AppDynamics Prometheus モニタリング拡張機能

概要

AppDynamics Prometheus モニタリング拡張機能を使用すると、Prometheus 互換のエクスポータからメトリックをシームレスに収集し、カスタム指標として AppDynamics コントローラに転送できます。この統合により、Prometheus の拡張エクスポータエコシステムを活用して、システムリソース、データベース、コンテナ、GPU、カスタムアプリケーションなどのさまざまなインフラストラクチャおよびアプリケーション コンポーネントをモニターできます。

主な機能

  • ノードエクスポータ、cAdvisor、DCGM（NVIDIA GPU）、データベースエクスポータ、カスタム端末などの任意の Prometheus エクスポータからの [Collect Metrics]。
  • フィルタリング、集約、およびパスマッピング（計算された数式とラベルベースのパスを含む）に対するビルトインサポートによる [Transform Metrics]。
  • 統合モニタリング、分析、およびアラートのための AppDynamics コントローラにおける [Customize Metrics ]。
  • モジュラー構成を使用した少数から数百のエクスポータからの [Seamless Scalability]。

ワークフロー

Prometheus モニタリング拡張機能を正常に統合するには、次の手順を実行する必要があります。
  1. controller-info.xml で拡張機能を有効にし、モニター対象のホストでノードエクスポータなどの Prometheus エクスポータを開始します。
  2. 各 YAML ファイルでグローバル設定とエクスポータ固有の設定を行います。
  3. マシンエージェントを再起動します。

Prometheus 拡張機能の使用

  1. 以下の手順で Controller-info.xml を更新します。
    CODE 
    <prometheus-enabled>true</prometheus-enabled>
  2. ノードエクスポータ、cAdvisor などの必要な Prometheus エクスポータを実行します。エクスポータがすでに実行されている場合は、この手順を無視します。

YAML ファイルの設定

この拡張機能により、conf/prometheus の下にある 2 つの新しい構成ファイルが導入されます。

  • prometheus-config.yaml：この YAML には、すべてのエクスポータのグローバル設定と、アクティブなエクスポータ設定の参照リストが含まれています。設定オプションと使用方法については、prometheus-config-reference.yaml を参照してください。

    JAVASCRIPT 
    # ==================================================================================
# AppDynamics Prometheus Configuration - Reference Template
# ==================================================================================
# This is a comprehensive reference file documenting all available configuration
# options for the AppDynamics Prometheus integration.
# 
# Use this file as a reference when creating your own prometheus-config.yaml
#
# Documentation: <add link>
# Version: 1.0
# ==================================================================================

prometheus:
  
  # ================================================================================
  # GLOBAL SETTINGS
  # ================================================================================
  # These settings apply to all exporters unless explicitly overridden in the
  # individual exporter configuration files.
  # ================================================================================
  global:
    
    # --------------------------------------------------------------------------
    # scrapeInterval (integer, seconds)
    # --------------------------------------------------------------------------
    # How often to scrape metrics from all exporters
    # 
    # Default: 60 seconds
    # Range: 10-3600 seconds
    # Recommendation: 
    #   - 30-60s for most use cases
    #   - 10-30s for high-frequency monitoring
    #   - 120-300s for low-frequency infrastructure metrics
    # 
    # Note: Can be overridden per exporter
    # --------------------------------------------------------------------------
    scrapeInterval: 60
    
    # --------------------------------------------------------------------------
    # scrapeTimeout (integer, seconds)
    # --------------------------------------------------------------------------
    # Maximum time to wait for a scrape request to complete
    # 
    # Default: 10 seconds
    # Range: 5-60 seconds
    # Recommendation: Set to < scrapeInterval to avoid overlapping scrapes
    # 
    # Note: If a scrape takes longer than this, it will be cancelled and an
    # error will be logged. The next scrape will still occur on schedule.
    # --------------------------------------------------------------------------
    scrapeTimeout: 10
    
    # --------------------------------------------------------------------------
    # connectTimeout (integer, seconds)
    # --------------------------------------------------------------------------
    # Maximum time to wait for HTTP connection establishment
    # 
    # Default: 5 seconds
    # Range: 1-30 seconds
    # Recommendation: 5-10s for remote exporters, 2-5s for local exporters
    # --------------------------------------------------------------------------
    connectTimeout: 5
    
    # --------------------------------------------------------------------------
    # maxMetricsPerScrape (integer)
    # --------------------------------------------------------------------------
    # Maximum number of metrics to collect in a single scrape operation
    # 
    # Default: 10000
    # Range: 100-100000
    # Purpose: Prevents memory issues and runaway metric collection
    # 
    # Note: If an exporter returns more metrics than this limit, only the
    # first N metrics will be processed and a warning will be logged.
    # --------------------------------------------------------------------------
    maxMetricsPerScrape: 10000
    
    # --------------------------------------------------------------------------
    # defaultMetricPathPrefix (string)
    # --------------------------------------------------------------------------
    # Default prefix for all metrics 
    # 
    # Default: "Custom Metrics|Prometheus"
    # Examples:
    #   - "Custom Metrics|Prometheus"
    #   - "Custom Metrics|Infrastructure"
    #   - "Hardware Resources"
    # 
    # Note: Individual exporters can override this with their own prefix
    # --------------------------------------------------------------------------
    defaultMetricPathPrefix: "Custom Metrics|Prometheus"
    
    # --------------------------------------------------------------------------
    # includeUnmappedMetrics (boolean)
    # --------------------------------------------------------------------------
    # Whether to include metrics that don't have explicit mappings
    # 
    # Default: false
    # Options: true | false
    # 
    # When true: All metrics from exporters will be reported, even if they
    #           don't have a mapping rule defined
    # When false: Only metrics with explicit mappings will be reported
    # 
    # Recommendation: Start with false to avoid metric explosion, then enable
    #                for specific exporters as needed
    # --------------------------------------------------------------------------
    includeUnmappedMetrics: false
    
