quickdid / test-scripts / docker-test.sh

#!/bin/bash

# Comprehensive test script for Telegraf/TimescaleDB metrics setup
# This script validates the entire metrics pipeline

set -e

echo "========================================="
echo "Telegraf/TimescaleDB Metrics Test Suite"
echo "========================================="
echo ""

# Check if Docker is running
if ! docker info > /dev/null 2>&1; then
    echo "❌ Docker is not running. Please start Docker first."
    exit 1
fi

# Function to wait for a service to be healthy
wait_for_service() {
    local service=$1
    local max_attempts=30
    local attempt=1
    
    echo -n "Waiting for $service to be healthy"
    while [ $attempt -le $max_attempts ]; do
        if docker-compose ps $service | grep -q "healthy"; then
            echo " ✅"
            return 0
        fi
        echo -n "."
        sleep 2
        attempt=$((attempt + 1))
    done
    echo " ❌"
    echo "Service $service failed to become healthy after $max_attempts attempts"
    return 1
}

# Function to run SQL query
run_query() {
    docker exec -i timescaledb psql -U postgres -d metrics -t -c "$1" 2>/dev/null
}

# Function to check table exists
check_table() {
    local table=$1
    local result=$(run_query "SELECT EXISTS (SELECT 1 FROM information_schema.tables WHERE table_name = '$table');")
    if [[ "$result" =~ "t" ]]; then
        echo "✅ Table '$table' exists"
        return 0
    else
        echo "❌ Table '$table' does not exist"
        return 1
    fi
}

# Navigate to the metrics-stack directory (create if needed)
if [ ! -d "metrics-stack" ]; then
    echo "Creating metrics-stack directory..."
    mkdir -p metrics-stack/telegraf
    mkdir -p metrics-stack/test-scripts
    mkdir -p metrics-stack/init-scripts
fi

cd metrics-stack

# Create .env file if it doesn't exist
if [ ! -f ".env" ]; then
    echo "Creating .env file..."
    cat > .env << 'EOF'
# PostgreSQL/TimescaleDB Configuration
POSTGRES_DB=metrics
POSTGRES_USER=postgres
POSTGRES_PASSWORD=secretpassword

# Telegraf Database User
TELEGRAF_DB_USER=postgres
TELEGRAF_DB_PASSWORD=secretpassword

# TimescaleDB Settings
TIMESCALE_TELEMETRY=off
EOF
fi

# Copy configuration files if they don't exist
if [ ! -f "telegraf/telegraf.conf" ]; then
    echo "Creating telegraf.conf..."
    cat > telegraf/telegraf.conf << 'EOF'
[agent]
  interval = "10s"
  round_interval = true
  metric_batch_size = 1000
  metric_buffer_limit = 10000
  collection_jitter = "0s"
  flush_interval = "10s"
  flush_jitter = "0s"
  precision = ""
  debug = false
  quiet = false
  hostname = "telegraf-agent"
  omit_hostname = false

[[inputs.statsd]]
  service_address = ":8125"
  protocol = "udp"
  delete_gauges = true
  delete_counters = true
  delete_sets = true
  delete_timings = true
  percentiles = [50, 90, 95, 99]
  metric_separator = "."
  allowed_pending_messages = 10000
  datadog_extensions = true
  datadog_distributions = true

[[outputs.postgresql]]
  connection = "host=timescaledb user=${TELEGRAF_DB_USER} password=${TELEGRAF_DB_PASSWORD} dbname=${POSTGRES_DB} sslmode=disable"
  schema = "public"
  create_templates = [
    '''CREATE TABLE IF NOT EXISTS {{.table}} ({{.columns}})''',
    '''SELECT create_hypertable({{.table|quoteLiteral}}, 'time', if_not_exists => TRUE)''',
  ]
  tags_as_jsonb = true
  fields_as_jsonb = false
EOF
fi

