local store assumption (review Minor #5).</b> This mirror assumes the site's
/// <see cref="ISecretStore"/> is node-local (plain local SQLite, or the gRPC-hub follower half,
/// both of which keep writes local). It is NOT safe to run against a site node whose store is
/// decorated by SQL-Server secrets REPLICATION (<c>Secrets:Replication:Mode=SqlServer</c>) — its
/// <c>DeleteAsync</c> publishes the tombstone to the shared hub, so a site merely un-referencing
/// a secret would delete it fleet-wide. Task 11 rejects that configuration on the Site role;
/// no code change is made here.
/// </para>
/// <para>
/// <b>Resolve semantics and the resolver cache.</b> The trial-resolve verify step goes through
/// <see cref="ISecretResolver.GetAsync"/>, which the concrete resolver caches for a TTL
/// (<c>ResolveCacheTtl</c>, 30 s in the shipped implementation). If this node's master key
/// changes (e.g. a KEK rotation) between an earlier successful resolve and a reconcile pass
/// run within the cache TTL, the verify can return the STALE cached plaintext rather than
/// re-decrypting — that is still a successful decrypt of <em>a</em> value under the key that
/// produced the cache entry, which is all this check is trying to prove (this node's master
/// key can decrypt what central most recently shipped), so a cache hit is an acceptable false
/// negative for the "did the key change" question and not a bug in this mirror.
/// (<c>ResolveCacheTtl</c>, 30 s in the shipped implementation). Because every successful apply
/// or delete now invalidates that cache first (see above), the remaining exposure is narrow: a
/// KEK rotation on this node's OWN master key (not a secret rotation at central) landing inside
/// the cache window of an unrelated resolve. That residual case is still an acceptable false
/// negative for the "did the key change" question — the cache hit still proves this node once
/// decrypted a value with a key that produced that cache entry — and is not a bug in this
/// mirror.
/// </para>
/// <para>
/// <b>Error-log de-spam.</b> A secret that keeps failing to decrypt would otherwise log an
/// Error line every reconcile pass forever. This mirror logs Error only the FIRST time a given
/// (name, revision, kekId) tuple fails on this node; identical repeats log at Debug instead.
/// The moment that same name resolves successfully (a fixed master key, or a newer revision
/// central re-shipped), the remembered tuple is cleared, so a LATER failure — even for the
/// exact same revision/kekId, e.g. a flapping key — logs at Error again. This state is
/// in-memory only (per node process), which matches the ledger's own node-local scope.
/// <b>Error-log de-spam.</b> A secret that keeps failing (to apply OR to decrypt) would
/// otherwise log an Error line every reconcile pass forever. This mirror logs Error only the
/// FIRST time a given failure fingerprint (kind + revision + relevant detail, e.g. KekId for a
/// decrypt failure) occurs for a name on this node; identical repeats log at Debug instead. The
/// moment that same name applies/resolves successfully, the remembered fingerprint is cleared,
/// so a LATER failure — even one with an identical fingerprint — logs at Error again. This state
/// is in-memory only (per node process), which matches the ledger's own node-local scope.
/// </para>
/// </remarks>
public sealed class DeployedSecretsMirror : BackgroundService
{
private const string DeleteActor = "deployed-secrets-mirror";
private readonly SiteStorageService _storage;
private readonly ISecretStore _store;
private readonly ISecretResolver _resolver;
private readonly ISecretCacheInvalidator? _invalidator;
private readonly IOptions<SiteRuntimeOptions> _options;
private readonly ILogger<DeployedSecretsMirror> _logger;
private readonly TimeProvider _time;
private readonly SemaphoreSlim _gate = new(1, 1);
private bool _disposed;
/// <summary>
/// Per-secret-name record of the (Revision, KekId) tuple whose decrypt failure this node
/// has already logged at Error. Cleared for a name the moment it resolves successfully, so
/// a later failure — even an identical one — logs at Error again. In-memory only; not a
/// durable ledger, and deliberately so (it exists purely to de-spam this process's logs).
/// Per-secret-name record of the failure fingerprint this node has already logged at Error
/// (an apply-conflict or a decrypt failure — see the class remarks). Cleared for a name the
/// moment it applies/resolves successfully, so a later failure — even an identical one —
/// logs at Error again. In-memory only; not a durable ledger, and deliberately so (it exists
/// purely to de-spam this process's logs).
/// </summary>
private readonly Dictionary<string, (long Revision, string KekId)> _lastLoggedFailure = new(StringComparer.Ordinal);
private readonly Dictionary<string, string> _lastLoggedFailure = new(StringComparer.Ordinal);
/// <summary>Initializes a new instance of <see cref="DeployedSecretsMirror"/>.</summary>
/// <param name="storage">Site-local storage exposing <c>deployed_secrets</c> and the <c>mirrored_secrets</c> ledger.</param>
/// <param name="store">This node's own secret store.</param>
/// <param name="resolver">This node's own secret resolver, used to trial-decrypt each deployed secret.</param>
/// <param name="invalidator">
/// Optional cache invalidator for <paramref name="resolver"/> (the concrete
/// <c>DefaultSecretResolver</c> implements this itself; <c>null</c> if the host never
/// registered one, in which case no invalidation is attempted).