    # --------------------------------------------------------------------------
    # userAgent (string)
    # --------------------------------------------------------------------------
    # User-Agent header sent with all HTTP requests to exporters
    # 
    # Default: "AppDynamics-MachineAgent-Prometheus/1.0"
    # Purpose: Helps identify the scraper in exporter logs and metrics
    # 
    # Format: Usually "<Product>/<Version>"
    # --------------------------------------------------------------------------
    userAgent: "AppDynamics-MachineAgent-Prometheus/1.0"
    
    # --------------------------------------------------------------------------
    # THREADING CONFIGURATION
    # --------------------------------------------------------------------------
    # Controls the thread pool used for concurrent scraping operations
    # --------------------------------------------------------------------------
    threading:
      
      # ------------------------------------------------------------------------
      # poolSize (integer)
      # ------------------------------------------------------------------------
      # Number of worker threads in the scraper thread pool
      # 
      # Default: max(4, number_of_cpu_cores)
      # Range: 1-100
      # 
      # Recommendation:
      #   - 1-2 threads: For 1-2 exporters or limited resources
      #   - 4-8 threads: For 3-10 exporters (typical use case)
      #   - 10+ threads: For 10+ exporters with sufficient CPU/memory
      # 
      # Note: Each scrape operation uses one thread. If you have 5 exporters
      # scraping every 60s, a pool of 5-10 threads is sufficient.
      # ------------------------------------------------------------------------
      poolSize: 4
      
      # ------------------------------------------------------------------------
      # threadNamePrefix (string)
      # ------------------------------------------------------------------------
      # Prefix for thread names in the pool (useful for debugging and monitoring)
      # 
      # Default: "appd-prom-scraper"
      # 
      # Purpose: Makes it easy to identify Prometheus scraper threads in
      # thread dumps and profilers
      # 
      # Thread naming pattern: "{prefix}-{number}"
      # Example: "appd-prom-scraper-1", "appd-prom-scraper-2", etc.
      # ------------------------------------------------------------------------
      threadNamePrefix: "appd-prom-scraper"
      
      # ------------------------------------------------------------------------
      # daemonThreads (boolean)
      # ------------------------------------------------------------------------
      # Whether scraper threads should be daemon threads
      # 
      # Default: true
      # Options: true | false
      # 
      # When true: Threads won't prevent JVM shutdown
      # When false: JVM will wait for threads to complete before shutdown
      # 
      # Recommendation: Keep as true (default) for normal operation
      # ------------------------------------------------------------------------
      daemonThreads: true
  
  # ================================================================================
  # EXPORTER REFERENCES
  # ================================================================================
  # List of individual exporter configuration files to load
  # 
  # Each exporter configuration is defined in a separate YAML file, which allows:
  #   - Modular configuration management
  #   - Easy addition/removal of exporters
  #   - Clean separation of concerns
  #   - Version control friendly
  # 
  # File Path Resolution:
  #   - Relative paths are resolved relative to this configuration file's directory
  #   - Absolute paths can also be used
  #   - Example: "exporters/my-exporter.yaml" → "conf/prometheus/exporters/my-exporter.yaml"
  # 
  # Loading Behavior:
  #   - Files are loaded in the order specified
  #   - If a file is missing or invalid, an error is logged and that exporter is skipped
  #   - Other exporters continue to function normally
  # ================================================================================
  exporters:
    
    # Example 1: NVIDIA DCGM GPU Exporter (GPU monitoring)
    - configFile: "exporters/dcgm-exporter.yaml"
    
    # Example 2: Node Exporter (System metrics - CPU, memory, disk, network)
    # Uncomment to enable:
    # - configFile: "exporters/node-exporter.yaml"
    
    # Example 3: Custom application exporter
    # - configFile: "exporters/my-app-exporter.yaml"
    
    # Example 4: Kafka exporter
    # - configFile: "exporters/kafka-exporter.yaml"
    
    # Example 5: MongoDB exporter
    # - configFile: "exporters/mongodb-exporter.yaml"
    
    # You can add as many exporter references as needed
    # Each exporter runs independently with its own scrape schedule


# ==================================================================================
# TROUBLESHOOTING
# ==================================================================================
#
# Problem: No metrics appearing in Controller
# Solution: 
#   - Check Machine Agent logs for errors
#   - Verify exporter is enabled: enabled: true
#   - Ensure generic monitoring is enabled: -Dappdynamics.sim.prometheus.enabled=true
#   - Verify exporter endpoint is accessible
#
# Problem: Too many metrics being collected
# Solution:
#   - Add filters to restrict metric collection
#   - Set includeUnmappedMetrics: false
#   - Reduce maxMetricsPerScrape if needed
#
# Problem: Scrapes timing out
# Solution:
#   - Increase scrapeTimeout
#   - Reduce maxMetricsPerScrape
#   - Check exporter performance
#   - Verify network connectivity
#
# Problem: Metrics delayed or missing
# Solution:
#   - Check scrape interval isn't too long
#   - Verify thread pool size is sufficient
#   - Review logs for scrape failures
#
# ==================================================================================

  • exporters/exporter-template.yaml：この YAML には、エクスポータ固有の設定を行うためのグローバル設定が含まれています。モニターする Prometheus エクスポータごとにこのファイルを複製し、必要に応じてカスタマイズします。

    JAVASCRIPT 
    # ==================================================================================
# AppDynamics Prometheus Exporter Configuration - Template
# ==================================================================================
# This is a comprehensive template documenting all available configuration options
# for a single Prometheus exporter.
# 
# Use this file as a reference when creating configuration for a new exporter.
# 
# Copy this file, rename it (e.g., "my-exporter.yaml"), customize the values,
# and reference it in prometheus-config.yaml
#
# Documentation: <add link>
# Version: 1.0
# ==================================================================================

# ==============================================================================
# EXPORTER IDENTIFICATION
# ==============================================================================

