github/k3k
1
0
Fork 0
mirror of https://github.com/rancher/k3k.git synced 2026-08-18 11:56:22 +00:00
Code Issues Packages Projects Releases Wiki Activity

Refactor startup command to wait for node IP changes (#598)

Browse Source 
* Patch node ip when server pod restarts

Signed-off-by: galal-hussein <hussein.galal.ahmed.11@gmail.com>

* wsl

Signed-off-by: galal-hussein <hussein.galal.ahmed.11@gmail.com>

* Refactor startup command and adding safe mode

Signed-off-by: galal-hussein <hussein.galal.ahmed.11@gmail.com>

* Add date/time logging to the startup script

Signed-off-by: galal-hussein <hussein.galal.ahmed.11@gmail.com>

* fixes

Signed-off-by: galal-hussein <hussein.galal.ahmed.11@gmail.com>

---------

Signed-off-by: galal-hussein <hussein.galal.ahmed.11@gmail.com>
This commit is contained in:
Hussein Galal
2026-01-09 16:29:47 +02:00
committed by GitHub
parent c16eae99c7
commit a871917aec
6 changed files with 136 additions and 19 deletions
Download Patch File Download Diff File Expand all files Collapse all files
pkg/controller/cluster/client.go
+1 -1
View File
@@ -23,7 +23,7 @@ func newVirtualClient(ctx context.Context, hostClient ctrlruntimeclient.Client,
}
if err := hostClient.Get(ctx, kubeconfigSecretName, &clusterKubeConfig); err != nil {
return nil, fmt.Errorf("failed to get kubeconfig secret: %w", err)
return nil, err
}
restConfig, err := clientcmd.RESTConfigFromKubeConfig(clusterKubeConfig.Data["kubeconfig.yaml"])
pkg/controller/cluster/server/server.go
+6 -2
View File
@@ -416,9 +416,11 @@ func (s *Server) setupDynamicPersistence() v1.PersistentVolumeClaim {
func (s *Server) setupStartCommand() (string, error) {
var output bytes.Buffer
tmpl := singleServerTemplate
tmpl := StartupCommand
mode := "single"
if *s.cluster.Spec.Servers > 1 {
tmpl = HAServerTemplate
mode = "ha"
}
tmplCmd, err := template.New("").Parse(tmpl)
@@ -430,6 +432,8 @@ func (s *Server) setupStartCommand() (string, error) {
"ETCD_DIR": "/var/lib/rancher/k3s/server/db/etcd",
"INIT_CONFIG": "/opt/rancher/k3s/init/config.yaml",
"SERVER_CONFIG": "/opt/rancher/k3s/server/config.yaml",
"CLUSTER_MODE": mode,
"K3K_MODE": string(s.cluster.Spec.Mode),
"EXTRA_ARGS": strings.Join(s.cluster.Spec.ServerArgs, " "),
}); err != nil {
return "", err
pkg/controller/cluster/server/template.go
+98 -12
View File
@@ -1,16 +1,102 @@
package server
var singleServerTemplate string = `
if [ -d "{{.ETCD_DIR}}" ]; then
# if directory exists then it means its not an initial run
/bin/k3s server --cluster-reset --config {{.INIT_CONFIG}} {{.EXTRA_ARGS}} 2>&1 | tee /var/log/k3s.log
fi
rm -f /var/lib/rancher/k3s/server/db/reset-flag
/bin/k3s server --config {{.INIT_CONFIG}} {{.EXTRA_ARGS}} 2>&1 | tee /var/log/k3s.log`
var StartupCommand string = `
info()
{
echo "[INFO] [$(date +"%c")]" "$@"
}
fatal()
{
echo "[FATAL] [$(date +"%c")] " "$@" >&2
exit 1
}
# safe mode function to reset node IP after pod restarts
safe_mode() {
CURRENT_IP=""
if [ -f /var/lib/rancher/k3s/k3k-node-ip ]; then