/// </param>
/// <param name="options">Supplies <see cref="SiteRuntimeOptions.DeployedSecretsMirrorInterval"/>.</param>
/// <param name="logger">Logger instance. Never logs secret values — names, revisions, and KEK ids only.</param>
/// <param name="time">Clock driving the reconcile loop's <see cref="PeriodicTimer"/>.</param>
/// <param name="time">Clock driving the reconcile loop's <see cref="PeriodicTimer"/> and the tombstone-recovery stamp.</param>
public DeployedSecretsMirror(
SiteStorageService storage,
ISecretStore store,
ISecretResolver resolver,
ISecretCacheInvalidator? invalidator,
IOptions<SiteRuntimeOptions> options,
ILogger<DeployedSecretsMirror> logger,
TimeProvider time)
@@ -111,6 +173,7 @@ public sealed class DeployedSecretsMirror : BackgroundService
_storage = storage ?? throw new ArgumentNullException(nameof(storage));
_store = store ?? throw new ArgumentNullException(nameof(store));
_resolver = resolver ?? throw new ArgumentNullException(nameof(resolver));
_invalidator = invalidator;
_options = options ?? throw new ArgumentNullException(nameof(options));
_logger = logger ?? throw new ArgumentNullException(nameof(logger));
_time = time ?? throw new ArgumentNullException(nameof(time));
@@ -118,10 +181,11 @@ public sealed class DeployedSecretsMirror : BackgroundService
/// <summary>
/// Runs one reconcile pass: applies every live <c>deployed_secrets</c> row into this
/// node's own secret store, deletes rows this node previously mirrored that are now
/// tombstoned upstream, and trial-decrypts every live row to verify it resolves under
/// this node's own master key. Safe to call concurrently with itself (and with the
/// hosted loop) — a single in-flight pass is enforced internally.
/// node's own secret store (confirmed by read-back before ledgering), deletes rows this
/// node previously mirrored that are now tombstoned upstream, and trial-decrypts every
/// live, confirmed-applied row to verify it resolves under this node's own master key.
/// Safe to call concurrently with itself (and with the hosted loop) — a single in-flight
/// pass is enforced internally.
/// </summary>
/// <param name="ct">A token to cancel the operation.</param>
/// <returns>The pass's <see cref="MirrorResult"/>.</returns>
@@ -140,38 +204,25 @@ public sealed class DeployedSecretsMirror : BackgroundService
foreach (var row in rows)
{
if (row.IsRemoved)
try
{
if (!mirrored.ContainsKey(row.Name))
if (row.IsRemoved)
{
// Never mirrored by this node (or already un-mirrored) — an
// out-of-band, same-named local secret is left alone.
continue;
await ProcessRemovedRowAsync(row, mirrored, removed, ct).ConfigureAwait(false);
}
else
{
await ProcessLiveRowAsync(row, mirrored, applied, failed, ct).ConfigureAwait(false);
}
await _store.DeleteAsync(new SecretName(row.Name), "deployed-secrets-mirror", ct)
.ConfigureAwait(false);
await _storage.RemoveMirroredSecretAsync(row.Name).ConfigureAwait(false);
removed.Add(row.Name);
_lastLoggedFailure.Remove(row.Name);
_logger.LogInformation(
"Removed deployed secret {SecretName} (no longer referenced by this site)",
row.Name);
continue;
}
if (!mirrored.TryGetValue(row.Name, out var mirroredRevision)
|| mirroredRevision != row.Secret.Revision)
catch (Exception ex) when (!ct.IsCancellationRequested)
{
await _store.ApplyReplicatedAsync(ToStored(row.Secret), ct).ConfigureAwait(false);
await _storage.SetMirroredSecretAsync(row.Name, row.Secret.Revision).ConfigureAwait(false);
applied.Add(row.Name);
_logger.LogInformation(
"Applied deployed secret {SecretName} r{Revision} (kek {KekId})",
row.Name, row.Secret.Revision, row.Secret.KekId);
ReportFailure(
row.Name,
$"error:{row.Secret.Revision}:{ex.GetType().Name}:{ex.Message}",
$"Secret '{row.Name}' failed to process on this node ({ex.GetType().Name}: {ex.Message})",
failed);
}
await VerifyAsync(row, failed, ct).ConfigureAwait(false);
}
var result = new MirrorResult(applied, removed, failed);
@@ -185,7 +236,7 @@ public sealed class DeployedSecretsMirror : BackgroundService
else
{
_logger.LogDebug(
"Deployed secrets mirror in sync ({Count} secret(s), {FailedCount} failing to decrypt)",
"Deployed secrets mirror in sync ({Count} secret(s), {FailedCount} failing)",
rows.Count(r => !r.IsRemoved), failed.Count);
}
@@ -197,12 +248,146 @@ public sealed class DeployedSecretsMirror : BackgroundService
}
}
/// <summary>
/// Handles a tombstoned <c>deployed_secrets</c> row: deletes it from this node's store, but
/// only if this node had previously ledgered it as mirrored — an out-of-band, same-named
/// local secret this mirror never applied is left alone.