# ------------------------------------------------------------------------------
# name (string, REQUIRED)
# ------------------------------------------------------------------------------
# Unique identifier for this exporter
# 
# Requirements:
#   - Must be unique across all exporters
#   - Use alphanumeric characters, hyphens, and underscores
#   - No spaces allowed
# 
# Examples: "node-exporter", "kafka-broker-1", "app-metrics"
# 
# Purpose: Used in logs, metrics paths, and internal tracking
# ------------------------------------------------------------------------------
name: "example-exporter"

# ------------------------------------------------------------------------------
# type (string, optional)
# ------------------------------------------------------------------------------
# Exporter type/category for documentation and organization
# 
# Common types:
#   - "node" - System metrics (CPU, memory, disk, network)
#   - "dcgm" - NVIDIA GPU metrics
#   - "application" - Application-specific metrics
#   - "database" - Database metrics (MySQL, PostgreSQL, MongoDB)
#   - "messaging" - Message queue metrics (Kafka, RabbitMQ)
#   - "container" - Container runtime metrics (cAdvisor)
#   - "custom" - Custom/proprietary metrics
# 
# Purpose: Helps with categorization and filtering
# ------------------------------------------------------------------------------
type: "custom"

# ------------------------------------------------------------------------------
# enabled (boolean)
# ------------------------------------------------------------------------------
# Whether this exporter should be actively scraped
# 
# Default: true
# Options: true | false
# 
# Use Cases:
#   - Temporarily disable an exporter without removing its configuration
#   - Environment-specific enablement (dev vs prod)
#   - A/B testing of different configurations
# ------------------------------------------------------------------------------
enabled: true


# ==============================================================================
# SCRAPE CONFIGURATION
# ==============================================================================

# ------------------------------------------------------------------------------
# scrapeInterval (integer, seconds, optional)
# ------------------------------------------------------------------------------
# How often to scrape metrics from this exporter
# 
# Default: Inherits from global.scrapeInterval
# Range: 10-3600 seconds
# 
# Override the global setting when:
#   - This exporter updates metrics more/less frequently than others
#   - You want to reduce load on a slow exporter
#   - You need high-frequency monitoring for specific metrics
# 
# ------------------------------------------------------------------------------
scrapeInterval: 60

# ------------------------------------------------------------------------------
# scrapeTimeout (integer, seconds, optional)
# ------------------------------------------------------------------------------
# Maximum time to wait for this exporter to respond
# 
# Default: Inherits from global.scrapeTimeout
# Range: 5-60 seconds
# 
# Recommendation: Set < scrapeInterval to avoid overlapping scrapes
# 
# Override when:
#   - This exporter is slow to respond (large metric payloads)
#   - Network latency is higher than usual
#   - Exporter performs computation during scrape
# ------------------------------------------------------------------------------
scrapeTimeout: 10


# ==============================================================================
# SERVICE DISCOVERY
# ==============================================================================
# Configuration for locating and connecting to the Prometheus exporter endpoint
# 
# Currently supported discovery type:
#   - static: Direct connection using host/port/path
# ==============================================================================
serviceDiscovery:
  
  # ----------------------------------------------------------------------------
  # type (string)
  # ----------------------------------------------------------------------------
  # Service discovery mechanism
  # 
  # Supported Values:
  #   - "static"      : Direct host/IP connection (bare metal, VMs, or Kubernetes DNS)
  #   - "kubernetes"  : Kubernetes service discovery (constructs DNS automatically)
  # 
  # Default: "static"
  # ----------------------------------------------------------------------------
  type: "static"
  
  # ============================================================================
  # STATIC DISCOVERY CONFIGURATION
  # ============================================================================
  # Use this for bare metal, VMs, or when you want to specify the full hostname
  # (including Kubernetes service DNS names manually)
  
  # ----------------------------------------------------------------------------
  # host (string, REQUIRED for static)
  # ----------------------------------------------------------------------------
  # Hostname or IP address of the exporter
  # 
  # Examples:
  #   - "localhost"                                    - Local exporters
  #   - "192.168.1.100"                               - Direct IP address
  #   - "metrics.example.com"                         - DNS hostname
  #   - "exporter.namespace.svc.cluster.local"        - Kubernetes service (manual)
  # 
  # Environment Variables: Supports ${env:VAR_NAME:default} syntax
  # Example: "${env:EXPORTER_HOST:localhost}"
  # ----------------------------------------------------------------------------
  host: "localhost"
  
  # ----------------------------------------------------------------------------
  # port (integer, REQUIRED)
  # ----------------------------------------------------------------------------
  # TCP port number where the exporter is listening
  # 
  # Range: 1-65535
  # 
  # Common Prometheus exporter ports:
  #   - 9100: Node Exporter
  #   - 9400: DCGM Exporter (GPU)
  #   - 9090: Prometheus itself
  #   - 9308: Kafka Exporter
  #   - 9216: MongoDB Exporter
  #   - 9104: MySQL Exporter
  # 
  # Check your exporter's documentation for the default port
  # 
  # Can use environment variables: ${env:EXPORTER_PORT:9090}
  # ----------------------------------------------------------------------------
  port: 9090
  
  # ----------------------------------------------------------------------------
  # path (string)
  # ----------------------------------------------------------------------------
  # URL path to the metrics endpoint
  # 
  # Default: "/metrics" (Prometheus standard)
  # 
  # Examples:
  #   - "/metrics"                - Standard Prometheus endpoint
  #   - "/actuator/prometheus"    - Spring Boot Actuator
  #   - "/federate"               - Prometheus federation endpoint
  #   - "/api/metrics"            - Custom application endpoint
  # 
  # Can use environment variables: ${env:METRICS_PATH:/metrics}
  # ----------------------------------------------------------------------------
  path: "/metrics"
  
