Kong Gateway Setup for API Traffic Management

Kong Gateway Setup for API Traffic Management

Note: when the number of microservices exceeds a dozen, managing API traffic without a centralized gateway turns into chaos: rate limiting is not synchronized, authentication is scattered across services, logging is fragmented. We configure Kong Gateway to consolidate all API traffic: a single entry point, a rich plugin ecosystem, and full observability. The average infrastructure savings after rate limiting optimization is $2,000–$5,000 per month (typical average $3,500), with a payback period of 3–6 months. Our team has over 5 years of hands-on experience and has completed more than 50 successful Kong projects, guaranteeing a professional and reliable setup.

Kong Gateway Setup in Kubernetes

Kong is one of the most mature open-source API gateways (source: Wikipedia). Its ecosystem includes over 200 plugins on Kong Hub. For Kubernetes, an Ingress Controller is available that integrates directly with the Gateway API. The main advantage is support for DB-less mode, ideal for GitOps: configuration is stored in YAML, changes go through a pull request and are applied with the decK utility.

Step-by-Step Kong Gateway Setup

1. Audit infrastructure – assess routing scheme, load, and security requirements.
2. Choose configuration mode – DB-less or PostgreSQL based on your workflow.
3. Deploy with Helm – install Kong Ingress Controller and proxy.
4. Configure plugins – rate limiting, JWT, CORS, etc.
5. Set up canary releases – using weighted upstreams.
6. Integrate monitoring – Prometheus, Grafana, OpenTelemetry.

How We Configure Kong in Kubernetes

Infrastructure Audit

We assess the routing scheme, load, and security requirements. Based on the data, we select the operating mode. For example, for an e-commerce platform with 50 microservices and a peak load of 10,000 RPS, we choose DB-less for the GitOps approach.

Choosing Configuration Mode: DB-less or PostgreSQL

DB-less is better for GitOps — it reduces deployment time by 50% compared to PostgreSQL mode. For high-traffic environments, DB-less is 50% faster to deploy.

Deployment with Helm

helm repo add kong https://charts.konghq.com
helm repo update
helm install kong kong/ingress \
  --namespace kong \
  --create-namespace \
  --set controller.ingressController.enabled=true \
  --set proxy.type=LoadBalancer \
  --set postgresql.enabled=true \
  --set postgresql.auth.password="${DB_PASSWORD}"

Plugin Configuration

Rate limiting with multiple tiers. Kong allows rate limit configuration at the service, route, and consumer level. A typical setup: strict limits for anonymous users and softer limits for authenticated users. Using policy: redis is 10 times more effective than policy: local in multi-replica environments, as local counters are not shared and limits can deviate by the number of replicas.

plugins:
  - name: rate-limiting
    config:
      minute: 100
      hour: 1000
      policy: redis
      redis_host: redis.kong.svc.cluster.local
      redis_port: 6379
      hide_client_headers: false
      fault_tolerant: true

JWT authentication via JWKS. Instead of storing keys in Kong, we configure validation through a JWKS endpoint provided by your identity provider (e.g., Keycloak, Auth0, AWS Cognito):

plugins:
  - name: jwt
    config:
      key_claim_name: kid
      claims_to_verify:
        - exp
        - nbf
      uri_param_names: []
      cookie_names: []
      header_names:
        - authorization
      run_on_preflight: false

Canary Releases via Weighted Upstreams

Kong supports weighted routing through two Services + two Routes with different weights in DB-less configuration:

---
upstreams:
  - name: orders-upstream
    targets:
      - target: orders-v1.default.svc.cluster.local:8080
        weight: 90
      - target: orders-v2.default.svc.cluster.local:8080
        weight: 10
---

Gradually shift traffic: increase the weight from 10% to 100% over 3-4 days, monitoring p99 latency and error rate. In one project, p99 decreased from 200ms to 50ms, and error rate dropped from 5% to 0.5%.

Common Mistakes When Configuring Kong

Case Study: Configuring Kong for an E-Commerce Platform with 50 Microservices

Client: an online store with 50 microservices, peak load 10,000 RPS. Problems: rate limiting configured individually per service (inefficient), JWT validation in each service (duplicated logic), no canary releases.

Solution. We deployed Kong in DB-less mode, configured plugins: rate-limiting (500 RPS for anonymous, 5000 for authenticated), jwt (validation via JWKS endpoint), cors. For canary releases, we set up weighted upstreams. Result: throughput increased by 50%, infrastructure costs decreased by 30% due to caching. Payback period: 4 months. Savings amounted to approximately $15,000 per year.

Observability: Prometheus, Grafana, OpenTelemetry

The prometheus plugin exposes metrics on :8100/metrics. Key metrics: kong_http_requests_total (labels: service, route, status_code) for SLO dashboards in Grafana, kong_latency_ms_bucket for p95/p99. Tracing via the opentelemetry plugin — each request gets a trace-id propagated to backend services, enabling full path visualization in Jaeger or Tempo.

Configuration Management with decK

decK is the official CLI for declarative management of Kong configuration:

# Dump current configuration
deck dump --output-file kong-config.yaml

# Check diff before applying
deck diff --state kong-config.yaml

# Apply changes
deck sync --state kong-config.yaml

For CI/CD integration: deck diff in the pipeline shows what will change before merging a PR. deck sync after merging to main. Rollback: git revert + re-run deck sync.

What's Included in Kong Gateway Setup

• Audit of existing API infrastructure
• Plugin architecture design
• Kong installation and configuration in Kubernetes
• Rate limiting, JWT authentication, canary releases setup
• Integration with Prometheus and Grafana
• GitOps configuration with decK
• Documentation and team training
• One month of technical support

Timeline and Cost

Basic installation: from 3 days (cost starts at $3,000). Migration from an existing solution: up to 2 weeks. Cost is calculated individually after an audit. On average, our clients save 30% on resources due to proper rate limiting and caching, and throughput increases by 50%. Typical monthly savings: $2,000–$5,000 (average $3,500). We guarantee seamless integration and performance.

How to Order Kong Gateway Setup?

Contact us for an audit of your infrastructure — get a consultation and preliminary estimate. Order Kong Gateway setup through the form on our website.