Unable to connect to the server: dial tcp: i/o timeout (Kubernetes) — how to fix
Unable to connect to the server: dial tcp 10.0.12.34:6443: i/o timeout
E1110 09:12:03.123456 4321 memcache.go:265] couldn't get current server API group list: Get "https://10.0.12.34:6443/api?timeout=32s": dial tcp 10.0.12.34:6443: i/o timeout
Quick fix: Check kubectl config current-context and the server URL (kubectl config view --minify). Test the port with nc -vz <host> 6443. Connect the VPN or bastion, add your IP to the cluster's authorized networks, start the local cluster (minikube start), or refresh the kubeconfig (aws eks update-kubeconfig ... / gcloud container clusters get-credentials ...).
What it means
i/o timeout is a network-level failure: kubectl's TCP connection attempt got no response (packets silently dropped), unlike connection refused where the host actively rejects it. Authentication never happened.
Start by checking what kubectl is trying to reach:
kubectl config current-context
kubectl config view --minify -o jsonpath='{.clusters[0].cluster.server}'; echo
nc -vz -w 5 10.0.12.34 6443 # or: curl -k https://10.0.12.34:6443/version
If nc also times out, it's networking, not kubectl.
Cause 1: wrong context or stale kubeconfig
You might be pointing at an old or deleted cluster, or a cluster whose IP changed after recreation.
kubectl config get-contexts
kubectl config use-context the-right-one
Regenerate credentials for managed clusters:
aws eks update-kubeconfig --region eu-west-1 --name my-cluster
gcloud container clusters get-credentials my-cluster --region europe-west1
az aks get-credentials --resource-group rg --name my-cluster
Also check KUBECONFIG isn't set to a different file: echo $KUBECONFIG.
Cause 2: private API endpoint / VPN
A private IP like 10.x, 172.16-31.x or 192.168.x is only reachable from inside the cluster's network. Connect the VPN, use a bastion host, or tunnel:
ssh -L 6443:10.0.12.34:6443 user@bastion
# then point the kubeconfig server at https://127.0.0.1:6443 (TLS name must match, or use tls-server-name)
- EKS: if only the private endpoint is enabled, you need network access to the VPC. Check with
aws eks describe-cluster --name my-cluster --query cluster.resourcesVpcConfig. - GKE: private clusters need you to be in the VPC or on the authorized networks.
- AKS: private clusters resolve through a private DNS zone; use
az aks command invokeas a fallback.
Cause 3: your IP isn't allowed
Public endpoints often restrict source IPs. If your IP changed (new Wi-Fi, home vs office), you'll time out:
curl -s https://ifconfig.me
# EKS
aws eks update-cluster-config --name my-cluster --resources-vpc-config publicAccessCidrs="203.0.113.7/32"
# GKE
gcloud container clusters update my-cluster --enable-master-authorized-networks --master-authorized-networks 203.0.113.7/32
# AKS
az aks update -g rg -n my-cluster --api-server-authorized-ip-ranges 203.0.113.7/32
For self-managed clusters, check cloud security groups and host firewalls allow TCP 6443 from your machine.
Cause 4: the cluster or control plane is down
- minikube:
minikube status, thenminikube start. After a restart the IP may change;minikube update-contextfixes kubeconfig. - kind / k3d / Docker Desktop: make sure Docker is running and the cluster container is up (
docker ps). - kubeadm: on the control-plane node, check
sudo systemctl status kubeletandsudo crictl ps | grep kube-apiserver. An expired certificate or full disk can stop the API server. - Stopped cloud VMs or a scaled-to-zero cluster also look exactly like this.
Cause 5: proxy settings
If HTTPS_PROXY is set, kubectl sends traffic through the proxy, which may not reach internal addresses:
env | grep -i proxy
export NO_PROXY=$NO_PROXY,10.0.12.34,.internal.example.com
Or set proxy-url per cluster in kubeconfig instead of a global environment variable.
If it's intermittent
Occasional timeouts on large clusters can be API server overload; check API server metrics and raise the timeout for a single request with kubectl --request-timeout=60s get pods.
Last updated 2026-11-10.