microsoft · TedHartMS · May 19, 2026 · May 14, 2026 · May 14, 2026 · May 14, 2026
@@ -36,6 +36,9 @@ jobs:
     name: BDNBenchmark
     needs: [changes]
     runs-on: ${{ matrix.os }}
+    concurrency:
+      group: bdn-benchmark-${{ matrix.gh-pages-branch }}-${{ matrix.os }}-${{ matrix.framework }}-${{ matrix.configuration }}-${{ matrix.test }}
+      cancel-in-progress: false
     strategy:
       fail-fast: false
       matrix:
@@ -51,37 +54,30 @@ jobs:
     steps:
       - name: Check out code
         uses: actions/checkout@v4
+
       - name: Setup .NET
         uses: actions/setup-dotnet@v4
+
       - name: Install dependencies
         run: dotnet restore
 
       - name: Run BDN.benchmark Perf Test
         run: ./test/BDNPerfTests/run_bdnperftest.ps1 ${{ matrix.test }} ${{ matrix.framework }}
         shell: pwsh
         continue-on-error: false
-
-      - name: Random pause between tasks so multiple inserts don't run at the same time
-        run: |
-          $delay = Get-Random -Minimum 1 -Maximum 600
-          Write-Host "Sleeping for $delay seconds..."
-          Start-Sleep -Seconds $delay
-        shell: pwsh
 
       - name: Upload test results to artifacts
         uses: actions/upload-artifact@v4
         with:
           name: Results-${{ matrix.os }}-${{ matrix.framework }}-${{ matrix.configuration }}-${{ matrix.test }}
-          path: |
-             ./test/BDNPerfTests/results
+          path: ./test/BDNPerfTests/results
         if: ${{ always() }}
 
       - name: Upload Error Log to artifacts
         uses: actions/upload-artifact@v4
         with:
           name: ErrorLog-${{ matrix.os }}-${{ matrix.framework }}-${{ matrix.configuration }}-${{ matrix.test }}
-          path: |
-            ./test/BDNPerfTests/errorlog
+          path: ./test/BDNPerfTests/errorlog
         if: ${{ always() }}
 
       # Run `github-action-benchmark` action for the Continuous Benchmarking Charts (https://microsoft.github.io/garnet/charts/)

@@ -144,7 +144,7 @@ jobs:
       - name: Upload test results
         uses: actions/upload-artifact@v4
         with:
-          name: dotnet-garnet-results-${{ matrix.os }}-${{ matrix.framework }}-${{ matrix.configuration }}-${{ matrix.test }}
+          name: dotnet-standalone-results-${{ matrix.os }}-${{ matrix.framework }}-${{ matrix.configuration }}-${{ matrix.test }}
           path: GarnetTestResults-${{ matrix.os }}-${{ matrix.framework }}-${{ matrix.configuration }}-${{ matrix.test }}
         if: ${{ always() }}
 
@@ -159,7 +159,7 @@ jobs:
         os: [ ubuntu-latest, windows-latest ]
         framework: [ 'net8.0' , 'net10.0']
         configuration: [ 'Debug', 'Release' ]
-        test: [ 'Garnet.test.cluster', 'Garnet.test.cluster.migrate', 'Garnet.test.cluster.replication', 'Garnet.test.cluster.replication.tls', 'Garnet.test.cluster.replication.asyncreplay', 'Garnet.test.cluster.replication.disklesssync', 'Garnet.test.cluster.vectorsets', 'Garnet.test.cluster.multilog' ]
+        test: [ 'Garnet.test.cluster', 'Garnet.test.cluster.migrate', 'Garnet.test.cluster.replication', 'Garnet.test.cluster.replication.tls', 'Garnet.test.cluster.replication.disklesssync', 'Garnet.test.cluster.vectorsets', 'Garnet.test.cluster.multilog' ]
     if: needs.changes.outputs.garnet == 'true'
     steps:
       - name: Check out code
@@ -180,7 +180,7 @@ jobs:
       - name: Upload test results
         uses: actions/upload-artifact@v4
         with:
-          name: dotnet-garnet-results-${{ matrix.os }}-${{ matrix.framework }}-${{ matrix.configuration }}-${{ matrix.test }}
+          name: dotnet-cluster-results-${{ matrix.os }}-${{ matrix.framework }}-${{ matrix.configuration }}-${{ matrix.test }}
           path: GarnetTestResults-${{ matrix.os }}-${{ matrix.framework }}-${{ matrix.configuration }}-${{ matrix.test }}
         if: ${{ always() }}
 
