Dynamic Thresholding
Skala resource aktif berdasarkan load traffic, queue depth, dan SLA target.
Documentation / Sovereign Autoscale
Sovereign Autoscale
Sistem scaling cerdas untuk memastikan kapasitas selalu sesuai kebutuhan operasional, tanpa pemborosan resource.
Policy Guardrail
Batas minimum-maksimum node dijaga agar stabilitas dan cost control tetap seimbang.
Node Lifecycle
Provision, warmup, handover, dan drain node dikelola tanpa downtime operasional.
Autoscale Lifecycle
- Signal collector membaca telemetry per node secara kontinu.
- Planner menghitung skenario scale up/down berbasis policy.
- Executor menjalankan provisioning dan traffic redistribution.
- Verifier memantau health score pasca scaling dan melakukan remediation jika perlu.
Authentication & Policy Access
Endpoint autoscale hanya dapat diakses oleh role superadmin atau service account dengan scope khusus. Semua operasi scale wajib menyertakan request signature untuk audit trail.
# Required Headers
Authorization: Bearer <service_jwt>
X-Tenant-ID: mitra_abc123
X-Request-ID: uuid-v4
X-Signature: hmac_sha256(payload, secret)
# Required Scopes
autoscale:read — baca telemetry & policy
autoscale:write — update policy & trigger scale
autoscale:override — bypass guardrail (superadmin only)
Autoscale API Endpoints
GET
/api/autoscale/policyAmbil konfigurasi policy autoscale aktif untuk tenant.
{
"tenant_id": "mitra_abc123",
"min_nodes": 2,
"max_nodes": 12,
"target_cpu_pct": 65,
"target_queue_depth": 120,
"scale_up_step": 2,
"scale_down_step": 1,
"cooldown_seconds": 300,
"health_gate_min_score": 85,
"updated_at": "2026-05-19T00:00:00Z"
}
PUT
/api/autoscale/policyUpdate policy autoscale tenant.
{
"min_nodes": 3,
"max_nodes": 20,
"target_cpu_pct": 70,
"target_queue_depth": 180,
"cooldown_seconds": 240,
"health_gate_min_score": 88
}
GET
/api/autoscale/telemetryAmbil snapshot telemetry cluster untuk decisioning dan monitoring.
{
"timestamp": "2026-05-19T00:05:00Z",
"cluster": {
"active_nodes": 6,
"avg_cpu_pct": 58,
"avg_memory_pct": 61,
"queue_depth": 94,
"req_per_sec": 132
},
"nodes": [
{ "id": "node-sg-01", "cpu": 62, "memory": 60, "health": 92 },
{ "id": "node-jkt-01", "cpu": 55, "memory": 59, "health": 95 }
]
}
POST
/api/autoscale/simulateSimulasikan keputusan autoscale tanpa melakukan perubahan nyata (dry-run decision engine).
{
"input": {
"active_nodes": 4,
"avg_cpu_pct": 83,
"queue_depth": 260
}
}
# Response
{ "decision": "scale_up", "add_nodes": 2, "confidence": 0.93 }
POST
/api/autoscale/actionTrigger aksi manual scale up/down dengan audit log.
{
"action": "scale_up" | "scale_down",
"count": 2,
"reason": "traffic_spike" | "cost_optimization" | "manual_override",
"override_guardrail": false
}
Decision Engine Rules
1. Jika avg_cpu_pct > target_cpu_pct selama 3 interval berturut-turut, trigger scale up.
2. Jika queue_depth > target_queue_depth selama 2 interval, trigger scale up prioritas tinggi.
3. Jika avg_cpu_pct < target_cpu_pct - 20% selama 15 menit, evaluasi scale down.
4. Scale down hanya dilakukan jika node health score semua node > 90 dan tidak ada incident aktif.
5. Semua aksi scale diblok selama cooldown window aktif untuk mencegah flapping.
Rate Limiting
| Endpoint Group | Limit | Window | Notes |
|---|---|---|---|
| GET /policy, /telemetry | 120 req | 1 menit | Read-heavy safe |
| PUT /policy | 30 req | 1 menit | Config mutation |
| POST /action, /simulate | 20 req | 1 menit | Scale decision control |
Error Codes
| Code | Meaning | Resolution |
|---|---|---|
| AS-001 | Scale operation blocked by cooldown window | Tunggu cooldown selesai atau override via superadmin |
| AS-002 | Node provisioning failed | Cek quota provider dan network bootstrap script |
| AS-003 | Health gate failed after scale-up | Node baru gagal health check, auto-drain aktif |
| AS-004 | Scale-down prevented by minimum capacity policy | Turunkan min_nodes policy atau tingkatkan drain timeout |
| AS-005 | Telemetry stale, decision engine paused | Periksa collector pipeline dan heartbeat agent |
Observability Metrics
| Metric | Unit | Deskripsi |
|---|---|---|
| as.scale.events_total | count | Total event scale up/down |
| as.scale.up.duration_ms | ms | Durasi rata-rata scale-up |
| as.scale.down.duration_ms | ms | Durasi rata-rata scale-down |
| as.node.health_score | 0-100 | Skor kesehatan node agregat |
| as.node.active_total | count | Jumlah node aktif saat ini |
| as.queue.depth | count | Kedalaman queue request |
| as.policy.block_count | count/hr | Jumlah scale event diblok policy |
| as.cost.estimate_hourly | currency/hr | Estimasi biaya per jam |
Optimization Tip
Aktifkan cooldown window dan anomaly dampener untuk mencegah flapping saat beban traffic berubah cepat.