Fix InfluxDB concurrent reads and MongoDB benchmark hosts

Snapshot each completed InfluxDB WAL file through the native persistence path so concurrent point queries do not materialize the entire million-record mutable buffer. Extend the existing real million-record control with sixteen independently owned concurrent readers.

Use the Ubuntu 22.04 family consistently for benchmark preparation and workers to retain a kernel supported by the pinned MongoDB image, and record actual host/Docker kernels.

Verified: strict Debug comparison-host build, real 1M-record control failed before and passed after with all sixteen bodies checked and physical container/volume cleanup, complete-solution formatting, repository governance, actionlint and diff whitespace checks. Complete solution checks and full Linux session qualification remain pending at this checkpoint. KeyLoad runtime and adapter sources are unchanged.

Author: KSemenenko

.github/workflows/Features/BenchmarkComparisons/SessionSetup/action.yml

@@ -23,6 +23,8 @@ runs:
       shell: bash
       run: |
         docker version
+        uname -a
+        docker info --format 'Docker kernel: {{.KernelVersion}}'
         sudo sysctl -w vm.max_map_count=262144
     - name: Verify this GitHub session job
       shell: bash

.github/workflows/benchmarks.yml

@@ -19,7 +19,7 @@ jobs:
   comparison-runtime:
     name: Prepare comparison runtime
     if: github.repository == 'managedcode/KeyLoad' && github.ref == 'refs/heads/main'
-    runs-on: ubuntu-24.04
+    runs-on: ubuntu-22.04
     timeout-minutes: 150
     permissions: {contents: read, actions: read, packages: read}
     outputs:
@@ -71,7 +71,10 @@ jobs:
           if-no-files-found: error
           retention-days: 90
       - name: Check Docker
-        run: docker version
+        run: |
+          docker version
+          uname -a
+          docker info --format 'Docker kernel: {{.KernelVersion}}'
       - name: Resolve the latest released KeyLoad server image to one digest
         id: keyload-image
         if: (github.event_name == 'push' || inputs.mode == 'full')
@@ -137,7 +140,7 @@ jobs:
     name: ${{ matrix.jobName }}
     if: (github.event_name == 'push' || inputs.mode == 'full')
     needs: [comparison-runtime]
-    runs-on: ubuntu-24.04
+    runs-on: ubuntu-22.04
     timeout-minutes: 240
     permissions: {contents: read, actions: read, packages: read}
     strategy:
@@ -190,7 +193,7 @@ jobs:
     name: ${{ matrix.jobName }}
     if: (github.event_name == 'push' || inputs.mode == 'full')
     needs: [comparison-runtime]
-    runs-on: ubuntu-24.04
+    runs-on: ubuntu-22.04
     timeout-minutes: 240
     strategy:
       fail-fast: false
@@ -201,7 +204,7 @@ jobs:
     name: ${{ matrix.jobName }}
     if: (github.event_name == 'push' || inputs.mode == 'full')
     needs: [comparison-runtime]
-    runs-on: ubuntu-24.04
+    runs-on: ubuntu-22.04
     timeout-minutes: 240
     strategy:
       fail-fast: false
@@ -212,7 +215,7 @@ jobs:
     name: ${{ matrix.jobName }}
     if: (github.event_name == 'push' || inputs.mode == 'full')
     needs: [comparison-runtime]
-    runs-on: ubuntu-24.04
+    runs-on: ubuntu-22.04
     timeout-minutes: 240
     strategy:
       fail-fast: false
@@ -223,7 +226,7 @@ jobs:
     name: ${{ matrix.jobName }}
     if: (github.event_name == 'push' || inputs.mode == 'full')
     needs: [comparison-runtime]
-    runs-on: ubuntu-24.04
+    runs-on: ubuntu-22.04
     timeout-minutes: 240
     strategy:
       fail-fast: false
@@ -234,7 +237,7 @@ jobs:
     name: ${{ matrix.jobName }}
     if: (github.event_name == 'push' || inputs.mode == 'full')
     needs: [comparison-runtime]
-    runs-on: ubuntu-24.04
+    runs-on: ubuntu-22.04
     timeout-minutes: 240
     strategy:
       fail-fast: false
@@ -245,7 +248,7 @@ jobs:
     name: ${{ matrix.jobName }}
