Soak testing

EchoFox ships with three M6 scripts for long-running stability validation.

Quick reference

Script	Purpose
scripts/heap-snapshot.js	One-shot v8 heap snapshot to disk
scripts/heap-diff.js	Compare two snapshots, report class-level growth
scripts/soak.js	Synthetic load harness with periodic sampling

In-process leak detector

Always-on runtime watchdog enabled via config.runtime.leakDetection:

runtime: {
  // ...
  leakDetection: {
    enabled:                true,
    sampleIntervalMs:       600000,    // 10 min
    windowSize:             144,        // 24h rolling
    growthThresholdPercent: 30,         // alert at 30% growth
  },
},

Mechanism: every sampleIntervalMs it records the heap-used MB into a rolling window of windowSize samples. If the MIN heap in the window's later half exceeds the MAX heap in the earlier half by growthThresholdPercent, it logs a warn + bumps leak_alerts_total and sets the gauge leak_suspected = 1.

This complements memoryGuard (which catches instantaneous heap pressure for OOM protection); leak detection catches slow growth over hours.

Heap snapshots

Capture a snapshot at any moment:

node scripts/heap-snapshot.js
# → ./snapshots/heap-2026-06-09T14-30-00.heapsnapshot

Or write to a specific path:

node scripts/heap-snapshot.js --out=./before.heapsnapshot

Open the resulting file in Chrome DevTools (chrome://inspect → Memory → Load).

Heap diff

Compare two snapshots to find class-level retention growth:

node scripts/heap-diff.js ./before.heapsnapshot ./after.heapsnapshot

Output:

=== Top 30 class deltas (by abs retained-size change) ===
CLASS                                         +Δ count        +Δ size      pct
-------------------------------------------------------------------------------
object:Promise                                     +423      +12.8 KB    42.1%
string:                                            +156       +8.4 KB    18.2%
...
Total heap delta: 0.34 MB

If the total exceeds 50 MB, you get a warning to review suspect retainers.

Synthetic soak

For a self-contained leak check WITHOUT a real WhatsApp connection:

node scripts/soak.js --durationMin=60 --rate=10 --snapshot=15

Options:

• --durationMin — how long to run (default 60)
• --rate — synthetic messages per minute (default 10)
• --snapshot — take a heap snapshot every N minutes (default 0 = off)
• --out — output directory (default ./soak-out)

The harness uses an in-memory SQLite + mock socket, fires synthetic .ping messages, samples heap + RSS every minute, and writes:

• <out>/report.json — per-minute samples
• <out>/summary.txt — verdict (PASSED / WARN / FAILED)
• <out>/heap-T<N>.heapsnapshot — periodic snapshots (if requested)

Recommended soak profiles

Profile	Command
Smoke (5 min)	node scripts/soak.js --durationMin=5 --rate=20
Short (1 h)	node scripts/soak.js --durationMin=60 --rate=10 --snapshot=15
Medium (8 h)	node scripts/soak.js --durationMin=480 --rate=5 --snapshot=60
Long (24 h)	node scripts/soak.js --durationMin=1440 --rate=2 --snapshot=180

Production soak

For real-world soak on a deployed bot:

1. Set runtime.leakDetection.enabled: true in config.js (default)
2. Configure alert channel: config.alerts.notifyChannel
3. Let it run with normal user traffic for ≥48 hours
4. Watch the dashboard's Metrics tab for the leak_suspected gauge
5. If the leak alert fires, take a heap snapshot, restart, take another, and diff them

Verdict thresholds

Synthetic soak verdict (in scripts/soak.js):

• Heap growth ≤ 20 MB → ✓ PASSED
• Heap growth 20-50 MB → ⚠️ WARN (investigate)
• Heap growth > 50 MB → ❌ FAILED (likely leak)

These thresholds assume a baseline bot with no long-lived caches. Tune per your workload.