Standard Migrations fails on agents if VDDK uploaded post scaling the agents

[geet-pf9]

What steps did you take and what happened?

Standard Migrations fails on agents if VDDK uploaded post scaling the agents -

Error -

2026/09/01 16:05:50 failed to live replicate disks: failed to copy disk Hard disk 1 (DeviceKey=2000): failed to run nbdcopy: nbdkit: error: /home/fedora/vmware-vix-disklib-distrib/lib64/libvixDiskLib.so.9: cannot open shared object file: No such file or directory: exec: already started. Trying to perform cleanup

What did you expect to happen?

Agents should get vddk with sync and should work

Environment

main

vCenter version

NA

Anything else you would like to add?

No response

[spai-p9]

Steps done: Migrations failed on agents with nbdkit: error: /home/fedora/vmware-vix-disklib-distrib/lib64/libvixDiskLib.so.9: cannot open shared object file, even though the master had VDDK 8.0.3 uploaded and the UI reported it as available.

Root cause: a stale bind mount on the master, not the agents.

The sync-daemon pod mounts the child path /home/ubuntu/vmware-vix-disklib-distrib as a DirectoryOrCreate hostPath, so its mount is pinned to that directory's inode. On the master, the directory was rm -rf'd while the pod was running, then recreated by the VDDK re-upload. The pod kept the orphaned (deleted) inode.

Restarted the sync-daemon, the sync started again:
kubectl -n kube-system rollout restart daemonset/sync-daemon

The reference refreshed and sync worked.