feat(server): add observability and operator controls (#16)
quality-gate / quality (push) Failing after 1m1s

This commit is contained in:
KyuubiYoru
2026-07-16 13:22:16 +02:00
parent 88ef946af5
commit be732de7c9
31 changed files with 4149 additions and 140 deletions
@@ -0,0 +1,14 @@
using System.ComponentModel.DataAnnotations;
namespace FinalFactory.Rendezvous.Server.Observability;
internal sealed class AuditOptions
{
public const string SectionName = "Rendezvous:Audit";
[Range(100, 100_000)]
public int MaxEntries { get; set; } = 10_000;
[Range(1, 30)]
public int RetentionDays { get; set; } = 30;
}
@@ -0,0 +1,134 @@
using System.Security.Cryptography;
using System.Text;
using Microsoft.Extensions.Options;
namespace FinalFactory.Rendezvous.Server.Observability;
internal sealed partial class AuditTrail
{
private readonly object _gate = new();
private readonly LinkedList<AuditEntry> _entries = [];
private readonly AuditOptions _options;
private readonly TimeProvider _timeProvider;
private readonly ILogger<AuditTrail> _logger;
private readonly RendezvousTelemetry _telemetry;
public AuditTrail(
IOptions<AuditOptions> options,
ILogger<AuditTrail> logger,
RendezvousTelemetry telemetry,
TimeProvider? timeProvider = null)
{
_options = options.Value;
_logger = logger;
_telemetry = telemetry;
_timeProvider = timeProvider ?? TimeProvider.System;
}
public void Record(
string actorSubject,
string action,
string result,
string targetKind,
string targetIdentifier,
string correlationId)
{
DateTimeOffset now = _timeProvider.GetUtcNow();
AuditEntry entry = new(
now,
Fingerprint(actorSubject),
action,
result,
targetKind,
Fingerprint(targetIdentifier),
correlationId);
lock (_gate)
{
PurgeExpired(now);
while (_entries.Count >= _options.MaxEntries)
{
_entries.RemoveFirst();
}
_entries.AddLast(entry);
}
_telemetry.RecordAudit(action, result);
LogOperatorAction(
_logger,
entry.Timestamp,
entry.ActorFingerprint,
action,
result,
targetKind,
entry.TargetFingerprint,
correlationId);
}
public IReadOnlyDictionary<string, long> GetAggregateCounts()
{
lock (_gate)
{
PurgeExpired(_timeProvider.GetUtcNow());
return _entries
.GroupBy(static entry => $"{entry.Action}:{entry.Result}", StringComparer.Ordinal)
.ToDictionary(
static group => group.Key,
static group => (long)group.Count(),
StringComparer.Ordinal);
}
}
internal IReadOnlyList<AuditEntry> GetEntriesForTests()
{
lock (_gate)
{
PurgeExpired(_timeProvider.GetUtcNow());
return _entries.ToArray();
}
}
private void PurgeExpired(DateTimeOffset now)
{
DateTimeOffset oldest = now.AddDays(-_options.RetentionDays);
while (_entries.First is { Value.Timestamp: var timestamp }
&& timestamp < oldest)
{
_entries.RemoveFirst();
}
}
private static string Fingerprint(string value)
{
byte[] digest = SHA256.HashData(Encoding.UTF8.GetBytes(value));
return Convert.ToHexString(digest.AsSpan(0, 12));
}
[LoggerMessage(
EventId = 100,
Level = LogLevel.Information,
Message = "Operator audit at {Timestamp}: actor {ActorFingerprint} action {Action} completed with {Result} for {TargetKind} target {TargetFingerprint}; correlation {CorrelationId}")]
private static partial void LogOperatorAction(
ILogger logger,
DateTimeOffset timestamp,
string actorFingerprint,
string action,
string result,
string targetKind,
string targetFingerprint,
string correlationId);
}
internal sealed record AuditEntry(
DateTimeOffset Timestamp,
string ActorFingerprint,
string Action,
string Result,
string TargetKind,
string TargetFingerprint,
string CorrelationId)
{
public override string ToString() =>