     if: (github.event_name == 'push' || inputs.mode == 'full')
     needs: [comparison-runtime]
-    runs-on: ubuntu-24.04
+    runs-on: ubuntu-22.04
     timeout-minutes: 240
     strategy:
       fail-fast: false
@@ -256,7 +259,7 @@ jobs:
     name: ${{ matrix.jobName }}
     if: (github.event_name == 'push' || inputs.mode == 'full')
     needs: [comparison-runtime]
-    runs-on: ubuntu-24.04
+    runs-on: ubuntu-22.04
     timeout-minutes: 240
     strategy:
       fail-fast: false
@@ -267,7 +270,7 @@ jobs:
     name: ${{ matrix.jobName }}
     if: (github.event_name == 'push' || inputs.mode == 'full')
     needs: [comparison-runtime]
-    runs-on: ubuntu-24.04
+    runs-on: ubuntu-22.04
     timeout-minutes: 240
     strategy:
       fail-fast: false
@@ -278,7 +281,7 @@ jobs:
     name: ${{ matrix.jobName }}
     if: (github.event_name == 'push' || inputs.mode == 'full')
     needs: [comparison-runtime]
-    runs-on: ubuntu-24.04
+    runs-on: ubuntu-22.04
     timeout-minutes: 240
     strategy:
       fail-fast: false
@@ -289,7 +292,7 @@ jobs:
     name: ${{ matrix.jobName }}
     if: (github.event_name == 'push' || inputs.mode == 'full')
     needs: [comparison-runtime]
-    runs-on: ubuntu-24.04
+    runs-on: ubuntu-22.04
     timeout-minutes: 240
     strategy:
       fail-fast: false
@@ -300,7 +303,7 @@ jobs:
     name: ${{ matrix.jobName }}
     if: (github.event_name == 'push' || inputs.mode == 'full')
     needs: [comparison-runtime]
-    runs-on: ubuntu-24.04
+    runs-on: ubuntu-22.04
     timeout-minutes: 240
     strategy:
       fail-fast: false
@@ -311,7 +314,7 @@ jobs:
     name: ${{ matrix.jobName }}
     if: (github.event_name == 'push' || inputs.mode == 'full')
     needs: [comparison-runtime]
-    runs-on: ubuntu-24.04
+    runs-on: ubuntu-22.04
     timeout-minutes: 240
     strategy:
       fail-fast: false
@@ -322,7 +325,7 @@ jobs:
     name: ${{ matrix.jobName }}
     if: (github.event_name == 'push' || inputs.mode == 'full')
     needs: [comparison-runtime]
-    runs-on: ubuntu-24.04
+    runs-on: ubuntu-22.04
     timeout-minutes: 240
     strategy:
       fail-fast: false
@@ -333,7 +336,7 @@ jobs:
     name: ${{ matrix.jobName }}
     if: (github.event_name == 'push' || inputs.mode == 'full')
     needs: [comparison-runtime]
-    runs-on: ubuntu-24.04
+    runs-on: ubuntu-22.04
     timeout-minutes: 240
     strategy:
       fail-fast: false
@@ -344,7 +347,7 @@ jobs:
     name: ${{ matrix.jobName }}
     if: (github.event_name == 'push' || inputs.mode == 'full')
     needs: [comparison-runtime]
-    runs-on: ubuntu-24.04
+    runs-on: ubuntu-22.04
     timeout-minutes: 240
     strategy:
       fail-fast: false
@@ -355,7 +358,7 @@ jobs:
     name: Combine benchmark results
     needs: [comparison-runtime, comparison-keyload, comparison-postgresql, comparison-qdrant, comparison-rabbitmq, comparison-redis, comparison-neo4j, comparison-mongodb, comparison-opensearch, comparison-kurrentdb, comparison-surrealdb, comparison-helixdb, comparison-timescaledb, comparison-sqlserver, comparison-mysql, comparison-clickhouse, comparison-influxdb]
     if: ${{ always() && !cancelled() && (github.event_name == 'push' || inputs.mode == 'full') && needs.comparison-runtime.result == 'success' }}
-    runs-on: ubuntu-24.04
+    runs-on: ubuntu-22.04
     timeout-minutes: 150
     env:
       GH_TOKEN: ${{ github.token }}
@@ -392,7 +395,7 @@ jobs:
       (github.event_name == 'push' || inputs.mode == 'full') &&
       github.repository == 'managedcode/KeyLoad' && github.event.repository.id == 477801965 &&
       github.ref == 'refs/heads/main'
-    runs-on: ubuntu-24.04
+    runs-on: ubuntu-22.04
     timeout-minutes: 5
     permissions: {actions: write}
     env:

benchmarks/KeyLoad.ComparisonHost/DatabaseAdapters/Cases/InfluxDocumentReadback.cs

