If the split-lock detection is enabled by default on the cluster node when the resource is migrated or bought online, the cluster crashes.
By permanently disabling the split_lock_detect, this problem can be fixed:
Log in to the node as the root user, then open a terminal console.
Verify if split_lock_detect is enabled.
cat /proc/cpuinfo | grep split_lock_detect
The command will display split lock detection: enabled.
Open the file /etc/default/grub.
Locate "GRUB_CMDLINE_LINUX_DEFAULT=" and add "split_lock_detect=off" at the end of the line.
"GRUB_CMDLINE_LINUX_DEFAULT=<…> split_lock_detect=off"
Update the bootloader.
update-bootloader -refresh
Reboot the cluster node.
Verify if split_lock_detect is disabled by running any one of the following commands:
journalctl -b |grep split
dmesg -T
The command will display split lock detection: disabled.