$"[AuditEntry {Action}/{Result}; actor and target fingerprinted]";
}
@@ -0,0 +1,30 @@
using FinalFactory.Rendezvous.Contracts;
namespace FinalFactory.Rendezvous.Server.Observability;
internal static class HealthEndpoints
{
public static IEndpointRouteBuilder MapRendezvousHealthEndpoints(
this IEndpointRouteBuilder endpoints)
{
endpoints.MapGet(
"/health/live",
static () => Results.Ok(new HealthResponse { Status = "live" }))
.Produces<HealthResponse>()
.Produces<ApiError>(StatusCodes.Status429TooManyRequests)
.WithName("GetLiveness")
.WithTags("Health");
endpoints.MapGet(
"/health/ready",
static (RendezvousReadiness readiness) =>
!readiness.GetSnapshot().IsReady
? Results.StatusCode(StatusCodes.Status503ServiceUnavailable)
: Results.Ok(new HealthResponse { Status = "ready" }))
.Produces<HealthResponse>()
.Produces<ApiError>(StatusCodes.Status429TooManyRequests)
.Produces(StatusCodes.Status503ServiceUnavailable)
.WithName("GetReadiness")
.WithTags("Health");
return endpoints;
}
}
@@ -0,0 +1,41 @@
using FinalFactory.Rendezvous.Server.Provisioning;
using FinalFactory.Rendezvous.Server.State;
using FinalFactory.Rendezvous.Server.Transport;
using Microsoft.Extensions.Options;
namespace FinalFactory.Rendezvous.Server.Observability;
internal sealed class RendezvousReadiness(
UdpMediatorService mediator,
ProvisioningReadiness provisioning,
IEphemeralRendezvousStore state,
IOptions<UdpMediatorOptions> udpOptions)
{
public ReadinessSnapshot GetSnapshot()
{
bool ipv6Required = !string.IsNullOrWhiteSpace(udpOptions.Value.Ipv6ListenAddress);
return new ReadinessSnapshot(
HttpListenerReady: true,
UdpIpv4ListenerReady: mediator.LocalEndpoint is not null,
UdpIpv6ListenerReady: !ipv6Required || mediator.LocalIpv6Endpoint is not null,
ProvisioningReady: provisioning.IsReady,
StoreAvailable: state.IsAvailable,
Draining: state.IsDraining);
}
}
internal sealed record ReadinessSnapshot(
bool HttpListenerReady,
bool UdpIpv4ListenerReady,
bool UdpIpv6ListenerReady,
bool ProvisioningReady,
bool StoreAvailable,
bool Draining)
{
public bool IsReady => HttpListenerReady
&& UdpIpv4ListenerReady
&& UdpIpv6ListenerReady
&& ProvisioningReady
&& StoreAvailable
&& !Draining;
}
@@ -0,0 +1,126 @@
using System.Diagnostics;
using System.Diagnostics.Metrics;
using FinalFactory.Rendezvous.Server.State;
namespace FinalFactory.Rendezvous.Server.Observability;
internal sealed class RendezvousTelemetry : IDisposable
{
public const string MeterName = "FinalFactory.Rendezvous";
public const string ActivitySourceName = "FinalFactory.Rendezvous.Server";
private readonly InMemoryEphemeralRendezvousStore _store;
private readonly Meter _meter = new(MeterName, "1.0.0");
private readonly ActivitySource _activities = new(ActivitySourceName, "1.0.0");
private readonly Counter<long> _httpRequests;
private readonly Histogram<double> _httpDuration;
private readonly Counter<long> _udpResults;
private readonly Histogram<double> _udpDuration;
private readonly Counter<long> _limiterDrops;
private readonly Counter<long> _auditEvents;
private readonly Counter<long> _connectionOutcomes;
private readonly Counter<long> _operatorAuthentication;
private readonly Histogram<double> _pairingLatency;
public RendezvousTelemetry(InMemoryEphemeralRendezvousStore store)
{
_store = store;
_httpRequests = _meter.CreateCounter<long>("rendezvous.http.requests");
_httpDuration = _meter.CreateHistogram<double>(
"rendezvous.http.duration",
"ms");
_udpResults = _meter.CreateCounter<long>("rendezvous.udp.results");
_udpDuration = _meter.CreateHistogram<double>(
"rendezvous.udp.duration",
