" But when trying to mount with this label, mount hangs up for a while(feel about one minute), then brain-split happened to cluster, next, n1(local) reboot, after a while, cluster go normal status, finally, mount return. " This don't appear again after reboot all nodes on the cluster. So, adding "udevadm trigger" can solve mount by "LABEL=" on remote nodes. Should we close this bug? Or try to find out solution that trigger udev in low level to create corresponding syslink in /dev/disk/by-label.