  # ============================================================================
  # KUBERNETES DISCOVERY CONFIGURATION
  # ============================================================================
  # When type is "kubernetes", the Machine Agent will automatically construct
  # the Kubernetes service DNS name as: <serviceName>.<namespace>.svc.cluster.local
  #
  # Example Configuration (uncomment to use):
  # ----------------------------------------------------------------------------
  # type: "kubernetes"
  # 
  # # ----------------------------------------------------------------------------
  # # serviceName (string, REQUIRED for kubernetes)
  # # ----------------------------------------------------------------------------
  # # The Kubernetes service name
  # # 
  # # Examples:
  # #   - "dcgm-exporter"
  # #   - "node-exporter"
  # #   - "prometheus-server"
  # #
  # # Can use environment variables: ${env:K8S_SERVICE_NAME:dcgm-exporter}
  # # ----------------------------------------------------------------------------
  # serviceName: "dcgm-exporter"
  # 
  # # ----------------------------------------------------------------------------
  # # namespace (string, REQUIRED for kubernetes)
  # # ----------------------------------------------------------------------------
  # # The Kubernetes namespace where the service is deployed
  # # 
  # # Examples:
  # #   - "default"
  # #   - "monitoring"
  # #   - "gpu-operator"
  # #
  # # Can use environment variables: ${env:K8S_NAMESPACE:default}
  # # ----------------------------------------------------------------------------
  # namespace: "gpu-operator"
  # 
  # # Port and path work the same way as static discovery
  # port: 9400
  # path: "/metrics"
  # ----------------------------------------------------------------------------
  
  # ----------------------------------------------------------------------------
  # scheme (string, optional)
  # ----------------------------------------------------------------------------
  # Protocol scheme for the connection
  # 
  # Supported: "http", "https"
  # Default: "http"
  # 
  # Use "https" if your exporter uses TLS/SSL
  # ----------------------------------------------------------------------------
  # scheme: "http"


# ==============================================================================
# AUTHENTICATION
# ==============================================================================
# Configure authentication for accessing the exporter endpoint
# 
# Supported types:
#   - none: No authentication (default)
#   - basic: HTTP Basic Authentication (username/password)
#   - bearer: Bearer token authentication (OAuth2, API tokens)
# ==============================================================================
authentication:
  
  # ----------------------------------------------------------------------------
  # type (string)
  # ----------------------------------------------------------------------------
  # Authentication mechanism
  # 
  # Options: "none" | "basic" | "bearer"
  # Default: "none"
  # ----------------------------------------------------------------------------
  type: "none"
  
  # ----------------------------------------------------------------------------
  # BASIC AUTHENTICATION
  # ----------------------------------------------------------------------------
  # Used when type: "basic"
  # 
  # Sends credentials in Authorization header: "Basic base64(username:password)"
  # 
  # Security Best Practices:
  #   - Use environment variables for credentials: ${env:USERNAME}
  # ----------------------------------------------------------------------------
  # basic:
  #   # Username for authentication
  #   username: "${env:EXPORTER_USERNAME}"
  #   
  #   # Password for authentication
  #   password: "${env:EXPORTER_PASSWORD}"
  
  # ----------------------------------------------------------------------------
  # BEARER TOKEN AUTHENTICATION
  # ----------------------------------------------------------------------------
  # Used when type: "bearer"
  # 
  # Sends token in Authorization header: "Bearer <token>"
  # 
  # Two options for providing the token:
  #   1. Directly in configuration (less secure, use env vars)
  #   2. Read from a file (more secure, e.g., Kubernetes secrets)
  # 
  # ----------------------------------------------------------------------------
  # bearer:
  #   # Option 1: Token directly (use env var)
  #   token: "${env:EXPORTER_TOKEN}"
  #   
  #   # Option 2: Read token from file (preferred for Kubernetes)
  #   # tokenFile: "/var/run/secrets/exporter/token"


# ==============================================================================
# METRIC CONFIGURATION
# ==============================================================================
# Controls how metrics are collected, filtered, transformed, and mapped
# ==============================================================================
metricConfig:
  
  # ----------------------------------------------------------------------------
  # metricPathPrefix (string, optional)
  # ----------------------------------------------------------------------------
  # Prefix for all metrics from this exporter in the AppDynamics metric tree
  # 
  # Default: Inherits from global.defaultMetricPathPrefix
  # 
  # Examples:
  #   - "Custom Metrics|Infrastructure|Node"
  #   - "Hardware Resources|GPU"
  #   - "Application|MyApp|Metrics"
  # 
  # Purpose: Organizes metrics in AppDynamics metric browser
  # 
  # Note: Individual metric mappings can override this with absolute paths
  # ----------------------------------------------------------------------------
  metricPathPrefix: "Custom Metrics|Prometheus"
  
  # ----------------------------------------------------------------------------
  # includeUnmappedMetrics (boolean, optional)
  # ----------------------------------------------------------------------------
  # Whether to include metrics without explicit mappings
  # 
  # Default: Inherits from global.includeUnmappedMetrics (usually false)
  # 
  # When true: All metrics are reported, mapped and unmapped
  #            Unmapped metrics will appear under: metricPathPrefix|exporterName|metricName
  # When false: Only metrics with explicit mappings are reported
  # 
  # Use Cases:
  #   - Set to true temporarily to discover available metrics
  #   - Enable for exporters with stable, well-known metrics
  #   - Disable to prevent metric explosion from dynamic labels
  # ----------------------------------------------------------------------------
  includeUnmappedMetrics: false
  
  # ============================================================================
  # METRIC FILTERS
  # ============================================================================
  # Include/exclude metrics based on regex patterns
  # 
  # Filter Order:
  #   1. Include patterns are evaluated first (if defined)
  #   2. Exclude patterns are evaluated second (if defined)
  #   3. A metric must match include AND not match exclude to be collected
  # 
  # Regex Syntax:
  #   - Uses Java regex syntax
  #   - Patterns are matched against the full metric name
  #   - Use .* for wildcards
  #   - Use ^ and $ for exact matches
  # ============================================================================
  filters:
    
