NGINX Ingress Connection Timed Out — Fix Guide (2026)

Getting connection timeout or upstream timed out errors through NGINX Ingress? Here's how to debug and fix timeout issues between NGINX and your backend services.

upstream timed out (110: Connection timed out) through NGINX Ingress means the backend pod isn't responding within the configured timeout window. Here's the systematic fix.

Common Error Messages

# In browser
504 Gateway Timeout

# In NGINX Ingress pod logs
upstream timed out (110: Connection timed out) while reading response header from upstream

# Or
connect() failed (110: Connection timed out) while connecting to upstream

Step 1: Check If the Backend Is Actually Slow

Before tweaking NGINX, verify if your backend is genuinely slow:

bash

# Exec into a pod in the same namespace and test directly
kubectl run curl-test --image=curlimages/curl --restart=Never -- sleep 3600
 
kubectl exec curl-test -- curl -v --max-time 30 \
  http://my-service.my-namespace.svc.cluster.local:8080/api/health
 
# Check pod resource usage
kubectl top pods -n my-namespace
 
# Check if your app pod is throttled
kubectl describe pod my-app-xxx -n my-namespace | grep -A5 "Limits\|Requests"

If the direct curl times out too, the issue is your application — not NGINX.

Fix 1: Increase NGINX Timeout Annotations

The most common fix — NGINX's default timeouts are often too low for slow operations (DB queries, ML inference, file processing):

yaml

apiVersion: networking.k8s.io/v1
kind: Ingress
metadata:
  name: my-app-ingress
  annotations:
    # How long to wait for backend to send response headers (default: 60s)
    nginx.ingress.kubernetes.io/proxy-read-timeout: "300"
    
    # How long to wait to establish connection to backend (default: 60s)
    nginx.ingress.kubernetes.io/proxy-connect-timeout: "30"
    
    # How long to wait for backend to send data between packets (default: 60s)
    nginx.ingress.kubernetes.io/proxy-send-timeout: "300"
    
    # For large file uploads
    nginx.ingress.kubernetes.io/proxy-body-size: "100m"
spec:
  ingressClassName: nginx
  rules:
  - host: myapp.example.com
    http:
      paths:
      - path: /
        pathType: Prefix
        backend:
          service:
            name: my-service
            port:
              number: 8080

Values are in seconds. For ML inference or long-running APIs, go to 300–600 seconds.

Fix 2: Service Port Mismatch

The Ingress points to the right service but the wrong port — connection times out because nothing is listening:

bash

# Verify service exists and ports match
kubectl get svc my-service -n my-namespace
 
# Check what port the pods actually listen on
kubectl get pods -n my-namespace -o yaml | grep containerPort
 
# Test the service directly
kubectl exec curl-test -- curl http://my-service.my-namespace:8080
# vs
kubectl exec curl-test -- curl http://my-service.my-namespace:3000

Fix the Ingress backend port to match what the service actually exposes.

Fix 3: Readiness Probe Failing — NGINX Sends Traffic to Unhealthy Pods

If your readiness probe is configured wrong, pods might be in Ready state but not actually serving traffic:

bash

# Check endpoint status
kubectl get endpoints my-service -n my-namespace
 
# If endpoints show "none" or wrong IPs — readiness probe is failing
kubectl describe pod my-app-xxx -n my-namespace | grep -A10 "Readiness"
kubectl logs my-app-xxx -n my-namespace | tail -20

Fix the readiness probe:

yaml

readinessProbe:
  httpGet:
    path: /health
    port: 8080
  initialDelaySeconds: 15    # give app time to start
  periodSeconds: 10
  failureThreshold: 3
  timeoutSeconds: 5           # probe timeout

Fix 4: keepalive Timeout Mismatch

NGINX reuses connections to backends (keepalive). If your backend closes connections faster than NGINX expects, you get timeouts:

yaml

annotations:
  nginx.ingress.kubernetes.io/upstream-keepalive-timeout: "60"
  nginx.ingress.kubernetes.io/upstream-keepalive-connections: "10"

Or in the NGINX ConfigMap:

yaml

apiVersion: v1
kind: ConfigMap
metadata:
  name: ingress-nginx-controller
  namespace: ingress-nginx
data:
  upstream-keepalive-timeout: "60"
  upstream-keepalive-connections: "10"
  keep-alive: "75"

Fix 5: Slow DNS Resolution

NGINX resolves backend service DNS on first request and caches it. If DNS is slow or stale, the first request times out:

bash

# Check CoreDNS health
kubectl get pods -n kube-system -l k8s-app=kube-dns
kubectl logs -n kube-system -l k8s-app=kube-dns | tail -20
 
# Check NGINX resolver config
kubectl exec -n ingress-nginx <nginx-pod> -- cat /etc/nginx/nginx.conf | grep resolver

Add resolver to NGINX ConfigMap:

yaml

data:
  use-forwarded-headers: "true"
  compute-full-forwarded-for: "true"

Fix 6: NGINX Worker Processes Saturated

If NGINX workers are handling too many concurrent requests, new requests queue and time out:

bash

# Check NGINX pod resource usage
kubectl top pods -n ingress-nginx
 
# Check current connections
kubectl exec -n ingress-nginx <nginx-pod> -- \
  curl localhost:10246/nginx_status

If Active connections is consistently high, increase NGINX resources:

yaml

# In ingress-nginx HelmChart values
controller:
  resources:
    requests:
      cpu: 500m
      memory: 512Mi
    limits:
      cpu: 2000m
      memory: 1Gi
  config:
    worker-processes: "4"      # default is auto (1 per CPU)
    worker-connections: "16384"

Debugging Checklist

bash

# 1. Check NGINX ingress controller logs
kubectl logs -n ingress-nginx -l app.kubernetes.io/name=ingress-nginx --tail=50
 
# 2. Check backend pod logs
kubectl logs -n my-namespace -l app=my-app --tail=50
 
# 3. Check endpoints (are pods registered?)
kubectl get endpoints -n my-namespace my-service
 
# 4. Test direct service access (bypass NGINX)
kubectl run test --image=curlimages/curl --restart=Never -- \
  curl -v --max-time 10 http://my-service.my-namespace:8080
 
# 5. Check current NGINX config
kubectl exec -n ingress-nginx <nginx-pod> -- nginx -T | grep timeout

Quick summary of fixes:

Slow app → increase proxy-read-timeout annotation (most common)
Wrong port → fix service port in Ingress spec
Unhealthy pods → fix readiness probe
keepalive mismatch → set upstream-keepalive-timeout
NGINX overloaded → increase resources and worker connections

NGINX Ingress Connection Timed Out — Fix Guide (2026)

Common Error Messages

Step 1: Check If the Backend Is Actually Slow

Fix 1: Increase NGINX Timeout Annotations

Fix 2: Service Port Mismatch

Fix 3: Readiness Probe Failing — NGINX Sends Traffic to Unhealthy Pods

Fix 4: keepalive Timeout Mismatch

Fix 5: Slow DNS Resolution

Fix 6: NGINX Worker Processes Saturated

Debugging Checklist

Stay ahead of the curve

Related Articles

AWS RDS Connection Timeout from EKS Pods — How to Fix It

Ingress-NGINX Is Being Retired: How to Migrate to Gateway API Before It Breaks

Kubernetes DNS Not Working: How to Fix CoreDNS Failures in Production

Comments