StatefulSet vs Operator — Stateful Workload K8s'de#

"Postgres'i K8s'e nasıl koyarsın? StatefulSet manuel yönet, ya da operator pattern. 2026'da operator standart, manual StatefulSet sadece basic state'li workload için."

Bu rehber StatefulSet'in nerede yeterli, operator'ün ne zaman zorunlu olduğunu, ve karar ağacını anlatır.

⚖️ Tek Cümlede#

Yaklaşım	Felsefe
Plain StatefulSet	"Manuel yönet, K8s sadece pod orchestration"
Operator (CRD-based)	"Domain-specific automation: failover, backup, upgrade"

📐 StatefulSet — Yeterli Olduğu Yerler#

Use case#

Stateless-ish state: Redis cache (data kaybedilebilir)
Manual-managed DB: tek-instance dev Postgres
Custom workload: senin yazdığın stateful app

Manifest#

apiVersion: apps/v1
kind: StatefulSet
metadata:
  name: redis
spec:
  serviceName: redis
  replicas: 3
  selector:
    matchLabels: {app: redis}
  template:
    spec:
      containers:
        - name: redis
          image: redis:7
          volumeMounts:
            - name: data
              mountPath: /data
  volumeClaimTemplates:
    - metadata: {name: data}
      spec:
        accessModes: [ReadWriteOnce]
        resources: {requests: {storage: 10Gi}}

✅ Pro#

K8s native, basit
Tüm K8s tooling uyumlu
Az operasyonel yük

❌ Con#

HA failover yok (manuel)
Backup yok (manuel)
Upgrade orchestration yok
Cluster-wide health management yok

🔑 StatefulSet yeter: Redis cache, NATS, custom worker.

🤖 Operator — Production DB için Zorunlu#

Niye?#

Stateful workload'lar karmaşık lifecycle ister: - Primary failover - Automated backup + restore - Version upgrade (rolling, validation) - Quorum management - Encryption rotation - Cross-region replication

Operator = bu logic'i K8s controller olarak embed eder.

Anatomi#

[CRD: Cluster]                  ← user manifest (declarative)
   │
   ▼
[Operator (controller pod)]     ← reconcile loop
   │
   ├── StatefulSet oluşturur
   ├── PVC yönetir
   ├── Failover otomatik
   ├── Backup tetikler
   └── Upgrade orchestrate

Use case#

Postgres: CloudNativePG, Crunchy, Zalando
MySQL: Vitess, Percona, MySQL Operator
Redis: Redis Operator, Redis Enterprise Operator
Kafka: Strimzi
MongoDB: MongoDB Community Operator
Elasticsearch: ECK (Elastic Cloud on K8s)
etcd: etcd Operator

Detay: Operator-Patterns.md (Postgres odaklı).

🌳 Karar Ağacı#

START
  │
  ├── Production DB (Postgres / MySQL / Mongo)?
  │     │
  │     └── Operator zorunlu (CloudNativePG / Crunchy / vb.)
  │
  ├── Stateful queue (Kafka / NATS)?
  │     │
  │     ├── Kafka → Strimzi
  │     └── NATS → StatefulSet OK (basit setup)
  │
  ├── Cache (Redis / Memcached)?
  │     │
  │     ├── Stateless cache (data kaybı OK) → StatefulSet
  │     └── Persistent / sentinel HA → Redis Operator
  │
  ├── Search (Elasticsearch / OpenSearch)?
  │     │
  │     └── ECK / OpenSearch Operator
  │
  ├── Custom stateful workload (kendi yazdığın)?
  │     │
  │     └── StatefulSet (gerek varsa custom controller yaz)
  │
  └── Dev / lab DB?
        │
        └── StatefulSet basit (operator gereksiz overhead)

📊 Karşılaştırma Tablosu#

Boyut	StatefulSet	Operator
Setup	YAML	Operator install + CRD
Failover	Manuel	Otomatik
Backup	Manuel script	Native (CRD bazlı)
Upgrade	Manuel rolling	Orchestrated
Configuration drift	Manuel detect	Continuous reconcile
Multi-cluster	Manuel	Native (operator-bağlı)
Operasyonel yük	Yüksek	Düşük (operator'a bağlı)
Öğrenme eğrisi	Düşük	Orta (CRD spec)
Vendor lock-in	Yok	Operator-spesifik
2026 prod öneri	Niche	Standart

🛠️ StatefulSet Best Practices (Kullanıyorsan)#

1. PodDisruptionBudget zorunlu#

apiVersion: policy/v1
kind: PodDisruptionBudget
metadata:
  name: redis
spec:
  minAvailable: 2
  selector:
    matchLabels: {app: redis}

2. Pod anti-affinity (farklı node)#

spec:
  template:
    spec:
      affinity:
        podAntiAffinity:
          requiredDuringSchedulingIgnoredDuringExecution:
            - labelSelector:
                matchLabels: {app: redis}
              topologyKey: kubernetes.io/hostname

3. Storage class fast SSD#

spec:
  volumeClaimTemplates:
    - spec:
        storageClassName: <FAST_SSD>

4. Headless service#

apiVersion: v1
kind: Service
metadata:
  name: redis
spec:
  clusterIP: None
  selector: {app: redis}

→ Pod-level DNS: redis-0.redis.<NS>.svc.cluster.local.

5. PreStop hook + graceful shutdown#

lifecycle:
  preStop:
    exec:
      command: ["redis-cli", "SHUTDOWN", "NOSAVE"]

🚫 Anti-Pattern Tablosu#

Anti-pattern	Niye kötü	Doğru
Postgres prod StatefulSet (operator yok)	Manuel failover, backup yok	CloudNativePG / Crunchy
Operator + manuel `kubectl edit`	Drift	GitOps + ArgoCD
StatefulSet + tek replica prod	SPOF	3+ replica
PVC ReadWriteMany cache için	Performance	RWO + per-pod PVC
PDB yok	Drain'da pod kayıp	minAvailable set
Storage class yavaş HDD	DB latency	NVMe / SSD
StatefulSet upgrade `OnDelete` strategy	Manuel pod delete	RollingUpdate
Volume size tek tahmin	Disk full	Generous + alarm + expand
Operator + StatefulSet karışık	Hangi yöneliyor belirsiz	Operator yönetir, StatefulSet read-only

📋 Decision Checklist#

[ ] Workload türü: DB / queue / cache / search / custom?
[ ] Production mu, dev/lab mı?
[ ] HA gereksinim (failover otomatik mi?)?
[ ] Backup ihtiyacı?
[ ] Operator var mı (community / commercial)?
[ ] Operator olgunluğu (CNCF / production-grade)?
[ ] Vendor lock-in kabul edilebilir mi?
[ ] Operasyonel yük kapasitesi var mı?

Karar: StatefulSet vs Operator

📚 Referanslar#

K8s StatefulSet — kubernetes.io/docs/concepts/workloads/controllers/statefulset/
Operator Pattern — kubernetes.io/docs/concepts/extend-kubernetes/operator/
OperatorHub.io — operatorhub.io
Awesome Operators — github.com/operator-framework/awesome-operators
Operator-Patterns.md — Postgres operator'lar
Postgres-Production-Guide.md
HA-Patroni-Stolon.md
Backup-Restore-Patterns.md
Connection-Pooling.md

"StatefulSet K8s'in temel araç'ı; operator uzman aracı. Production DB → operator zorunlu; cache/queue → StatefulSet yeter. Yanlış seçim = 6 ay sonra failover'da öğrenirsin."