# Copy docker-compose.yml if it doesn't exist
if [ ! -f "docker-compose.yml" ]; then
    echo "Creating docker-compose.yml..."
    cat > docker-compose.yml << 'EOF'
version: '3.8'

services:
  timescaledb:
    image: timescale/timescaledb:latest-pg17
    container_name: timescaledb
    restart: unless-stopped
    environment:
      POSTGRES_DB: ${POSTGRES_DB}
      POSTGRES_USER: ${POSTGRES_USER}
      POSTGRES_PASSWORD: ${POSTGRES_PASSWORD}
      TIMESCALE_TELEMETRY: ${TIMESCALE_TELEMETRY}
    ports:
      - "5442:5432"
    volumes:
      - timescale_data:/home/postgres/pgdata/data
      - ./init-scripts:/docker-entrypoint-initdb.d:ro
    command:
      - postgres
      - -c
      - shared_buffers=256MB
      - -c
      - effective_cache_size=1GB
      - -c
      - maintenance_work_mem=64MB
      - -c
      - work_mem=8MB
    healthcheck:
      test: ["CMD-SHELL", "pg_isready -U ${POSTGRES_USER} -d ${POSTGRES_DB}"]
      interval: 10s
      timeout: 5s
      retries: 5
    networks:
      - metrics_network

  telegraf:
    image: telegraf:1.35
    container_name: telegraf
    restart: unless-stopped
    environment:
      TELEGRAF_DB_USER: ${TELEGRAF_DB_USER}
      TELEGRAF_DB_PASSWORD: ${TELEGRAF_DB_PASSWORD}
      POSTGRES_DB: ${POSTGRES_DB}
    ports:
      - "8125:8125/udp"
    volumes:
      - ./telegraf/telegraf.conf:/etc/telegraf/telegraf.conf:ro
    depends_on:
      timescaledb:
        condition: service_healthy
    networks:
      - metrics_network
    command: ["telegraf", "--config", "/etc/telegraf/telegraf.conf"]

networks:
  metrics_network:
    driver: bridge

volumes:
  timescale_data:
EOF
fi

# Create init script
if [ ! -f "init-scripts/01-init.sql" ]; then
    echo "Creating init script..."
    cat > init-scripts/01-init.sql << 'EOF'
-- Enable TimescaleDB extension
CREATE EXTENSION IF NOT EXISTS timescaledb;
CREATE EXTENSION IF NOT EXISTS pg_stat_statements;
EOF
fi

echo ""
echo "Step 1: Starting Docker services..."
echo "========================================="
docker-compose down -v 2>/dev/null || true
docker-compose up -d

echo ""
echo "Step 2: Waiting for services to be healthy..."
echo "========================================="
wait_for_service timescaledb
sleep 5  # Extra time for Telegraf to connect

echo ""
echo "Step 3: Sending test metrics..."
echo "========================================="

# Send various types of metrics
echo "Sending counter metrics..."
for i in {1..5}; do
  echo "quickdid.http.request.count:1|c|#method:GET,path:/resolve,status:200" | nc -u -w0 localhost 8125
  echo "quickdid.http.request.count:1|c|#method:POST,path:/api,status:201" | nc -u -w0 localhost 8125
done

echo "Sending gauge metrics..."
echo "quickdid.resolver.rate_limit.available_permits:10|g" | nc -u -w0 localhost 8125
sleep 1
echo "quickdid.resolver.rate_limit.available_permits:5|g" | nc -u -w0 localhost 8125

echo "Sending timing metrics..."
for i in {1..10}; do
  duration=$((RANDOM % 100 + 10))
  echo "quickdid.http.request.duration_ms:${duration}|ms|#method:GET,path:/resolve,status:200" | nc -u -w0 localhost 8125
done

echo "Sending histogram metrics..."
for i in {1..5}; do
  resolution_time=$((RANDOM % 500 + 50))
  echo "quickdid.resolver.resolution_time:${resolution_time}|h|#resolver:redis" | nc -u -w0 localhost 8125
done

echo "Waiting 15 seconds for Telegraf to flush metrics..."
sleep 15

echo ""
echo "Step 4: Verifying table creation..."
echo "========================================="

