Skip to content

TechCrazi/StressTest

BranchesTags

Repository files navigation

Microsoft SQL stress test service (.NET 8)

ASP.NET Core service to pressure-test Traefik, EKS, and SQL Server (RDS SQL Web). Supports HTTP and gRPC writes with random/HL7/passthrough payloads (default 1 KB–50 MB), optional SQL disable, ring-based aggregation, OTLP telemetry (service.name = sql-stress), and a separate SQL read dashboard. Kestrel/gRPC limits are raised to 60 MB to allow 50 MB payloads.

Features

  • Three roles via APP_ROLE:
    • ingester: handles /write and gRPC writes; can disable SQL to isolate LB/app behavior.
    • aggregator: ring registry + /dashboard + /stats only (no SQL writes/reads).
    • messages: SQL reader only; serves /messages JSON and /sqlread manual dashboard.
  • HTTP /write and gRPC stress.StressTest/Write insert random/HL7/passthrough payloads (or simulate when SQL is disabled).
  • Auto-creates table stress_writes on startup if missing.
  • Dashboard /dashboard + JSON /stats shows per-second totals, per-pod stats (HTTP/gRPC splits, SQL disabled flag), per-IP metrics; aggregator pod keeps a ring of active pods even when idle.
  • SQL read dashboard /sqlread (messages role) has buttons to fetch:
    • Last 10 rows (size + per-row fetch ms, no payload),
    • Last 5 rows (full payload, truncated to 5 MB per row to avoid OOM).
  • OTLP tracing/metrics (AspNetCore, HttpClient, SqlClient + custom meters sqlstress.http.*, sqlstress.grpc.*), service.name = sql-stress.
  • Ports: 8080 HTTP (write/status/health/dashboards/stats/messages/sqlread), 8081 gRPC (h2c). Traefik IngressRoute splits HTTP vs gRPC and buffers to 64 MB.

Environment variables

Required DB:

  • SQL_SERVER – RDS endpoint.
  • SQL_DATABASE
  • SQL_USER
  • SQL_PASSWORD

Optional DB/tuning (defaults in parentheses):

  • SQL_PORT (1433)
  • SQL_TABLE (stress_writes)
  • SQL_ENCRYPT (true)
  • SQL_TRUST_SERVER_CERT (false)
  • SQL_TIMEOUT_SECONDS (60) – command timeout
  • SQL_CONNECT_TIMEOUT_SECONDS (30)
  • SQL_MAX_POOL_SIZE (200)
  • SQL_MIN_POOL_SIZE (10)
  • MAX_INFLIGHT_SQL (0 = unlimited) – cap concurrent SQL writes per pod; use 8–32 to protect RDS

Payload/behavior:

  • PAYLOAD_MIN_BYTES (1024)
  • PAYLOAD_MAX_BYTES (52428800) – 50 MB
  • DISABLE_SQL (false) – if true, skip DB writes but still count traffic
  • SQL_INSERT_MODE (random) – random generates payload; body uses HTTP POST body as payload (for full end-to-end tests); hl7 generates synthetic HL7 text
  • APP_ROLE (ingester) – ingester | aggregator | messages

Ports and identity:

  • APP_PORT (8080) – HTTP
  • APP_GRPC_PORT (8081) – gRPC
  • POD_NAME (defaults to hostname)
  • POD_NAMESPACE – required for peer discovery when using K8s

Ring/peer aggregation:

  • RING_ENDPOINT – URL to post heartbeat snapshots (e.g., http://sql-stress-aggregator.stresstest.svc.cluster.local/register)
  • PEER_SERVICE_NAME (for K8s endpoint discovery, e.g., sql-stress)
  • PEER_SERVICE_PORT (8080)
  • PEER_DASHBOARD_URLS – optional comma-separated base URLs to pull peer /stats?scope=local

Telemetry:

  • OTLP_ENDPOINT – OTLP gRPC/HTTP endpoint (e.g., http://otel-collector:4317)
  • OTLP_HEADERS – extra headers (e.g., authorization=Bearer abc123)
  • OTLP_INSECURE (false) – skip TLS validation

Other:

  • DOTNET_SYSTEM_GLOBALIZATION_INVARIANT (0 in manifests to keep ICU)
  • SQL_TABLE (stress_writes)

Endpoints

  • GET /healthz – DB ping (SQL required).
  • GET /status – pod status snapshot.
  • GET|POST /write – single write (or simulated if DISABLE_SQL=true).
  • GET /messages?limit=10 – latest rows from SQL; size-only by default.
  • GET /messages?limit=5&mode=full – full message payload (truncated to 5 MB per row) + per-row fetch duration.
  • GET /sqlread – manual SQL read dashboard (messages role).
  • GET /dashboard – HTML dashboard (aggregator role).
  • GET /stats – JSON used by the dashboard (aggregated ring-aware).
  • gRPC stress.StressTest/Write – inserts random payload.
  • gRPC stress.StressTest/Healthz – DB health.
  • POST /register – heartbeat ingestion (aggregator).