/// </summary>
private async Task ProcessRemovedRowAsync(
DeployedSecretRow row, Dictionary<string, long> mirrored, List<string> removed, CancellationToken ct)
{
if (!mirrored.ContainsKey(row.Name))
{
return;
}
await _store.DeleteAsync(new SecretName(row.Name), DeleteActor, ct).ConfigureAwait(false);
_invalidator?.Invalidate(new SecretName(row.Name));
await _storage.RemoveMirroredSecretAsync(row.Name).ConfigureAwait(false);
removed.Add(row.Name);
_lastLoggedFailure.Remove(row.Name);
_logger.LogInformation(
"Removed deployed secret {SecretName} (no longer referenced by this site)",
row.Name);
}
/// <summary>
/// Handles a live <c>deployed_secrets</c> row: applies it (confirmed by read-back) when not
/// already ledgered at the shipped revision, then verifies it decrypts — unless the apply
/// outcome makes a verify meaningless (an out-of-band local secret won LWW), in which case a
/// single, distinct failure is reported instead.
/// </summary>
private async Task ProcessLiveRowAsync(
DeployedSecretRow row,
Dictionary<string, long> mirrored,
List<string> applied,
List<string> failed,
CancellationToken ct)
{
var name = new SecretName(row.Name);
var alreadyLedgered =
mirrored.TryGetValue(row.Name, out var mirroredRevision) && mirroredRevision == row.Secret.Revision;
if (!alreadyLedgered)
{
var outcome = await ApplyAndConfirmAsync(row, name, ct).ConfigureAwait(false);
switch (outcome)
{
case ApplyOutcome.Applied:
await _storage.SetMirroredSecretAsync(row.Name, row.Secret.Revision).ConfigureAwait(false);
_invalidator?.Invalidate(name);
applied.Add(row.Name);
_lastLoggedFailure.Remove(row.Name);
_logger.LogInformation(
"Applied deployed secret {SecretName} r{Revision} (kek {KekId})",
row.Name, row.Secret.Revision, row.Secret.KekId);
break;
case ApplyOutcome.LocalNewer:
ReportFailure(
row.Name,
$"outofband:{row.Secret.Revision}",
$"Secret '{row.Name}' exists on this node with a newer local value; " +
"the deployed revision was not applied",
failed);
return; // Do not verify — that would decrypt the operator's own value.
case ApplyOutcome.Unconfirmed:
ReportFailure(
row.Name,
$"unconfirmed:{row.Secret.Revision}",
$"Secret '{row.Name}' could not be confirmed applied on this node " +
"after a tombstone re-apply attempt",
failed);
return;
}
}
await VerifyAsync(row, failed, ct).ConfigureAwait(false);
}
/// <summary>The confirmed outcome of applying one shipped row into this node's store. See the class remarks.</summary>
private enum ApplyOutcome
{
/// <summary>The local row is now live and its ciphertext matches what was shipped.</summary>
Applied,
/// <summary>The local row is live with DIFFERENT ciphertext — an out-of-band secret that legitimately won LWW.</summary>
LocalNewer,
/// <summary>Neither of the above — a tombstone re-apply attempt did not read back as applied.</summary>
Unconfirmed,
}
/// <summary>
/// Calls <see cref="ISecretStore.ApplyReplicatedAsync"/> and confirms the outcome by reading
/// the row back — see the class remarks for why this mirror never trusts the call blindly.