@@ -276,7 +276,7 @@ jobs:
           echo "TSAVORITE_TEST_DIR=$dir" >> $env:GITHUB_ENV
 
       - name: Run tests ${{ matrix.test }}
-        run: dotnet test ${{ env.TSAVORITE_TEST_DIR }} -f ${{ matrix.framework }} --configuration ${{ matrix.configuration }} --logger "console;verbosity=detailed" --logger trx --results-directory "TsavoriteTestResults-${{ matrix.os }}-${{ matrix.framework }}-${{ matrix.configuration }}-${{ matrix.test }}"
+        run: dotnet test ${{ env.TSAVORITE_TEST_DIR }} -f ${{ matrix.framework }} --configuration ${{ matrix.configuration }} --logger "console;verbosity=detailed" --logger trx --results-directory "TsavoriteTestResults-${{ matrix.os }}-${{ matrix.framework }}-${{ matrix.configuration }}-${{ matrix.test }}" -- NUnit.DisplayName=FullName
         timeout-minutes: 45
       - name: Upload test results
         uses: actions/upload-artifact@v4
@@ -305,6 +305,146 @@ jobs:
       - name: Build website
         run: yarn build
 
+  # Job to generate combined test summaries per category
+  test-summary:
+    name: Test Summary
+    runs-on: ubuntu-latest
+    needs: [changes, test-garnet-standalone, test-garnet-cluster, test-tsavorite]
+    if: always() && (needs.test-garnet-standalone.result != 'skipped' || needs.test-garnet-cluster.result != 'skipped' || needs.test-tsavorite.result != 'skipped')
+    steps:
+      - name: Download standalone test results
+        if: needs.test-garnet-standalone.result != 'skipped'
+        uses: actions/download-artifact@v4
+        with:
+          pattern: dotnet-standalone-results-*
+          path: results/standalone
+          merge-multiple: true
+      - name: Download cluster test results
+        if: needs.test-garnet-cluster.result != 'skipped'
+        uses: actions/download-artifact@v4
+        with:
+          pattern: dotnet-cluster-results-*
+          path: results/cluster
+          merge-multiple: true
+      - name: Download Tsavorite test results
+        if: needs.test-tsavorite.result != 'skipped'
+        uses: actions/download-artifact@v4
+        with:
+          pattern: dotnet-tsavorite-results-*
+          path: results/tsavorite
+          merge-multiple: true
+      - name: Generate summaries
+        shell: pwsh
+        run: |
+          function Write-CategorySummary($category, $dir) {
+            $trxFiles = Get-ChildItem -Path $dir -Filter "*.trx" -Recurse -ErrorAction SilentlyContinue
+            if (-not $trxFiles -or $trxFiles.Count -eq 0) { return }
+            $totalTests = 0; $passed = 0; $failed = 0; $skipped = 0
+            $durations = @()
+            $failedTests = @()
+            foreach ($trx in $trxFiles) {
+              $xml = [System.Xml.XmlDocument]::new()
+              try { $xml.Load($trx.FullName) }
+              catch {
+                Write-Warning "Skipping malformed TRX file: $($trx.FullName) - $_"
+                continue
+              }
+              $ns = @{ t = "http://microsoft.com/schemas/VisualStudio/TeamTest/2010" }
+              $results = Select-Xml -Xml $xml -XPath "//t:UnitTestResult" -Namespace $ns
+              foreach ($r in $results) {
+                $totalTests++
+                $outcome = $r.Node.outcome
+                if ($outcome -eq "Passed") { $passed++ }
+                elseif ($outcome -eq "Failed") {
+                  $failed++
+                  $dur = $r.Node.duration
+                  $durSec = 0
+                  if ($dur) { try { $durSec = [TimeSpan]::Parse($dur).TotalSeconds } catch { } }
+                  $failedTests += @{ Name = $r.Node.testName; Seconds = $durSec }
+                }
+                else { $skipped++ }
+                $dur = $r.Node.duration
+                if ($dur) {
+                  try { $ts = [TimeSpan]::Parse($dur); $durations += @{ Name = $r.Node.testName; Seconds = $ts.TotalSeconds } }
+                  catch { }