@@ -19,6 +19,7 @@ internal static async Task VerifyAsync(IComparisonTarget target, CancellationTok
             .Select(index => index == Updated ? corpus.Replacement(index) : corpus.CreateDocument(index));
         var observed = await DocumentComparisonOracle.VerifyAsync(independent.ReadDocumentsAsync(token), expected, token);
         await Assert.That(observed.Records).IsEqualTo(Records);
+        await ReadConcurrentlyAsync(documents, corpus, token);
         using var cancelled = CancellationTokenSource.CreateLinkedTokenSource(token);
         await using (var cursor = independent.ReadDocumentsAsync(cancelled.Token).GetAsyncEnumerator(cancelled.Token))
         {
@@ -30,4 +31,21 @@ internal static async Task VerifyAsync(IComparisonTarget target, CancellationTok
         await Assert.That(BenchmarkDataset.SameDocument((await independent.ReadAsync(corpus.CreateDocument(Records), token)).Value,
             corpus.CreateDocument(Records))).IsTrue();
     }
+
+    private static async Task ReadConcurrentlyAsync(IDocumentComparisonTarget target, DocumentComparisonCorpus corpus,
+        CancellationToken token)
+    {
+        const int StableStart = 200_000;
+        var start = new TaskCompletionSource(TaskCreationOptions.RunContinuationsAsynchronously);
+        var clients = Enumerable.Range(Empty, Clients).Select(async client =>
+        {
+            await using var session = await target.OpenDocumentSessionAsync(token);
+            await start.Task;
+            var document = corpus.CreateDocument(StableStart + client);
+            var observed = await session.ReadAsync(document, token);
+            await Assert.That(BenchmarkDataset.SameDocument(observed.Value, document)).IsTrue();
+        }).ToArray();
+        start.SetResult();
+        await Task.WhenAll(clients);
+    }
 }

build/KeyLoad.LoadTests.AppHost/Configuration/BenchmarkDeploymentOptions.cs

@@ -10,6 +10,8 @@ internal sealed record BenchmarkDeploymentOptions
     private const int DefaultNeo4jPageCache = 256;
     private const int DefaultOpenSearchHeap = 1024;
     private const int DefaultInfluxQueryMemoryPercent = 70;
+    private const int DefaultInfluxWalFilesPerSnapshot = 1;
+    private const int MaximumInfluxWalFilesPerSnapshot = 600;
     private const int MaximumQueryMemoryPercent = 90;
     private const int DefaultPostgresSlotRetention = 512;
     private const int DefaultDocumentPostgresConnections = 512;
@@ -24,12 +26,14 @@ internal sealed record BenchmarkDeploymentOptions
     public int Neo4jPageCacheMegabytes { get; init; } = DefaultNeo4jPageCache;
     public int OpenSearchHeapMegabytes { get; init; } = DefaultOpenSearchHeap;
     public int InfluxQueryMemoryPercent { get; init; } = DefaultInfluxQueryMemoryPercent;
+    public int InfluxWalFilesPerSnapshot { get; init; } = DefaultInfluxWalFilesPerSnapshot;
     public int PostgresSlotWalRetentionMegabytes { get; init; } = DefaultPostgresSlotRetention;
     public int DocumentPostgresMaximumConnections { get; init; } = DefaultDocumentPostgresConnections;
     internal bool IsValid() => DocumentPostgresMaximumConnections is >= MinimumDocumentPostgresConnections and <= MaximumDocumentPostgresConnections
         && ValidMemory(Neo4jInitialHeapMegabytes) && ValidMemory(Neo4jMaximumHeapMegabytes)
         && Neo4jInitialHeapMegabytes <= Neo4jMaximumHeapMegabytes && ValidMemory(Neo4jPageCacheMegabytes)
         && ValidMemory(OpenSearchHeapMegabytes) && InfluxQueryMemoryPercent is >= 1 and <= MaximumQueryMemoryPercent
+        && InfluxWalFilesPerSnapshot is >= 1 and <= MaximumInfluxWalFilesPerSnapshot
         && PostgresSlotWalRetentionMegabytes is >= 1 and <= MaximumWalMegabytes;
     private static bool ValidMemory(int value) => value is >= 1 and <= MaximumMemoryMegabytes;
 }