/// </summary>
private async Task<ApplyOutcome> ApplyAndConfirmAsync(DeployedSecretRow row, SecretName name, CancellationToken ct)
{
var shipped = ToStored(row.Secret);
await _store.ApplyReplicatedAsync(shipped, ct).ConfigureAwait(false);
var local = await _store.GetAsync(name, ct).ConfigureAwait(false);
if (local is not null && !local.IsDeleted && MatchesShippedCiphertext(local, shipped))
{
return ApplyOutcome.Applied;
}
if (local is not null && local.IsDeleted)
{
// Neither UpdatedUtc nor Revision is bound into the AEAD associated data (confirmed
// against AesGcmEnvelopeCipher — only SecretName and KekId are AAD), so re-stamping
// strictly past the tombstone is safe: it changes nothing the cipher verifies.
var nowUtc = _time.GetUtcNow();
var tombstoneUtcPlusOneTick = local.UpdatedUtc.AddTicks(1);
var bumpedUpdatedUtc = nowUtc > tombstoneUtcPlusOneTick ? nowUtc : tombstoneUtcPlusOneTick;
var bumped = shipped with { UpdatedUtc = bumpedUpdatedUtc, Revision = local.Revision + 1 };
await _store.ApplyReplicatedAsync(bumped, ct).ConfigureAwait(false);
var reconfirmed = await _store.GetAsync(name, ct).ConfigureAwait(false);
return reconfirmed is not null && !reconfirmed.IsDeleted && MatchesShippedCiphertext(reconfirmed, shipped)
? ApplyOutcome.Applied
: ApplyOutcome.Unconfirmed;
}
// local is live but its ciphertext differs from what was shipped: an out-of-band value
// with a strictly newer (UpdatedUtc, Revision) won LWW. (local is null only in a
// pathological concurrent-delete race; Unconfirmed covers that too.)
return local is null ? ApplyOutcome.Unconfirmed : ApplyOutcome.LocalNewer;
}
private static bool MatchesShippedCiphertext(StoredSecret local, StoredSecret shipped) =>
local.Ciphertext.AsSpan().SequenceEqual(shipped.Ciphertext)
&& local.WrappedDek.AsSpan().SequenceEqual(shipped.WrappedDek);
/// <summary>
/// Trial-resolves <paramref name="row"/>'s secret through <see cref="ISecretResolver"/> to
/// verify it decrypts on this node. On failure, appends a message to
/// <paramref name="failed"/> and logs — Error the first time this exact (name, revision,
/// kekId) has failed since it last resolved successfully on this node, Debug on every
/// repeat (see the class remarks on error-log de-spam).
/// <paramref name="failed"/> and logs with the shared de-spam rule (see the class remarks).
/// </summary>
private async Task VerifyAsync(DeployedSecretRow row, List<string> failed, CancellationToken ct)
{
@@ -211,7 +396,7 @@ public sealed class DeployedSecretsMirror : BackgroundService
{
resolved = await _resolver.GetAsync(new SecretName(row.Name), ct).ConfigureAwait(false) is not null;
}
catch (OperationCanceledException)
catch (OperationCanceledException) when (ct.IsCancellationRequested)
{
throw;
}
@@ -229,16 +414,25 @@ public sealed class DeployedSecretsMirror : BackgroundService
var message =
$"Secret '{row.Name}' (kek {row.Secret.KekId}) did not decrypt on this node — " +
"the node's master key differs from central's";
ReportFailure(row.Name, $"decrypt:{row.Secret.Revision}:{row.Secret.KekId}", message, failed);
}
/// <summary>
/// Appends <paramref name="message"/> to <paramref name="failed"/> and logs it — Error the
/// first time <paramref name="fingerprint"/> is seen for <paramref name="name"/> since it
/// last succeeded, Debug on every identical repeat.
/// </summary>
private void ReportFailure(string name, string fingerprint, string message, List<string> failed)
{
failed.Add(message);
var current = (row.Secret.Revision, row.Secret.KekId);
if (_lastLoggedFailure.TryGetValue(row.Name, out var previous) && previous == current)
if (_lastLoggedFailure.TryGetValue(name, out var previous) && previous == fingerprint)
{
_logger.LogDebug("{Failure}", message);
}
else
{
_lastLoggedFailure[row.Name] = current;
_lastLoggedFailure[name] = fingerprint;
_logger.LogError("{Failure}", message);
}
}
@@ -253,12 +447,15 @@ public sealed class DeployedSecretsMirror : BackgroundService
{
await ReconcileAsync(stoppingToken).ConfigureAwait(false);
}
catch (OperationCanceledException)
catch (OperationCanceledException) when (stoppingToken.IsCancellationRequested)
{
break;
}
catch (Exception ex)
{
// Covers every other failure, including an OperationCanceledException NOT
// attributable to this service's own shutdown token (e.g. a store/command
// timeout) — that must not silently end the loop (review Minor #4).
_logger.LogWarning(ex, "Deployed secrets mirror pass failed; retrying next interval.");
}
} while (await timer.WaitForNextTickAsync(stoppingToken).ConfigureAwait(false));
@@ -281,4 +478,16 @@ public sealed class DeployedSecretsMirror : BackgroundService
CreatedUtc = secret.CreatedUtc,
UpdatedUtc = secret.UpdatedUtc,
};
/// <inheritdoc />
public override void Dispose()
{
if (!_disposed)
{
_disposed = true;
_gate.Dispose();
}
base.Dispose();
}
}