"ms");
_limiterDrops = _meter.CreateCounter<long>("rendezvous.limiter.drops");
_auditEvents = _meter.CreateCounter<long>("rendezvous.audit.events");
_connectionOutcomes = _meter.CreateCounter<long>("rendezvous.connection.outcomes");
_operatorAuthentication = _meter.CreateCounter<long>("rendezvous.operator.authentication");
_pairingLatency = _meter.CreateHistogram<double>(
"rendezvous.pairing.latency",
"ms");
_meter.CreateObservableGauge(
"rendezvous.store.active_listings",
() => _store.GetMetricsSnapshot().ActiveListings);
_meter.CreateObservableGauge(
"rendezvous.store.active_leases",
() => _store.GetMetricsSnapshot().ActiveListings);
_meter.CreateObservableGauge(
"rendezvous.store.active_attempts",
() => _store.GetMetricsSnapshot().ActiveJoinAttempts);
_meter.CreateObservableGauge(
"rendezvous.queue.depth",
() => _store.GetMetricsSnapshot().ActiveJoinAttempts);
_meter.CreateObservableGauge(
"rendezvous.store.replay_markers",
() => _store.GetMetricsSnapshot().ReplayMarkers);
_meter.CreateObservableGauge(
"rendezvous.store.available",
() => _store.GetMetricsSnapshot().IsAvailable ? 1 : 0);
_meter.CreateObservableCounter(
"rendezvous.store.expiry_churn",
() => _store.GetMetricsSnapshot().ExpiryChurn);
}
public Activity? StartActivity(string name, ActivityKind kind = ActivityKind.Internal) =>
_activities.StartActivity(name, kind);
public void RecordHttp(string operation, int statusCode, double elapsedMilliseconds)
{
TagList tags = new()
{
{ "operation", operation },
{ "status_code", statusCode },
};
_httpRequests.Add(1, tags);
_httpDuration.Record(elapsedMilliseconds, tags);
}
public void RecordUdp(string operation, string result, double elapsedMilliseconds)
{
TagList tags = new()
{
{ "operation", operation },
{ "result", result },
};
_udpResults.Add(1, tags);
_udpDuration.Record(elapsedMilliseconds, tags);
}
public void RecordLimiterDrop(string transport, string partition) =>
_limiterDrops.Add(1, new TagList
{
{ "transport", transport },
{ "partition", partition },
});
public void RecordAudit(string action, string result) =>
_auditEvents.Add(1, new TagList
{
{ "action", action },
{ "result", result },
});
public void RecordConnectionOutcome(string outcome, string elapsedBucket) =>
_connectionOutcomes.Add(1, new TagList
{
{ "outcome", outcome },
{ "elapsed_bucket", elapsedBucket },
});
public void RecordOperatorAuthentication(string result) =>
_operatorAuthentication.Add(1, new TagList
{
{ "result", result },
});
public void RecordPairingLatency(double elapsedMilliseconds) =>
_pairingLatency.Record(elapsedMilliseconds);
public void Dispose()
{
_activities.Dispose();
_meter.Dispose();
}
}
@@ -0,0 +1,32 @@
using System.Diagnostics;
namespace FinalFactory.Rendezvous.Server.Observability;
internal sealed class TelemetryMiddleware(
RequestDelegate next,
RendezvousTelemetry telemetry)
{
public async Task InvokeAsync(HttpContext context)
{
string operation = context.GetEndpoint()?.Metadata.GetMetadata<IEndpointNameMetadata>()
?.EndpointName ?? "Unmatched";
long started = Stopwatch.GetTimestamp();
using Activity? activity = telemetry.StartActivity(
$"HTTP {operation}",
ActivityKind.Server);
string correlationId = activity?.TraceId.ToString() ?? Guid.NewGuid().ToString("N");
context.Response.Headers["X-Rendezvous-Correlation-ID"] = correlationId;
activity?.SetTag("rendezvous.operation", operation);
try
{
await next(context).ConfigureAwait(false);
}
finally
{
telemetry.RecordHttp(
operation,
context.Response.StatusCode,
Stopwatch.GetElapsedTime(started).TotalMilliseconds);
}
}
}