build/KeyLoad.LoadTests.AppHost/Resources/Additional/AdditionalHttpResources.cs

@@ -53,11 +53,13 @@ private static void AddInflux(IsolatedResourceContext context)
         const string Serve = "serve", NodeId = "--node-id=benchmark", Store = "--object-store=file";
         const string Directory = "--data-dir=" + Mount, Bind = "--http-bind=0.0.0.0:8181", Auth = "--without-auth";
         const string QueryMemory = "--exec-mem-pool-size=", Percent = "%";
+        const string SnapshotFiles = "--wal-files-per-snapshot=";
         var deployment = LoadTestsOptionsRegistration.Get(context.Builder).Deployment.Value;
         var queryMemory = QueryMemory + deployment.InfluxQueryMemoryPercent.ToString(CultureInfo.InvariantCulture) + Percent;
+        var snapshotFiles = SnapshotFiles + deployment.InfluxWalFilesPerSnapshot.ToString(CultureInfo.InvariantCulture);
         var node = context.Builder.AddContainer(Influx, BenchmarkAdditionalImages.InfluxImage, BenchmarkAdditionalImages.InfluxTag)
             .WithImageSHA256(BenchmarkAdditionalImages.InfluxDigest[BenchmarkAdditionalImages.DigestPrefixLength..])
-            .WithArgs(Serve, NodeId, Store, Directory, Bind, Auth, queryMemory).WithVolume(target: Mount)
+            .WithArgs(Serve, NodeId, Store, Directory, Bind, Auth, queryMemory, snapshotFiles).WithVolume(target: Mount)
             .WithHttpEndpoint(targetPort: InfluxPort, name: Http).WithHttpHealthCheck(Ready);
         context.BindEndpoint(EndpointIndex, node, Http);
         context.BindImage(Docker + BenchmarkAdditionalImages.InfluxImage + Separator + BenchmarkAdditionalImages.InfluxTag + DigestSeparator + BenchmarkAdditionalImages.InfluxDigest);

docs/Features/BenchmarkComparisons/AdditionalDatabases.md

@@ -33,8 +33,9 @@ JSON row, and the frozen corpus identity capacity bounds the total row count.
 The existing operation deadline bounds response headers and each read's idle wait;
 caller cancellation also stops buffered delivery. The independent ordered oracle
 still checks every identity, full body and digest. The native million-record
-control also exercises an update, cancellation during readback and a healthy write
-and read afterward. The vendor documents
+control also exercises an update, sixteen independent concurrent point reads after
+the complete readback, cancellation during readback and a healthy write and read
+afterward. The vendor documents
 [JSONL as its streaming HTTP query format](https://docs.influxdata.com/influxdb3/core/query-data/execute-queries/influxdb-v3-api/).
 
 The typed deployment policy assigns InfluxDB's query execution pool 70% of the
@@ -44,6 +45,15 @@ This changes the engine's internal pool ceiling, not the runner's CPU or RAM.
 The native million-record check retains failures from sorting or deduplication;
 query memory errors remain failures.
 
+InfluxDB snapshots each completed WAL file through its native
+[`wal-files-per-snapshot` setting](https://docs.influxdata.com/influxdb3/core/reference/config-options/#wal-files-per-snapshot).
+The default 600-file interval otherwise leaves this short million-record load in
+the mutable buffer: the pinned engine materializes that buffer while planning
+each query, so sixteen point reads can simultaneously copy the complete corpus.
+Native snapshots make Parquet available during the load and bound this planning
+pressure. The line protocol, `no_sync=false` acknowledgement, independent clients,
+operation deadlines and complete result checks stay the same.
+
 ## Deployment and qualification limits
 
 The added native deployments currently implement one node. TimescaleDB, SQL Server,
@@ -57,6 +67,15 @@ replicas.
 Aspire owns the official pinned container images, anonymous session volumes,
 readiness and joined cleanup. Native SQL readiness opens a real provider
 connection and executes `SELECT 1`; HTTP targets use their server health endpoints.
+The complete Benchmarks workflow uses the Ubuntu 22.04 runner family so preparation
+and every worker share one declared OS baseline. Its published kernel is 6.8;
+MongoDB's native startup guard rejects kernels 6.19 through 7.0.13 because of its
+TCMalloc dependency. Preparation and worker logs record the actual host and Docker
+kernels. The pinned MongoDB image and its startup guard are unchanged. Actual
+kernel, CPU and memory evidence still determines resource equivalence; a common
+runner label alone does not qualify figures. See the
+[runner image inventory](https://github.com/actions/runner-images/blob/main/images/ubuntu/Ubuntu2204-Readme.md)
+and [MongoDB's kernel requirement](https://www.mongodb.com/docs/manual/release-notes/8.0/).
 SQL Server uses Developer edition for nonproduction benchmarks and requires the
 Linux x86-64 runner. Its local execution on ARM is not qualified.
 

docs/Features/BenchmarkComparisons/SessionExecution.md

@@ -158,7 +158,8 @@ matrix, restarts or separate benchmark scenario.
   OpenSearch controls cover one million actual documents per topology. The HelixDB
   and ClickHouse controls also load one million documents and build the session indexes before
   mutations and complete readback. InfluxDB's million-record control verifies its
-  native JSONL readback, an update, cancellation during streaming and healthy
+  native JSONL readback, an update, sixteen concurrent point readers, cancellation
+  during streaming and healthy
   continuation. The eight-control run has a 40-minute timeout;
   each operation keeps its 30-second deadline and document cleanup keeps its
   20-minute deadline. All native