Build and run locally

dotnet restore

SQL_SERVER="<rds-endpoint>" \
SQL_USER="user" \
SQL_PASSWORD="pass" \
SQL_DATABASE="dbname" \
dotnet run

Container image

docker buildx build --platform linux/amd64,linux/arm64 -t ghcr.io/techcrazi/sql-stress:latest . --push

Container Scan via Trivy

Install Trivy

brew install trivy

Scan Image

trivy image ghcr.io/techcrazi/sql-stress:latest

Container Scan via Slim

Install Slim MAC

brew install docker-slim

Install Slim Windows

  • Enable WSL on Windows Desktop
  • Install Docker Desktop
  • Install Ubuntu WSL image
wsl --install -d Ubuntu

  • Update Docker Desktop Settings

  • Open Docker Desktop → Settings

  • Go to:

  • Resources → WSL Integration

  • Turn ON:

    • Enable integration with my default WSL distro
    • Ubuntu

  • Click Apply & Restart

  • SSH into Ubuntu WSL

  • Install Slim

curl -sL https://raw.githubusercontent.com/slimtoolkit/slim/master/scripts/install-slim.sh | sudo -E bash -

Run SQL Container for testing

docker run -d \
  --name sql-test \
  --platform linux/amd64 \
  -e "ACCEPT_EULA=Y" \
  -e "MSSQL_SA_PASSWORD=StrongPass123" \
  -p 1433:1433 \
  mcr.microsoft.com/mssql/server:2022-latest
Scan & Build Image AMD64 (On Intel or AMD Processor)
slim build \
  --target ghcr.io/techcrazi/sql-stress:latest \
  --tag ghcr.io/techcrazi/sql-stress:slim-amd64 \
  --image-build-arch amd64 \
  --publish-port 9080:8080 \
  --publish-port 9081:8081 \
  --http-probe-cmd 'GET:/healthz' \
  --http-probe-cmd 'GET:/status' \
  --http-probe-cmd 'GET:/write' \
  --http-probe-cmd 'GET:/dashboard' \
  --http-probe-cmd 'GET:/messages?limit=10' \
  --http-probe-cmd 'GET:/messages?limit=5&mode=full' \
  --http-probe-cmd 'GET:/sqlread' \
  --include-path '/app' \
  --include-path '/usr/share/dotnet/shared/Microsoft.NETCore.App' \
  --continue-after=probe \
  --copy-meta-artifacts ./slim-artifacts \
  --env SQL_SERVER=sql-test \
  --env SQL_DATABASE=stress \
  --env SQL_USER=sa \
  --env SQL_PASSWORD=StrongPass123 \
  --env SQL_ENCRYPT=false \
  --env SQL_TRUST_SERVER_CERT=true \
  --env SQL_CONNECT_TIMEOUT_SECONDS=1

  • --http-probe-cmd > Application endpoint that Slim will check against

  • --include-path '/app' > Retain OTEL DLLs

  • --include-path '/usr/share/dotnet/shared/Microsoft.NETCore.App' > Retain .NET DLLs

  • Original Image: 342.85 MB

  • Slim Image: 217.87 MB

Scan & Build Image ARM64 (On Apple or ARM Processor)
slim build \
  --target ghcr.io/techcrazi/sql-stress:latest \
  --tag ghcr.io/techcrazi/sql-stress:slim-arm64 \
  --image-build-arch arm64 \
  --publish-port 9080:8080 \
  --publish-port 9081:8081 \
  --http-probe-cmd 'GET:/healthz' \
  --http-probe-cmd 'GET:/status' \
  --http-probe-cmd 'GET:/write' \
  --http-probe-cmd 'GET:/dashboard' \
  --http-probe-cmd 'GET:/messages?limit=10' \
  --http-probe-cmd 'GET:/messages?limit=5&mode=full' \
  --http-probe-cmd 'GET:/sqlread' \
  --include-path '/app' \
  --include-path '/usr/share/dotnet/shared/Microsoft.NETCore.App' \
  --copy-meta-artifacts ./slim-artifacts \
  --env SQL_SERVER=sql-test \
  --env SQL_DATABASE=stress \
  --env SQL_USER=sa \
  --env SQL_PASSWORD=StrongPass123 \
  --env SQL_ENCRYPT=false \
  --env SQL_TRUST_SERVER_CERT=true \
  --env SQL_CONNECT_TIMEOUT_SECONDS=1

  • --http-probe-cmd > Application endpoint that Slim will check against

  • --include-path '/app' > Retain OTEL DLLs

  • --include-path '/usr/share/dotnet/shared/Microsoft.NETCore.App' > Retain .NET DLLs