    # --------------------------------------------------------------------------
    # include (array of strings, optional)
    # --------------------------------------------------------------------------
    # Regex patterns for metrics to INCLUDE
    # 
    # Behavior:
    #   - If defined, ONLY metrics matching at least one pattern are collected
    #   - If not defined, all metrics pass the include filter
    # 
    # Examples:
    #   - "http_.*" - All HTTP metrics
    #   - "node_cpu_.*" - All CPU metrics
    #   - "myapp_(requests|errors)_total" - Specific counters
    # 
    # --------------------------------------------------------------------------
    include:
      - ".*"  # Match all metrics (effectively no filtering)
      # - "http_requests_.*"
      # - "database_queries_.*"
      # - "cache_.*"
    
    # --------------------------------------------------------------------------
    # exclude (array of strings, optional)
    # --------------------------------------------------------------------------
    # Regex patterns for metrics to EXCLUDE
    # 
    # Behavior:
    #   - Evaluated AFTER include patterns
    #   - Metrics matching any exclude pattern are dropped
    # 
    # Common Exclusions:
    #   - ".*_bucket" - Histogram buckets (high cardinality)
    #   - ".*_created" - Metric creation timestamps
    #   - ".*_debug.*" - Debug/internal metrics
    #   - ".*_test.*" - Test metrics
    # 
    # --------------------------------------------------------------------------
    exclude:
      - ".*_bucket"      # Exclude histogram buckets
      - ".*_created"     # Exclude creation timestamps
      # - ".*_debug_.*"
      # - "go_.*"        # Exclude Go runtime metrics
  
  # ============================================================================
  # PROPERTY MAPPINGS
  # ============================================================================
  # Extract values from metric labels and report them as Machine Agent properties
  # ============================================================================
  propertyMappings:
    
    # --------------------------------------------------------------------------
    # Property Mapping Entry
    # --------------------------------------------------------------------------
    # Each entry extracts a value from metric labels and creates a property
    # --------------------------------------------------------------------------
    
    # Example 1: Count unique values (e.g., number of CPUs, GPUs, nodes)
    - propertyName: "Example|Total Instances"
      sourceLabelName: "instance"
      extractionType: "COUNT_UNIQUE_VALUES"
      description: "Total number of unique instances"
    
    # Example 2: Get first value (e.g., version, hostname)
    # - propertyName: "Example|Version"
    #   sourceLabelName: "version"
    #   extractionType: "FIRST_VALUE"
    #   description: "Application version"
    
    # Example 3: Concatenate all values (e.g., list of hosts)
    # - propertyName: "Example|All Instances"
    #   sourceLabelName: "instance"
    #   extractionType: "CONCATENATE_VALUES"
    #   description: "Comma-separated list of all instances"
    
    # Example 4: Pattern transformation (e.g., formatted strings)
    # - propertyName: "Hardware|GPU Model"
    #   sourceLabelName: "gpu_model"
    #   extractionType: "PATTERN_TRANSFORM"
    #   valuePattern: "GPU Model: {value}"
    #   description: "Formatted GPU model string"
    
    # Example 5: Filtered extraction (e.g., only specific label values)
    # - propertyName: "Application|Production Nodes"
    #   sourceLabelName: "hostname"
    #   extractionType: "CONCATENATE_VALUES"
    #   labelValueFilter: "prod-.*"
    #   description: "List of production hostnames only"
    
    # --------------------------------------------------------------------------
    # propertyName (string, REQUIRED)
    # --------------------------------------------------------------------------
    # Name of the property in Machine Agent
    # Use "|" as a separator for hierarchical names
    # 
    # Examples:
    #   - "Hardware|GPU|Count"
    #   - "Application|Database|Version"
    #   - "System|Hostname"
    # 
    # Best Practices:
    #   - Use clear, descriptive names
    #   - Organize hierarchically (Category|Subcategory|Property)
    #   - Avoid special characters except "|"
    # --------------------------------------------------------------------------
    
    # --------------------------------------------------------------------------
    # sourceLabelName (string, REQUIRED)
    # --------------------------------------------------------------------------
    # Name of the Prometheus label to extract the value from
    # 
    # Examples:
    #   - "instance" - Instance/host identifier
    #   - "version" - Application/library version
    #   - "cpu" - CPU core identifier
    #   - "gpu" - GPU device identifier
    #   - "model" - Hardware model name
    # 
    # Note: Label must exist in at least one metric for extraction to occur
    # --------------------------------------------------------------------------
    
    # --------------------------------------------------------------------------
    # extractionType (string, REQUIRED)
    # --------------------------------------------------------------------------
    # How to extract/aggregate values from the label across all metrics
    # 
    # Options:
    # 
    #   1. DIRECT_VALUE
    #      Extract the label value directly without transformation
    #      Use for: Static values, identifiers, names
    #      Example: Label "hostname"="server-01" → Property "server-01"
    # 
    #   2. FIRST_VALUE
    #      Use the first encountered value for this label
    #      Use for: Properties that should be consistent (version, hostname)
    #      Example: First metric has "version"="1.2.3" → Property "1.2.3"
    # 
    #   3. COUNT_UNIQUE_VALUES
    #      Count distinct values for this label across all metrics
    #      Use for: Counting resources (CPUs, GPUs, nodes, instances)
    #      Example: Label "gpu" has values "0","1","2" → Property "3"
    # 
    #   4. COUNT_OCCURRENCES
    #      Count total number of metrics that have this label
    #      Use for: Total metric count, activity indicators
    #      Example: 15 metrics have label "gpu" → Property "15"
    # 
    #   5. CONCATENATE_VALUES
    #      Join all unique values with commas (sorted alphabetically)
    #      Use for: Lists of resources, inventories
    #      Example: Label "cpu" has values "0","1","2" → Property "0,1,2"
    # 
    #   6. PATTERN_TRANSFORM
    #      Transform value using a pattern with {value} placeholder
    #      Use for: Formatted strings, prefixed/suffixed values
    #      Example: Pattern "GPU Count: {value}" + value "3" → "GPU Count: 3"
    #      Requires: valuePattern field to be specified
    # 
    # --------------------------------------------------------------------------
    
