How to fix Flannel Backend Error in Kubernetes

kubectl get pods -n kube-flannel  # Or -n kube-system depending on installation
kubectl get pods -n kube-flannel -o wide  # Shows which nodes have Flannel

# Check if DaemonSet is deployed:
kubectl get daemonset -n kube-flannel

# Describe the Flannel pod:
kubectl describe pod -n kube-flannel -l app=flannel

# Check logs:
kubectl logs -n kube-flannel -l app=flannel --tail=100
kubectl logs -n kube-flannel -l app=flannel --previous  # Previous crashed instance

# Check ConfigMap with network config:
kubectl get cm -n kube-flannel
kubectl describe cm -n kube-flannel kube-flannel-cfg

# View the actual configuration:
kubectl get cm kube-flannel-cfg -n kube-flannel -o jsonpath='{.data.net-conf\.json}' | jq .

# Expected output:
{
  "Network": "10.244.0.0/16",
  "Backend": {
    "Type": "vxlan",
    "VNI": 1,
    "Port": 4789
  }
}

# Get node IPs:
kubectl get nodes -o wide

# From control plane, test connectivity:
for node in $(kubectl get nodes -o jsonpath='{.items[*].status.addresses[?(@.type=="InternalIP")].address}'); do
  ping -c 1 $node && echo "Node $node reachable" || echo "Node $node unreachable"
done

# Test specific backend ports:
kubectl exec -it <pod> -- nc -zv <node-ip> 4789  # VXLAN port

# On each node, verify ports are open:
sudo iptables -L | grep -E "4789|8285|6783"  # Common Flannel ports

# Flannel requires:
# - Port 4789 UDP (VXLAN)
# - Port 8285 UDP (Flannel UDP backend)
# - Port 8472 UDP (Alternative VXLAN)
# - Port 6783 TCP/UDP (flannel control)

# Add iptables rules if missing:
sudo iptables -I INPUT -p udp -m multiport --dports 4789 -j ACCEPT
sudo iptables -I INPUT -p udp -m multiport --dports 8285 -j ACCEPT

# For cloud providers, check security groups:
# AWS: Check security group inbound rules
# Azure: Check NSG rules
# GCP: Check firewall rules

ssh <node-ip>
sudo modprobe vxlan  # Load module if missing
module -l | grep vxlan  # Check if loaded
lsmod | grep vxlan

# Verify VXLAN interface:
ip link show  # Look for "flannel.1" or similar

# If interface is missing:
sudo ip link add flannel.1 type vxlan id 1 port 4789 nolearning

# Check current MTU:
kubectl get cm kube-flannel-cfg -n kube-flannel -o jsonpath='{.data.net-conf\.json}' | jq .Backend.VXLANMtu

# VXLAN adds 50 bytes of overhead:
# If interface MTU is 1500, VXLAN MTU should be 1450
# If interface MTU is 9000 (jumbo frames), VXLAN can be 8950

# View actual MTU:
kubectl exec -it <pod> -- ifconfig eth0  # Check MTU

# Update ConfigMap if needed:
kubectl patch configmap kube-flannel-cfg -n kube-flannel --type merge -p \
  '{"data":{"net-conf.json":"...VXLANMtu: 1450..."}}'

# Restart Flannel to apply changes:
kubectl rollout restart daemonset/kube-flannel -n kube-flannel

# Get Pod CIDR:
kubectl get nodes -o jsonpath='{.items[0].spec.podCIDR}'

# Should be different from node CIDR:
kubectl get nodes -o jsonpath='{.items[*].status.addresses[?(@.type=="InternalIP")].address}'

# Check Flannel network in ConfigMap:
kubectl get cm kube-flannel-cfg -n kube-flannel -o jsonpath='{.data.net-conf\.json}' | jq .Network

# If there's overlap, update cluster network:
# This requires cluster recreation or complex migration
kubectl patch node <node> -p '{"spec":{"podCIDR":"10.200.0.0/24"}}'

# Restart the DaemonSet:
kubectl rollout restart daemonset/kube-flannel -n kube-flannel

# Watch pods restart:
kubectl get pods -n kube-flannel -w

# Monitor logs during restart:
kubectl logs -n kube-flannel -l app=flannel -f

# Verify network is restored:
kubectl run -it --rm test --image=alpine -- sh
# Inside pod:
ping $(kubectl get pods -o jsonpath='{.items[1].status.podIP}')