System Monitor

Reusable monitoring toolkit for Wippy applications. Provides a stats collector and HTTP endpoint that any app can include as a dependency.

Installation

Add to your _index.yaml:

- name: dep.monitor
  kind: ns.dependency
  component: butschster/monitor
  version: "*"

Add local replacement to wippy.lock (for development):

replacements:
  - from: butschster/monitor
    to: ../../monitor/src

Components

Stats Collector (monitor:collector)

Library that collects all system data in a single call:

local collector = require("collector")

local stats = collector.collect_stats()
-- stats.hostname        — string
-- stats.pid             — number
-- stats.cpu_count       — number
-- stats.max_procs       — number
-- stats.goroutines      — number
-- stats.mem_stats       — table (heap_alloc, heap_sys, stack_in_use, num_gc, ...)
-- stats.services        — table[] (id, status, desired, retry_count, started_at)
-- stats.modules         — table[] (name, description, class)
-- stats.total_entries   — number
-- stats.entry_counts    — table {[kind] = count}
-- stats.entry_details   — table {[id] = {kind, meta}}
-- stats.collected_at    — number (unix nanoseconds)

Import in your _index.yaml:

imports:
  collector: monitor:collector

Stats Collector Process (monitor:collector.process)

Long-running process that responds to "collect" messages with a stats snapshot. Registers as "monitor.collector" in the process registry for cross-node discovery.

# Add to your _index.yaml
- name: monitor.service
  kind: process.service
  process: monitor:collector.process
  host: app:processes
  lifecycle:
    auto_start: true

Requesting stats from another process:

-- Look up the collector (works across cluster nodes)
local pid = process.registry.lookup("monitor.collector")
process.send(pid, "collect")

-- Response arrives in inbox as topic "monitor.stats"
local msg = inbox:receive()
local stats = msg:payload():data()

HTTP Stats Endpoint (monitor:stats)

Function that returns system stats as JSON. Mount on any HTTP router:

- name: debug.stats
  kind: http.endpoint
  method: GET
  path: /debug/system
  function: monitor:stats

Response:

{
  "hostname": "myhost",
  "pid": 1234,
  "cpu_count": 4,
  "goroutines": 23,
  "memory": {
    "alloc": 12419072,
    "heap_alloc": 12419072,
    "heap_sys": 25624576,
    "heap_objects": 8421,
    "stack_in_use": 524288,
    "num_gc": 15
  },
  "services": [
    {
      "id": "app:http.server",
      "status": "running",
      "retry_count": 0
    }
  ],
  "total_entries": 12,
  "entry_counts": {
    "function.lua": 5,
    "process.service": 3
  },
  "collected_at": 1707744000000000000
}

Cluster Monitoring

In cluster mode, each node runs a collector.process service. A monitor can discover all nodes and collect stats from each:

┌─── Node A ──────────┐  ┌─── Node B ──────────┐
│ monitor:collector    │  │ monitor:collector    │
│ (process.service)    │  │ (process.service)    │
│ registered as        │  │ registered as        │
│ "monitor.collector"  │  │ "monitor.collector"  │
└──────────┬───────────┘  └──────────┬───────────┘
           │    process.send(pid)    │
           └────────────┬────────────┘
                        │
                ┌───────▼───────┐
                │   Monitor     │
                │   (any node)  │
                └───────────────┘

1. Subscribe to cluster events for node discovery
2. Look up "monitor.collector" on each node via process registry
3. Send "collect" to each — responses arrive via inbox
4. Aggregate and render

Registry Entries

| Entry | Kind | Purpose | |-----------------------------|----------------|-------------------------------------| | monitor:collector | library.lua | Importable stats collection library | | monitor:collector.process | process.lua | Long-running collector process | | monitor:stats | function.lua | HTTP endpoint returning JSON stats |