Files
scadalink-design/tests/ScadaLink.PerformanceTests/HealthAggregationTests.cs
Joseph Doherty f165ca2774 feat: wire all health metrics and add instance counts to dashboard
Wired ISiteHealthCollector calls for script errors (ScriptExecutionActor),
alarm eval errors (AlarmActor), dead letters (DeadLetterMonitorActor), and
S&F buffer depth placeholder. Added instance count tracking (deployed/
enabled/disabled) to SiteHealthReport via DeploymentManagerActor. Updated
Health Dashboard UI to show instance counts per site. All metrics flow
through the existing health report pipeline via ClusterClient.
2026-03-18 00:57:49 -04:00

167 lines
6.0 KiB
C#

using Microsoft.Extensions.Logging.Abstractions;
using Microsoft.Extensions.Options;
using ScadaLink.Commons.Messages.Health;
using ScadaLink.Commons.Types.Enums;
using ScadaLink.HealthMonitoring;
namespace ScadaLink.PerformanceTests;
/// <summary>
/// WP-4 (Phase 8): Performance test framework for health reporting aggregation.
/// Verifies health reporting from 10 sites can be aggregated correctly.
/// </summary>
public class HealthAggregationTests
{
private readonly CentralHealthAggregator _aggregator;
public HealthAggregationTests()
{
var options = Options.Create(new HealthMonitoringOptions
{
ReportInterval = TimeSpan.FromSeconds(30),
OfflineTimeout = TimeSpan.FromSeconds(60)
});
_aggregator = new CentralHealthAggregator(
options,
NullLogger<CentralHealthAggregator>.Instance);
}
[Trait("Category", "Performance")]
[Fact]
public void AggregateHealthReports_10Sites_AllTracked()
{
const int siteCount = 10;
for (var i = 0; i < siteCount; i++)
{
var siteId = $"site-{i + 1:D2}";
var report = new SiteHealthReport(
SiteId: siteId,
SequenceNumber: 1,
ReportTimestamp: DateTimeOffset.UtcNow,
DataConnectionStatuses: new Dictionary<string, ConnectionHealth>
{
[$"opc-{siteId}"] = ConnectionHealth.Connected
},
TagResolutionCounts: new Dictionary<string, TagResolutionStatus>
{
[$"opc-{siteId}"] = new(75, 72)
},
ScriptErrorCount: 0,
AlarmEvaluationErrorCount: 0,
StoreAndForwardBufferDepths: new Dictionary<string, int>
{
["ext-system"] = i * 2
},
DeadLetterCount: 0,
DeployedInstanceCount: 0,
EnabledInstanceCount: 0,
DisabledInstanceCount: 0);
_aggregator.ProcessReport(report);
}
var states = _aggregator.GetAllSiteStates();
Assert.Equal(siteCount, states.Count);
Assert.All(states.Values, s => Assert.True(s.IsOnline));
}
[Trait("Category", "Performance")]
[Fact]
public void AggregateHealthReports_RapidUpdates_HandlesVolume()
{
const int siteCount = 10;
const int updatesPerSite = 100;
for (var seq = 1; seq <= updatesPerSite; seq++)
{
for (var s = 0; s < siteCount; s++)
{
var report = new SiteHealthReport(
SiteId: $"site-{s + 1:D2}",
SequenceNumber: seq,
ReportTimestamp: DateTimeOffset.UtcNow,
DataConnectionStatuses: new Dictionary<string, ConnectionHealth>(),
TagResolutionCounts: new Dictionary<string, TagResolutionStatus>(),
ScriptErrorCount: seq % 5 == 0 ? 1 : 0,
AlarmEvaluationErrorCount: 0,
StoreAndForwardBufferDepths: new Dictionary<string, int>(),
DeadLetterCount: 0,
DeployedInstanceCount: 0,
EnabledInstanceCount: 0,
DisabledInstanceCount: 0);
_aggregator.ProcessReport(report);
}
}
var states = _aggregator.GetAllSiteStates();
Assert.Equal(siteCount, states.Count);
// Verify all sites have the latest sequence number
Assert.All(states.Values, s =>
{
Assert.Equal(updatesPerSite, s.LastSequenceNumber);
Assert.True(s.IsOnline);
});
}
[Trait("Category", "Performance")]
[Fact]
public void AggregateHealthReports_StaleReportsRejected()
{
var siteId = "site-01";
// Send report with seq 10
_aggregator.ProcessReport(new SiteHealthReport(
siteId, 10, DateTimeOffset.UtcNow,
new Dictionary<string, ConnectionHealth>(),
new Dictionary<string, TagResolutionStatus>(),
5, 0, new Dictionary<string, int>(), 0, 0, 0, 0));
// Send stale report with seq 5 — should be rejected
_aggregator.ProcessReport(new SiteHealthReport(
siteId, 5, DateTimeOffset.UtcNow,
new Dictionary<string, ConnectionHealth>(),
new Dictionary<string, TagResolutionStatus>(),
99, 0, new Dictionary<string, int>(), 0, 0, 0, 0));
var state = _aggregator.GetSiteState(siteId);
Assert.NotNull(state);
Assert.Equal(10, state!.LastSequenceNumber);
// The script error count from report 10 (5) should be kept, not replaced by 99
Assert.Equal(5, state.LatestReport!.ScriptErrorCount);
}
[Trait("Category", "Performance")]
[Fact]
public void HealthCollector_CollectReport_ResetsIntervalCounters()
{
var collector = new SiteHealthCollector();
// Simulate errors during an interval
for (var i = 0; i < 10; i++) collector.IncrementScriptError();
for (var i = 0; i < 3; i++) collector.IncrementAlarmError();
for (var i = 0; i < 7; i++) collector.IncrementDeadLetter();
collector.UpdateConnectionHealth("opc-1", ConnectionHealth.Connected);
collector.UpdateTagResolution("opc-1", 75, 72);
var report = collector.CollectReport("site-01");
Assert.Equal("site-01", report.SiteId);
Assert.Equal(10, report.ScriptErrorCount);
Assert.Equal(3, report.AlarmEvaluationErrorCount);
Assert.Equal(7, report.DeadLetterCount);
Assert.Single(report.DataConnectionStatuses);
// Second collect should have reset interval counters
var report2 = collector.CollectReport("site-01");
Assert.Equal(0, report2.ScriptErrorCount);
Assert.Equal(0, report2.AlarmEvaluationErrorCount);
Assert.Equal(0, report2.DeadLetterCount);
// Connection status persists (not interval-based)
Assert.Single(report2.DataConnectionStatuses);
}
}