  • Original Image: 372.56 MB

  • Slim Image: 226.60 MB

Image Testing
slim build \
  --target ghcr.io/techcrazi/sql-stress:latest \
  --tag sql-stress:slim \
  --publish-port 9080:8080 \
  --publish-port 9081:8081 \
  --continue-after=enter \
  --copy-meta-artifacts ./slim-artifacts \
  --env SQL_SERVER=sql-test \
  --env SQL_DATABASE=stress \
  --env SQL_USER=sa \
  --env SQL_PASSWORD=StrongPass123 \
  --env SQL_ENCRYPT=false \
  --env SQL_TRUST_SERVER_CERT=true \
  --env SQL_CONNECT_TIMEOUT_SECONDS=1 \
  --env DISABLE_SQL=true
Push gRPC message
grpcurl -vv -plaintext \
  -proto Protos/stress.proto \
  -d '{"sizeBytes":1024}' \
  localhost:9081 stress.StressTest/Write
Push Slim Image to GHCR
docker push ghcr.io/techcrazi/sql-stress:slim-amd64
docker push ghcr.io/techcrazi/sql-stress:slim-arm64

docker manifest create ghcr.io/techcrazi/sql-stress:slim \
  --amend ghcr.io/techcrazi/sql-stress:slim-amd64 \
  --amend ghcr.io/techcrazi/sql-stress:slim-arm64

docker manifest push ghcr.io/techcrazi/sql-stress:slim

Kubernetes (EKS + Traefik)

  1. Edit k8s/stress-app.yaml for your image, host, and DB settings, then apply:
kubectl --kubeconfig k8s/multicare-dev-eks.config apply -f k8s/stress-app.yaml
  1. Deployments/Services:
    • sql-stress-ingester (HTTP /write, gRPC) on ports 80->8080 and 81->8081; APP_ROLE=ingester.
    • sql-stress-aggregator (ring + /dashboard + /stats) on 80->8080; APP_ROLE=aggregator, DISABLE_SQL=true.
    • sql-stress-messages (SQL reader only; /messages, /sqlread) on 80->8080; APP_ROLE=messages, DISABLE_SQL=true (writes blocked; reads allowed).
  2. IngressRoute splits:
    • gRPC /stress.StressTest → ingester (h2c, buffered).
    • HTTP /write|/healthz|/status → ingester.
    • /dashboard|/stats → aggregator.
    • /messages|/sqlread → messages reader.
    • Traefik middleware sql-stress-buffer allows 64 MB request bodies.

Notes and limits

  • Max request size/message size raised to 60 MB to permit 50 MB payloads; ensure Traefik and any proxies allow similar limits.
  • Table schema (auto-created): id UNIQUEIDENTIFIER, created_at DATETIME2, payload_size INT, payload VARBINARY(MAX).
  • /messages full mode truncates per-row payload to 5 MB to avoid large in-memory blobs; size-only mode is lightweight.
  • Set PAYLOAD_MIN_BYTES = PAYLOAD_MAX_BYTES for deterministic size. For heavier loads, tune MAX_INFLIGHT_SQL, pool settings, and RDS instance size.

K6 Testing

  • Update k6/full-loadtest.js file

    • Test intensity

      • const VUS → 50 number of concurrent virtual users
      • const TEST_DURATION → "1h" how long to run the test

    • Common size units

      • const KB → 1024 file size in KB
      • const MB → 1024 * 1024 file size in MB

    • File size buckets (in bytes)

      • const SMALL_MIN → 1 * KB
      • const SMALL_MAX → 100 * KB
      • const MED_MIN → 100 * KB
      • const MED_MAX → 1 * MB
      • const LARGE_MIN → 1 * MB
      • const LARGE_MAX → 10 * MB
      • const HUGE_MIN → 10 * MB
      • const HUGE_MAX → 50 * MB

    • Bucket distribution (must roughly sum to 1.0)

      • const P_SMALL → 0.45 45% of messages in 1–100 KB
      • const P_MED → 0.45 45% in 100 KB–1 MB
      • const P_LARGE → 0.05 05% in 1–10 MB
      • const P_HUGE → 0.05 05% in 10–50 MB

    • Protocol split

      • const GRPC_RATIO → 0.8; 0.8 → 80% gRPC, 20% HTTP

    • Endpoint configuration - Point this to the right ingest FQDN

      • const BASE_HOST → "stresstest.mc.dev.testme.com"

  • Run K6 load test:

    • .\k6.exe run full-loadtest.js
    • .\k6.exe cloud run full-loadtest.js

About

No description, website, or topics provided.

Resources

Readme
Activity

Stars

1 star

Watchers

0 watching

Forks

0 forks
Report repository

Releases

No releases published

Packages

 
 
 

Languages