    # --------------------------------------------------------------------------
    # valuePattern (string, optional)
    # --------------------------------------------------------------------------
    # Pattern for transforming label values (only for PATTERN_TRANSFORM type)
    # 
    # Use {value} as placeholder for the actual label value
    # 
    # Examples:
    #   - "GPU Model: {value}"          → "GPU Model: Tesla V100"
    #   - "Version {value}"              → "Version 1.2.3"
    #   - "Host: {value}"                → "Host: server-01"
    #   - "{value} GPUs Available"       → "4 GPUs Available"
    # 
    # Note: Only used when extractionType is PATTERN_TRANSFORM
    # --------------------------------------------------------------------------
    
    # --------------------------------------------------------------------------
    # labelValueFilter (string, optional)
    # --------------------------------------------------------------------------
    # Regex pattern to filter which label values to include
    # 
    # Use Cases:
    #   - Extract properties only from specific instances/nodes
    #   - Filter by environment (prod, staging, dev)
    #   - Include only specific device types
    # 
    # Examples:
    #   - "prod-.*"              - Only production nodes
    #   - "gpu[0-3]"             - Only GPUs 0-3
    #   - "^worker.*"            - Only worker nodes
    #   - ".*-primary$"          - Only primary instances
    # 
    # Behavior:
    #   - If specified, only metrics with label values matching the pattern
    #     will be included in property extraction
    #   - Uses Java regex syntax
    #   - If not specified, all label values are included
    # 
    # Example Usage:
    #   propertyName: "Infrastructure|Production Nodes"
    #   sourceLabelName: "hostname"
    #   extractionType: "COUNT_UNIQUE_VALUES"
    #   labelValueFilter: "prod-.*"
    #   # Result: Counts only hostnames starting with "prod-"
    # --------------------------------------------------------------------------
    
    # --------------------------------------------------------------------------
    # description (string, optional)
    # --------------------------------------------------------------------------
    # Human-readable description of the property
    # Used for documentation purposes only
    # --------------------------------------------------------------------------
    
  
  # ============================================================================
  # METRIC MAPPINGS
  # ============================================================================
  # Transform Prometheus metrics into AppDynamics metrics
  # 
  # Purpose:
  #   - Map Prometheus metric names to AppDynamics metric paths
  #   - Transform metric values (unit conversion, scaling)
  #   - Transform labels into metric path segments
  #   - Set aggregation type for AppDynamics
  #   - Filter metrics based on label values
  # ============================================================================
  mappings:
    
    # --------------------------------------------------------------------------
    # Metric Mapping Entry
    # --------------------------------------------------------------------------
    # Each entry defines how one Prometheus metric maps to AppDynamics
    # --------------------------------------------------------------------------
    
    # Example 1: Simple metric with label transformation
    - sourceMetricName: "http_requests_total"
      targetMetricPath: "Application|HTTP|{method}|{status}|Requests"
      aggregationType: "SUM"
      labelMappings:
        method: "{value}"    # Keep as-is: GET, POST, etc.
        status: "status_{value}"  # Transform: 200 → status_200
    
    # Example 2: Metric with unit conversion (bytes to MB)
    # - sourceMetricName: "memory_bytes"
    #   targetMetricPath: "System|Memory|Used (MB)"
    #   aggregationType: "OBSERVATION"
    #   multiplier: 0.000001  # bytes to MB
    
    # Example 3: Metric with conditional filtering
    # - sourceMetricName: "node_cpu_seconds_total"
    #   targetMetricPath: "System|CPU|{cpu}|{mode}|Seconds"
    #   aggregationType: "AVERAGE"
    #   labelMappings:
    #     cpu: "cpu{value}"
    #     mode: "{value}"
    
    # --------------------------------------------------------------------------
    # sourceMetricName (string, REQUIRED)
    # --------------------------------------------------------------------------
    # Exact name of the Prometheus metric to map
    # 
    # Examples:
    #   - "http_requests_total"
    #   - "node_memory_MemAvailable_bytes"
    #   - "DCGM_FI_DEV_GPU_UTIL"
    # 
    # Note: Must match exactly (case-sensitive)
    # --------------------------------------------------------------------------
    
    # --------------------------------------------------------------------------
    # targetMetricPath (string, REQUIRED)
    # --------------------------------------------------------------------------
    # Metric path in AppDynamics metric tree
    # 
    # Supports placeholders: {labelName}
    # Placeholders are replaced with label values from the metric
    # 
    # Examples:
    #   - "Application|HTTP|Requests" - Static path
    #   - "Application|HTTP|{method}|Requests" - Dynamic with method label
    #   - "System|CPU|{cpu}|{mode}|Utilization" - Multiple placeholders
    # 
    # Best Practices:
    #   - Use clear, hierarchical paths
    #   - Place most specific values at the end
    #   - Include units in path: "Memory (MB)", "Latency (ms)"
    # --------------------------------------------------------------------------
    
    # --------------------------------------------------------------------------
    # aggregationType (string, REQUIRED)
    # --------------------------------------------------------------------------
    # How AppDynamics should aggregate this metric over time
    # 
    # Options:
    #   - OBSERVATION: Point-in-time measurement (gauges)
    #       Use for: CPU%, memory usage, queue size, temperature
    #   
    #   - AVERAGE: Average over time window
    #       Use for: Response times, latencies, utilization percentages
    #   
    #   - SUM: Sum over time window (counters)
    #       Use for: Request counts, error counts, bytes transferred
    # 
    # Prometheus Metric Type → Aggregation Type mapping:
    #   - Gauge → OBSERVATION or AVERAGE
    #   - Counter → SUM (use rate() or increase() on Prometheus side first)
    #   - Histogram → AVERAGE for quantiles, SUM for counts/sums
    #   - Summary → AVERAGE for quantiles, SUM for counts/sums
    # --------------------------------------------------------------------------
    