# Check if tables were created
check_table "quickdid.http.request.count"
check_table "quickdid.http.request.duration_ms"
check_table "quickdid.resolver.rate_limit.available_permits"
check_table "quickdid.resolver.resolution_time"

echo ""
echo "Step 5: Verifying data insertion..."
echo "========================================="

# Check row counts
for table in "quickdid.http.request.count" "quickdid.http.request.duration_ms" "quickdid.resolver.rate_limit.available_permits" "quickdid.resolver.resolution_time"; do
    count=$(run_query "SELECT COUNT(*) FROM \"$table\";" | tr -d ' ')
    if [ "$count" -gt 0 ]; then
        echo "✅ Table '$table' has $count rows"
    else
        echo "❌ Table '$table' is empty"
    fi
done

echo ""
echo "Step 6: Testing JSONB tag queries..."
echo "========================================="

# Test JSONB tag filtering
result=$(run_query "SELECT COUNT(*) FROM \"quickdid.http.request.count\" WHERE tags->>'method' = 'GET';" | tr -d ' ')
if [ "$result" -gt 0 ]; then
    echo "✅ JSONB tag filtering works (found $result GET requests)"
else
    echo "❌ JSONB tag filtering failed"
fi

echo ""
echo "Step 7: Testing TimescaleDB functions..."
echo "========================================="

# Test time_bucket function
result=$(run_query "SELECT COUNT(*) FROM (SELECT time_bucket('1 minute', time) FROM \"quickdid.http.request.count\" GROUP BY 1) t;" | tr -d ' ')
if [ "$result" -gt 0 ]; then
    echo "✅ time_bucket function works"
else
    echo "❌ time_bucket function failed"
fi

# Check if hypertables were created
hypertable_count=$(run_query "SELECT COUNT(*) FROM timescaledb_information.hypertables WHERE hypertable_name LIKE 'quickdid%';" | tr -d ' ')
if [ "$hypertable_count" -gt 0 ]; then
    echo "✅ Found $hypertable_count hypertables"
else
    echo "❌ No hypertables found"
fi

echo ""
echo "Step 8: Running comprehensive query tests..."
echo "========================================="

# Run the verify-queries.sql script if it exists
if [ -f "../test-scripts/verify-queries.sql" ]; then
    echo "Running verify-queries.sql..."
    docker exec -i timescaledb psql -U postgres -d metrics < ../test-scripts/verify-queries.sql > query_results.txt 2>&1
    if [ $? -eq 0 ]; then
        echo "✅ All queries executed successfully"
        echo "   Results saved to query_results.txt"
    else
        echo "❌ Some queries failed. Check query_results.txt for details"
    fi
else
    echo "⚠️  verify-queries.sql not found, skipping comprehensive query tests"
fi

echo ""
echo "========================================="
echo "Test Summary"
echo "========================================="

# Generate summary
failures=0
successes=0

# Count successes and failures from the output
if check_table "quickdid.http.request.count" > /dev/null 2>&1; then
    successes=$((successes + 1))
else
    failures=$((failures + 1))
fi

if [ "$hypertable_count" -gt 0 ]; then
    successes=$((successes + 1))
else
    failures=$((failures + 1))
fi

echo ""
if [ $failures -eq 0 ]; then
    echo "✅ All tests passed successfully!"
    echo ""
    echo "You can now:"
    echo "1. Connect to the database: docker exec -it timescaledb psql -U postgres -d metrics"
    echo "2. View logs: docker-compose logs -f"
    echo "3. Send more metrics: echo 'metric.name:value|type|#tag:value' | nc -u -w0 localhost 8125"
    echo "4. Stop services: docker-compose down"
else
    echo "⚠️  Some tests failed. Please check the output above for details."
    echo ""
    echo "Troubleshooting tips:"
    echo "1. Check Telegraf logs: docker-compose logs telegraf"
    echo "2. Check TimescaleDB logs: docker-compose logs timescaledb"
    echo "3. Verify connectivity: docker exec telegraf telegraf --test"
fi

echo ""
echo "Test complete!"