CURRENT_IP=$(cat /var/lib/rancher/k3s/k3k-node-ip)
fi
if [ -z "$CURRENT_IP" ] || [ "$CURRENT_IP" = "$POD_IP" ] || [ {{.K3K_MODE}} != "virtual" ]; then
return
fi
# skipping if the node is starting for the first time
if [ -d "{{.ETCD_DIR}}" ]; then
info "Starting K3s in Safe Mode (Network Policy Disabled) to patch Node IP from ${CURRENT_IP} to ${POD_IP}"
/bin/k3s server --disable-network-policy --config $1 {{.EXTRA_ARGS}} > /dev/null 2>&1 &
PID=$!
# Start the loop to wait for the nodeIP to change
info "Waiting for Node IP to update to ${POD_IP}."
count=0
until kubectl get nodes -o wide 2>/dev/null | grep -q "${POD_IP}"; do
if ! kill -0 $PID 2>/dev/null; then
fatal "safe Mode K3s process died unexpectedly!"
fi
sleep 2
count=$((count+1))
if [ $count -gt 60 ]; then
fatal "timed out waiting for node to change IP from $CURRENT_IP to $POD_IP"
fi
done
info "Node IP is set to ${POD_IP} successfully. Stopping Safe Mode process..."
kill $PID
wait $PID 2>/dev/null || true
fi
}
start_single_node() {
info "Starting single node setup..."
# checking for existing data in single server if found we must perform reset
if [ -d "{{.ETCD_DIR}}" ]; then
info "Existing data found in single node setup. Performing cluster-reset to ensure quorum..."
if ! /bin/k3s server --cluster-reset --config {{.INIT_CONFIG}} {{.EXTRA_ARGS}} > /dev/null 2>&1; then
fatal "cluster reset failed!"
fi
info "Cluster reset complete. Removing Reset flag file."
rm -f /var/lib/rancher/k3s/server/db/reset-flag
fi
# entering safe mode to ensure correct NodeIP
safe_mode {{.INIT_CONFIG}}
info "Adding pod IP file."
echo $POD_IP > /var/lib/rancher/k3s/k3k-node-ip
var HAServerTemplate string = `
if [ ${POD_NAME: -1} == 0 ] && [ ! -d "{{.ETCD_DIR}}" ]; then
/bin/k3s server --config {{.INIT_CONFIG}} {{.EXTRA_ARGS}} 2>&1 | tee /var/log/k3s.log
else
/bin/k3s server --config {{.SERVER_CONFIG}} {{.EXTRA_ARGS}} 2>&1 | tee /var/log/k3s.log
fi`
}
start_ha_node() {
info "Starting pod $POD_NAME in HA node setup"
if [ ${POD_NAME: -1} == 0 ] && [ ! -d "{{.ETCD_DIR}}" ]; then
info "Adding pod IP file."
echo $POD_IP > /var/lib/rancher/k3s/k3k-node-ip
/bin/k3s server --config {{.INIT_CONFIG}} {{.EXTRA_ARGS}} 2>&1 | tee /var/log/k3s.log
else
safe_mode {{.SERVER_CONFIG}}
info "Adding pod IP file."
echo $POD_IP > /var/lib/rancher/k3s/k3k-node-ip
/bin/k3s server --config {{.SERVER_CONFIG}} {{.EXTRA_ARGS}} 2>&1 | tee /var/log/k3s.info
fi
}
case "{{.CLUSTER_MODE}}" in
"ha")
start_ha_node
;;
"single"|*)
start_single_node
;;
esac`
pkg/controller/cluster/statefulset.go
+1 -1
View File
@@ -120,7 +120,7 @@ func (p *StatefulSetReconciler) handleServerPod(ctx context.Context, cluster v1b
if pod.DeletionTimestamp.IsZero() {
if controllerutil.AddFinalizer(pod, etcdPodFinalizerName) {
log.V(1).Info("Server Pod is being deleted. Removing finalizer", "pod", pod.Name, "namespace", pod.Namespace)
log.V(1).Info("Server Pod is being created. Adding finalizer", "pod", pod.Name, "namespace", pod.Namespace)
return p.Client.Update(ctx, pod)
}