    # --------------------------------------------------------------------------
    # multiplier (number, optional)
    # --------------------------------------------------------------------------
    # Multiply metric value by this factor
    # 
    # Use Cases:
    #   - Unit conversion:
    #       - Bytes to MB: 0.000001
    #       - Bytes to GB: 0.000000001
    #       - Seconds to milliseconds: 1000
    #       - Decimal to percentage: 100
    #   - Scaling:
    #       - Convert from basis points: 0.01
    # 
    # Examples:
    #   multiplier: 0.000001  # bytes → MB
    #   multiplier: 100       # decimal → percentage
    #   multiplier: 1000      # seconds → milliseconds
    # --------------------------------------------------------------------------
    
    # --------------------------------------------------------------------------
    # unit (string, optional)
    # --------------------------------------------------------------------------
    # Unit of measurement for documentation
    # 
    # Examples: "bytes", "MB", "ms", "%", "count", "ops/sec"
    # 
    # Note: This is informational only and doesn't affect metric values
    # Consider including units in targetMetricPath instead
    # --------------------------------------------------------------------------
    
    # --------------------------------------------------------------------------
    # labelMappings (map, optional)
    # --------------------------------------------------------------------------
    # Transform label values before inserting into metric path
    # 
    # Format: labelName: "pattern"
    # 
    # Two transformation types:
    #   1. Pattern with {value}: Replace {value} with original label value
    #      Example: "cpu{value}" transforms "0" → "cpu0"
    #   
    #   2. Static string: Replace label value with static string
    #      Example: "primary" transforms any value → "primary"
    # 
    # Examples:
    #   labelMappings:
    #     cpu: "cpu{value}"        # 0 → cpu0, 1 → cpu1
    #     mode: "{value}"          # keep as-is
    #     status: "HTTP_{value}"   # 200 → HTTP_200
    # --------------------------------------------------------------------------
    
  
  # ============================================================================
  # AGGREGATE METRICS
  # ============================================================================
  # Calculate aggregate statistics across multiple individual metrics
  # 
  # Purpose:
  #   - Report cluster-wide or system-wide statistics
  #   - Reduce metric cardinality for high-level views
  #   - Provide summary metrics for dashboards
  # 
  # Timing: Calculated after individual metrics are collected
  # 
  # Use Cases:
  #   - Average CPU utilization across all cores
  #   - Total memory used across all nodes
  #   - Maximum temperature across all GPUs
  #   - Count of active instances
  # ============================================================================
  aggregates:
    
    # --------------------------------------------------------------------------
    # Aggregate Metric Entry
    # --------------------------------------------------------------------------
    # Each entry defines one aggregate metric calculation
    # --------------------------------------------------------------------------
    
    # Example 1: Average across a label (e.g., average CPU usage across cores)
    - sourceMetricName: "cpu_usage_percent"
      targetMetricPath: "System|CPU|Average Usage (%)"
      aggregationFunction: "AVERAGE"
      aggregateAcrossLabel: "cpu"
    
    # Example 2: Maximum value (e.g., hottest temperature)
    # - sourceMetricName: "temperature_celsius"
    #   targetMetricPath: "Hardware|Maximum Temperature (C)"
    #   aggregationFunction: "MAX"
    #   aggregateAcrossLabel: "sensor"
    
    # Example 3: Sum (e.g., total memory across nodes)
    # - sourceMetricName: "memory_used_bytes"
    #   targetMetricPath: "System|Total Memory Used (MB)"
    #   aggregationFunction: "SUM"
    #   aggregateAcrossLabel: "node"
    #   multiplier: 0.000001  # Convert to MB
    
    # Example 4: Count (e.g., number of instances)
    # - sourceMetricName: "up"
    #   targetMetricPath: "System|Instance Count"
    #   aggregationFunction: "COUNT"
    #   aggregateAcrossLabel: "instance"
    
    # Example 5: Filtered aggregate (e.g., average for specific nodes)
    # - sourceMetricName: "cpu_usage_percent"
    #   targetMetricPath: "System|Worker Nodes|Average CPU (%)"
    #   aggregationFunction: "AVERAGE"
    #   aggregateAcrossLabel: "node"
    #   filterLabels:
    #     role: ["worker"]  # Only worker nodes
    
    # --------------------------------------------------------------------------
    # sourceMetricName (string, REQUIRED)
    # --------------------------------------------------------------------------
    # Prometheus metric name to aggregate
    # 
    # Note: This should match a sourceMetricName in the mappings section,
    # or be a metric that was collected
    # --------------------------------------------------------------------------
    
    # --------------------------------------------------------------------------
    # targetMetricPath (string, REQUIRED)
    # --------------------------------------------------------------------------
    # Metric path for the aggregate metric in AppDynamics
    # 
    # Best Practices:
    #   - Use descriptive names: "Average", "Total", "Maximum"
    #   - Include units: "(MB)", "(%)", "(ms)"
    #   - Place aggregates in logical hierarchies
    # --------------------------------------------------------------------------
    
    # --------------------------------------------------------------------------
    # aggregationFunction (string, REQUIRED)
    # --------------------------------------------------------------------------
    # Statistical function to apply
    # 
    # Options:
    #   - AVERAGE: Mean value across all instances
    #   - SUM: Total sum across all instances
    #   - MIN: Minimum value across all instances
    #   - MAX: Maximum value across all instances
    #   - COUNT: Number of instances (count of unique label values)
    # 
    # Use Cases:
    #   - AVERAGE: CPU utilization, memory usage, response time
    #   - SUM: Total requests, total bytes, total errors
    #   - MAX: Peak temperature, max latency, highest queue depth
    #   - MIN: Minimum available memory, lowest utilization
    #   - COUNT: Number of nodes, pods, instances
    # --------------------------------------------------------------------------
    
    # --------------------------------------------------------------------------
    # aggregateAcrossLabel (string, REQUIRED)
    # --------------------------------------------------------------------------
    # Label name to aggregate across
    # 
    # The aggregation will group all metrics with different values of this
    # label and apply the aggregation function
    # 
    # Examples:
    #   - "cpu" - Aggregate across all CPUs (cpu0, cpu1, cpu2, ...)
    #   - "node" - Aggregate across all nodes
    #   - "pod" - Aggregate across all pods
    #   - "instance" - Aggregate across all instances
    # --------------------------------------------------------------------------
    
