Resource Types - antimetal/system-agent GitHub Wiki

Resource Types

⚠️ Work in Progress: This documentation is currently being developed and may be incomplete or subject to change.

Overview

The Antimetal System Agent collects and monitors various types of Kubernetes resources and system metrics. This page provides a comprehensive reference of all resource types, their schemas, and collection behaviors.

Kubernetes Resources

Core Resources

Nodes

  • API Version: v1
  • Kind: Node
  • Collected Fields:
    • Metadata (name, labels, annotations, UID)
    • Status (conditions, capacity, allocatable)
    • Node info (kernel version, OS, container runtime)
    • Addresses (internal IP, hostname)
    • Taints and unschedulable status

Pods

  • API Version: v1
  • Kind: Pod
  • Collected Fields:
    • Metadata (name, namespace, labels, annotations, UID)
    • Spec (containers, volumes, node name)
    • Status (phase, conditions, container statuses)
    • QoS class and priority
    • Resource requests and limits
    • Owner references

Services

  • API Version: v1
  • Kind: Service
  • Collected Fields:
    • Metadata (name, namespace, labels, annotations)
    • Spec (type, ports, selector)
    • Status (load balancer info)
    • Cluster IP and external IPs
    • Session affinity

Endpoints

  • API Version: v1
  • Kind: Endpoints
  • Collected Fields:
    • Metadata (name, namespace)
    • Subsets (addresses, ports)
    • Ready and not-ready addresses

Namespaces

  • API Version: v1
  • Kind: Namespace
  • Collected Fields:
    • Metadata (name, labels, annotations)
    • Status (phase)

Workload Resources

Deployments

  • API Version: apps/v1
  • Kind: Deployment
  • Collected Fields:
    • Metadata (name, namespace, labels, annotations)
    • Spec (replicas, selector, strategy)
    • Status (replicas, conditions)
    • Revision history limit

StatefulSets

  • API Version: apps/v1
  • Kind: StatefulSet
  • Collected Fields:
    • Metadata (name, namespace, labels, annotations)
    • Spec (replicas, selector, service name)
    • Status (replicas, current revision)
    • Update strategy and pod management policy

DaemonSets

  • API Version: apps/v1
  • Kind: DaemonSet
  • Collected Fields:
    • Metadata (name, namespace, labels, annotations)
    • Spec (selector, update strategy)
    • Status (desired/current/ready numbers)
    • Node selector and tolerations

ReplicaSets

  • API Version: apps/v1
  • Kind: ReplicaSet
  • Collected Fields:
    • Metadata (name, namespace, labels, annotations)
    • Spec (replicas, selector)
    • Status (replicas, ready replicas)
    • Owner references (usually Deployment)

Performance Metrics

System Metrics

CPU Metrics

  • Type: cpu
  • Source: /proc/stat
  • Fields:
    • User, system, idle, iowait times
    • Per-core statistics
    • CPU count and frequency
    • Context switches and interrupts

Memory Metrics

  • Type: memory
  • Source: /proc/meminfo
  • Fields:
    • Total, free, available memory
    • Buffers and cache
    • Swap usage
    • Memory pressure indicators

Disk Metrics

  • Type: disk
  • Source: /proc/diskstats
  • Fields:
    • Read/write operations and bytes
    • I/O time and queue statistics
    • Device utilization
    • Per-device metrics

Network Metrics

  • Type: network
  • Source: /proc/net/dev
  • Fields:
    • Bytes/packets sent and received
    • Errors and drops
    • Per-interface statistics
    • Network queue information

Process Metrics

Process Stats

  • Type: process
  • Source: /proc/[pid]/stat
  • Fields:
    • PID, PPID, process name
    • CPU usage (user/system time)
    • Memory usage (RSS, VSZ)
    • State and priority
    • Thread count

Hardware Information

CPU Info

  • Type: cpu_info
  • Source: /proc/cpuinfo, /sys/devices/system/cpu
  • Fields:
    • Processor model and vendor
    • Core count and topology
    • Cache sizes
    • CPU features and flags

Memory Info

  • Type: memory_info
  • Source: /proc/meminfo, /sys/devices/system/node
  • Fields:
    • Total system memory
    • NUMA topology
    • Memory banks and channels

Disk Info

  • Type: disk_info
  • Source: /sys/block
  • Fields:
    • Device model and size
    • Queue configuration
    • Rotational status (SSD/HDD)

Data Formats

Kubernetes Resource Format

All Kubernetes resources follow this general structure:

{
  "apiVersion": "v1",
  "kind": "Pod",
  "metadata": {
    "name": "example-pod",
    "namespace": "default",
    "uid": "12345-67890",
    "labels": {
      "app": "example"
    },
    "annotations": {
      "description": "Example pod"
    }
  },
  "spec": {
    // Resource-specific specification
  },
  "status": {
    // Resource-specific status
  }
}

Performance Metric Format

Performance metrics follow this structure:

{
  "metric_type": "cpu",
  "timestamp": "2024-01-15T10:30:00Z",
  "data": {
    // Metric-specific data fields
  }
}

Collection Behavior

Collection Modes

  1. Continuous Collection

    • Kubernetes resources (watch API)
    • Performance metrics (polling)
    • Real-time updates

  2. One-Shot Collection

    • Hardware information
    • System configuration
    • Collected once at startup

Collection Intervals

Resource Type Default Interval Configurable
Kubernetes Resources Real-time (watch) No
CPU Metrics 10 seconds Yes
Memory Metrics 10 seconds Yes
Disk Metrics 30 seconds Yes
Network Metrics 10 seconds Yes
Process Metrics 30 seconds Yes

Resource Filters

Resources can be filtered by:

  • Namespace: Include/exclude specific namespaces
  • Labels: Select resources with specific labels
  • Annotations: Filter based on annotations
  • Resource names: Include/exclude by name patterns

Example configuration:

filters:
  namespaces:
    include: ["production", "staging"]
    exclude: ["kube-system"]
  labels:
    include:
      monitored: "true"
    exclude:
      temporary: "true"

Resource Relationships

Parent-Child Relationships

  • Deployment --> ReplicaSet --> Pod
  • StatefulSet --> Pod
  • DaemonSet --> Pod
  • Service --> Endpoints
  • Node --> Pod

Cross-References

Resources maintain references through:

  • Owner references
  • Label selectors
  • Service endpoints
  • Node assignments

API Integration

gRPC Message Types

Each resource type has corresponding Protocol Buffer definitions:

message Resource {
  string api_version = 1;
  string kind = 2;
  ObjectMeta metadata = 3;
  google.protobuf.Any spec = 4;
  google.protobuf.Any status = 5;
}

message Metric {
  string type = 1;
  google.protobuf.Timestamp timestamp = 2;
  google.protobuf.Any data = 3;
}

Batch Operations

Resources are sent in batches for efficiency:

  • Maximum batch size: 1000 resources
  • Compression: gzip
  • Chunking for large resources

See Also