+                }
+              }
+            }
+            # Compute total duration as sum of all individual test durations
+            $totalSec = ($durations | ForEach-Object { $_.Seconds } | Measure-Object -Sum).Sum
+            if ($totalSec -gt 0) {
+              $totalDurStr = "{0}m {1:D2}s" -f [int][math]::Floor($totalSec / 60), [int]($totalSec % 60)
+            } else { $totalDurStr = "N/A" }
+            $buckets = @(
+              @{ Label = "< 1s";      Min = 0;   Max = 1 },
+              @{ Label = "1s - 5s";   Min = 1;   Max = 5 },
+              @{ Label = "5s - 15s";  Min = 5;   Max = 15 },
+              @{ Label = "15s - 30s"; Min = 15;  Max = 30 },
+              @{ Label = "30s - 60s"; Min = 30;  Max = 60 },
+              @{ Label = "> 60s";     Min = 60;  Max = [double]::MaxValue }
+            )
+            $counts = @(0,0,0,0,0,0)
+            foreach ($d in $durations) {
+              for ($i = 0; $i -lt $buckets.Count; $i++) {
+                if ($d.Seconds -ge $buckets[$i].Min -and $d.Seconds -lt $buckets[$i].Max) { $counts[$i]++; break }
+              }
+            }
+            $maxCount = ($counts | Measure-Object -Maximum).Maximum
+            if ($maxCount -eq 0) { $maxCount = 1 }
+            $sb = [System.Text.StringBuilder]::new()
+            [void]$sb.AppendLine("## 🧪 $category")
+            [void]$sb.AppendLine("")
+            [void]$sb.AppendLine("⏱️ **Total: $totalDurStr** · $totalTests tests (✅ $passed · ❌ $failed · ⏭️ $skipped)")
+            [void]$sb.AppendLine("")
+            # Failed tests section (only if there are failures)
+            if ($failedTests.Count -gt 0) {
+              [void]$sb.AppendLine("### ❌ Failed Tests")
+              [void]$sb.AppendLine("")
+              [void]$sb.AppendLine("| Test | Duration |")
+              [void]$sb.AppendLine("|------|----------|")
+              foreach ($t in $failedTests) {
+                $durStr = "{0:F1}s" -f $t.Seconds
+                [void]$sb.AppendLine("| $($t.Name) | $durStr |")
+              }
+              [void]$sb.AppendLine("")
+            }
+            [void]$sb.AppendLine("### Distribution")
+            [void]$sb.AppendLine("")
+            [void]$sb.AppendLine("| Duration | Count | |")
+            [void]$sb.AppendLine("|----------|------:|--|")
+            for ($i = 0; $i -lt $buckets.Count; $i++) {
+              $barLen = [int][math]::Round(($counts[$i] / $maxCount) * 20)
+              if ($barLen -eq 0 -and $counts[$i] -gt 0) { $barLen = 1 }
+              $bar = "█" * $barLen
+              [void]$sb.AppendLine("| $($buckets[$i].Label) | $($counts[$i]) | $bar |")
+            }
+            $top10 = $durations | Sort-Object { $_.Seconds } -Descending | Select-Object -First 10
+            if ($top10.Count -gt 0) {
+              [void]$sb.AppendLine("")
+              [void]$sb.AppendLine("### 🐢 Top 10 Slowest")
+              [void]$sb.AppendLine("")
+              [void]$sb.AppendLine("| Test | Duration |")
+              [void]$sb.AppendLine("|------|----------|")
+              foreach ($t in $top10) {
+                $durStr = "{0:F1}s" -f $t.Seconds
+                [void]$sb.AppendLine("| $($t.Name) | $durStr |")
+              }
+            }
+            [void]$sb.AppendLine("")
+            $sb.ToString() >> $env:GITHUB_STEP_SUMMARY
+          }
+
+          # Generate per-category summaries
+          if (Test-Path "results/standalone") {
+            Write-CategorySummary "Garnet Standalone" "results/standalone"
+          }
+          if (Test-Path "results/cluster") {
+            Write-CategorySummary "Garnet Cluster" "results/cluster"
+          }
+          if (Test-Path "results/tsavorite") {
+            Write-CategorySummary "Tsavorite" "results/tsavorite"
+          }
+
   pipeline-success:
     name: Garnet CI (Complete)
     runs-on: ubuntu-latest