    # --------------------------------------------------------------------------
    # multiplier (number, optional)
    # --------------------------------------------------------------------------
    # Apply unit conversion to aggregated value
    # 
    # Same as multiplier in regular mappings
    # Applied AFTER aggregation
    # 
    # Example: Sum memory in bytes, then convert to MB with multiplier: 0.000001
    # --------------------------------------------------------------------------
    
    # --------------------------------------------------------------------------
    # filterLabels (map, optional)
    # --------------------------------------------------------------------------
    # Filter which metrics to include in aggregation based on label values
    # 
    # Format: labelName: [value1, value2, ...]
    # 
    # Use Cases:
    #   - Aggregate only specific subsets (e.g., worker nodes only)
    #   - Create multiple aggregates for different groups
    # 
    # Example:
    #   filterLabels:
    #     role: ["worker", "compute"]
    #     environment: ["production"]
    # --------------------------------------------------------------------------


# ==================================================================================
# USAGE EXAMPLES
# ==================================================================================
#
# Example 1: Simple HTTP metrics exporter
# -----------------------------------------
# name: "app-metrics"
# type: "application"
# enabled: true
# serviceDiscovery:
#   type: "static"
#   host: "localhost"
#   port: 8080
#   path: "/metrics"
# metricConfig:
#   metricPathPrefix: "Application|MyApp"
#   filters:
#     include:
#       - "http_.*"
#   mappings:
#     - sourceMetricName: "http_requests_total"
#       targetMetricPath: "Application|MyApp|HTTP|{method}|Requests"
#       aggregationType: "SUM"
#
# Example 2: Authenticated database exporter
# -------------------------------------------
# name: "postgres-metrics"
# type: "database"
# serviceDiscovery:
#   type: "static"
#   host: "postgres.prod.example.com"
#   port: 9187
# authentication:
#   type: "basic"
#   basic:
#     username: "${env:POSTGRES_EXPORTER_USER}"
#     password: "${env:POSTGRES_EXPORTER_PASSWORD}"
# metricConfig:
#   metricPathPrefix: "Database|PostgreSQL"
#   filters:
#     include:
#       - "pg_stat_.*"
#
# ==================================================================================
注: これらの YAML ファイルは参照用であり、このまま使用しないでください。要件に基づいて特定の値を編集する必要があります。

以下は、参照 YAML を編集する例です。

CODE 
prometheus:
  global:
    scrapeInterval: 60
    scrapeTimeout: 10
    connectTimeout: 5
    maxMetricsPerScrape: 10000
    defaultMetricPathPrefix: "Custom Metrics|Prometheus"
    includeUnmappedMetrics: false
    userAgent: "AppDynamics-MachineAgent-Prometheus/1.0"
    threading:
      poolSize: 4
      threadNamePrefix: "appd-prom-scraper"
      daemonThreads: true
  exporters:
    - configFile: "exporters/node-exporter.yaml"
    - configFile: "exporters/dcgm-exporter.yaml"
    # Add additional exporters as needed

この例では：

  • scrapeInterval：メトリックを取得する頻度（秒）を定義します。

  • scrapeTimeout：スクレイプ応答に待機する最大時間（秒）。

  • connectTimeout:：エクスポータへの HTTP 接続を確立するまでの最大時間（秒）。

  • maxMetricsPerScrape：スクレイプごとに収集するメトリックの最大数。
  • defaultMetricPathPrefix：AppDynamics コントローラのメトリックのデフォルトフォルダパス。
  • includeUnmappedMetrics：明示的なマッピングなしでメトリックを収集するかどうか（true または false）。
  • userAgent：スクレイプリクエストに使用される HTTP ユーザーエージェントのヘッダー値。
  • threading.poolSize：同時スクレイピングのスレッド数。
  • threading.threadNamePrefix：Prometheus スクレーパースレッド名のプレフィックス。
  • threading.daemonThreads：スクレーパースレッドがデーモンスレッドかどうか。
  • exporters.configFile：個々のエクスポータの YAML 設定へのパス。

以下は、エクスポータの例です。

JSON 
name: "node-exporter"
type: "node"
enabled: true
serviceDiscovery:
  type: "static"
  host: "localhost"
  port: 9100
  path: "/metrics"
metricConfig:
  metricPathPrefix: "Custom Metrics|Infrastructure|Node"
  filters:
    include:
      - "node_cpu_seconds_total"
      - "node_memory_MemAvailable_bytes"
    exclude:
      - ".*_debug_.*"
  mappings:
    - sourceMetricName: "node_cpu_seconds_total"
      targetMetricPath: "Infrastructure|Node|CPU|{cpu}|{mode}|Seconds"
      aggregationType: "SUM"
      labelMappings:
        cpu: "cpu{value}"
        mode: "{value}"

この例では：

  • name：エクスポータの固有識別子。

  • type：エクスポータのカテゴリ/タイプ。

  • enabled：このエクスポータを有効化/無効化します。

  • serviceDiscovery.type：サービスの発見メカニズム。

  • serviceDiscovery.host：エクスポータのホスト名または IP アドレス。

  • serviceDiscovery.port：エクスポータの TCP ポート。

  • serviceDiscovery.path：メトリックエンドポイントへの HTTP パス。

  • metricConfig.metricPathPrefix：AppDynamics のこのエクスポータからのすべてのメトリックのルートパス。

  • metricConfig.filters.include：これらの正規表現パターンに一致するメトリックのみを収集します。

  • metricConfig.filters.exclude：これらの正規表現パターンに一致するメトリックを除外します。

  • mappings[].sourceMetricName：収集する Prometheus メトリック名。

  • mappings[].targetMetricPath：AppDynamics のメトリックパス（ラベルのプレースホルダをサポート）。

  • mappings[].aggregationType：集約メソッド（SUM、OBSERVATION など）。

  • mappings[].labelMappings：メトリックパスのラベル値のフォーマット。