fix: remove NodeName from DaemonSet pod template

DaemonSet pods should not have NodeName set in their template spec.
The DaemonSet controller manages pod placement based on nodeAffinity
and other scheduling rules. Having NodeName hardcoded was forcing all
DaemonSet pods to only run on one specific node, preventing the
DaemonSet from functioning properly.

This was causing:
- Only one pod running instead of one per matching node
- DaemonSet controller creating hundreds of pods trying to satisfy requirements
- Pods unable to schedule on other matching nodes

The fix removes the NodeName field from the pod template, allowing the
DaemonSet controller to properly manage pod placement across all nodes
that match the configured nodeAffinity.

Author: elijah-rou

pkg/juicefs/mount/builder/daemonset.go

@@ -48,6 +48,8 @@ func (d *DaemonSetBuilder) NewMountDaemonSet(dsName string) (*appsv1.DaemonSet,
 	// Create template pod for DaemonSet
 	pod := podBuilder.genCommonJuicePod(podBuilder.genCommonContainer)
 	pod.Spec.RestartPolicy = corev1.RestartPolicyAlways
+	// Remove NodeName from DaemonSet pod template - DaemonSet controller manages pod placement
+	pod.Spec.NodeName = ""
 
 	// Generate mount command
 	mountCmd := d.genMountCommand()