@@ -81,7 +81,7 @@ jobs:
     strategy:
       fail-fast: false
       matrix:
-        test: [ 'Tsavorite.test', 'Tsavorite.test.recordops', 'Tsavorite.test.session', 'Tsavorite.test.session.context', 'Tsavorite.test.hlog', 'Tsavorite.test.recovery' ]
+        test: [ 'Tsavorite.test', 'Tsavorite.test.recordops', 'Tsavorite.test.session', 'Tsavorite.test.session.context', 'Tsavorite.test.hlog', 'Tsavorite.test.recovery', 'Tsavorite.test.stress' ]
         os: [ ubuntu-latest, windows-latest ]
         framework: [ 'net8.0', 'net10.0' ]
         configuration: [ 'Debug', 'Release' ]
@@ -124,6 +124,7 @@ jobs:
             'Tsavorite.test.session.context' = 'libs/storage/Tsavorite/cs/test/test.session.context'
             'Tsavorite.test.hlog' = 'libs/storage/Tsavorite/cs/test/test.hlog'
             'Tsavorite.test.recovery' = 'libs/storage/Tsavorite/cs/test/test.recovery'
+            'Tsavorite.test.stress' = 'libs/storage/Tsavorite/cs/test/test.stress'
           }
           $dir = $tsavoriteDirMap['${{ matrix.test }}']
           echo "TSAVORITE_TEST_DIR=$dir" >> $env:GITHUB_ENV

diff --git a/benchmark/BDN.benchmark/Cluster/ClusterContext.cs b/benchmark/BDN.benchmark/Cluster/ClusterContext.cs
@@ -16,7 +16,7 @@ unsafe class ClusterContext
         EmbeddedRespServer server;
         RespServerSession session;
         readonly BenchUtils benchUtils = new();
-        readonly int port = 7000;
+        readonly int port = 7000;   // This is not in the shared ClusterTestContext so does not conflict
 
         public static ReadOnlySpan<byte> keyTag => "{0}"u8;
         public Request[] singleGetSet;

diff --git a/libs/cluster/Server/Replication/CheckpointStore.cs b/libs/cluster/Server/Replication/CheckpointStore.cs
@@ -197,8 +197,16 @@ private void DeleteOutdatedCheckpoints()
             // This is safe because curr is the oldest entry still referenced by active readers.
             if (curr != null)
             {
-                var hlogSize = storeWrapper.store.GetLogFileSize(curr.metadata.storeHlogToken);
-                storeWrapper.store.Log.ShiftBeginAddress(hlogSize.hybridLogFileStartAddress, truncateLog: true);
+                try
+                {
+                    var hlogSize = storeWrapper.store.GetLogFileSize(curr.metadata.storeHlogToken);
+                    storeWrapper.store.Log.ShiftBeginAddress(hlogSize.hybridLogFileStartAddress, truncateLog: true);
+                }
+                catch (Exception ex)
+                {
+                    logger?.LogWarning(ex, "Skipping hlog truncation for checkpoint {token} because checkpoint metadata could not be read",
+                                           curr.metadata.storeHlogToken);
+                }
             }
 
             // Update head after delete

diff --git a/libs/cluster/Server/Replication/GarnetClusterCheckpointManager.cs b/libs/cluster/Server/Replication/GarnetClusterCheckpointManager.cs
@@ -76,15 +76,13 @@ private HybridLogRecoveryInfo ConvertMetadata(byte[] checkpointMetadata)
             // Try to parse new format where cookie is embedded inside the HybridLogRecoveryInfo
             try
             {
-                using (StreamReader s = new(new MemoryStream(checkpointMetadata)))
-                {
-                    recoveryInfo.Initialize(s);
-                }
+                using var s = new StreamReader(new MemoryStream(checkpointMetadata));
+                recoveryInfo.Initialize(s);
             }
             catch (Exception ex)
             {
                 logger?.LogError(ex, "Best effort read of checkpoint metadata failed");
-                throw ex.InnerException;
+                throw;
             }
 
             return recoveryInfo;

diff --git a/libs/host/Configuration/Options.cs b/libs/host/Configuration/Options.cs
@@ -564,7 +564,9 @@ internal sealed class Options : ICloneable
         public int IndexResizeThreshold { get; set; }
 
         // ValueOverflowThreshold must be at least 64 bytes and strictly less than PageSize (both after rounding down to the previous power of 2).
-        // Validated at server-options consumption time; see GarnetServerOptions.ValueOverflowThresholdBytes.
+        // Validated at server-options consumption time; see GarnetServerOptions.ValueOverflowThresholdBytes. Note that we do not have a KeyOverflowThreshold
+        // because it would complicate the minimum pagesize check that uses ValueOverflowThreshold check at startup; keys are usually small so calculating a
+        // minimum page size using a large(ish) Key threshold would have spurious errors. We'll defer that rare case to runtime checks.
         [MemorySizeValidation(isRequired: false)]
         [Option("value-overflow-threshold", Required = false, HelpText = "Max size of a value stored inline in the main-log page (larger values overflow to the heap). Accepts a memory size (e.g. 4k, 1m). Minimum 64 bytes; must be less than PageSize.")]
         public string ValueOverflowThreshold { get; set; }