Concept lesson

K8s Autoscaling (HPA & KEDA)

Horizontal Pod Autoscaler (HPA) CPU/memory scaling and KEDA event-driven autoscaling (queue depth/PromQL).

lesson
Freshness: current15 min read
Mastery
not started · 0%

Learning outcomes

  • Configure HPA resource metric autoscaling thresholds
  • Scale workloads to zero using KEDA event-driven queue depth triggers

Mental model

K8s Autoscaling (HPA & KEDA) defines a core pattern in modern production engineering, establishing deterministic contracts across distributed nodes or containerized cloud workloads.

Incoming Request / Trigger Event
Validate Protocol Schema & State Invariants
Execute Async Non-Blocking Pipeline
Enforce Resilience & Consensus Guards
Return Verified Execution State
Conceptual teaching model synthesized from:Kubernetes Official Production Systems Architecture & Control Plane Manual

Theory

Understanding k8s autoscaling (hpa & keda) requires analyzing system state machines, fault tolerance boundaries, and communication contracts.

# Production architectural configuration for hpa-keda-autoscaling
apiVersion: v1
kind: ProductionContract
metadata:
  name: hpa-keda-autoscaling-config
spec:
  resiliencePolicy: strict
  maxRetries: 3
  timeoutSeconds: 5

Alternatives and trade-offs

  • Synchronous Tightly-Coupled Architecture: Simple initial setup; vulnerable to cascading failures and thread blocking under heavy traffic.
  • Decoupled Asynchronous Systems (K8s Autoscaling (HPA & KEDA)): High resilience, scalable fault isolation; requires explicit handling of state synchronization and operational complexity.

Failure modes and misconceptions

  1. Unbounded Retries: Retrying failed operations without exponential backoff and jitter causes thundering herd spikes during system recovery.
  2. Missing Fencing Guards: Failing to enforce monotonic fencing tokens allows zombie process writes to overwrite valid state.
Reflect before revealing the guide

Decision scenario

Implement non-blocking execution pipelines, set explicit timeout bounds, and enforce monotonic fencing tokens to achieve high availability and fault isolation.

Learning outcomes

  • Structure production implementations of k8s autoscaling (hpa & keda).
  • Evaluate architectural trade-offs between consistency, availability, and latency.
  • Prevent common failure modes like thundering herd spikes and split-brain state corruption.

Trade-offs

K8s Autoscaling (HPA & KEDA) delivers high operational resilience and scalability, but increases system configuration and telemetry monitoring requirements.

Evidence assessment

Theory and decision mastery

not-started · 0%
theory0%
decision0%
activityNot mapped
projectNot mapped
1. What is the primary architectural goal of K8s Autoscaling HPA KEDA?
2. Which trade-off is introduced when implementing K8s Autoscaling HPA KEDA?
3. What common failure mode occurs when K8s Autoscaling HPA KEDA is misconfigured?

Decision scenario

You are designing a high-concurrency production cloud system that requires reliable deployment of K8s Autoscaling HPA KEDA.

Which decision provides the optimal balance of scalability, fault tolerance, and operational safety?

Primary sources