Description:
When standby cold sync occurs, there may have a chance that
node creation information is missed in standby node. Active node sending
node information to standby node (checkpointing) will lead to a standby amfd crash.
One way to get over this situation is to create node when the node is null.
Reproduction:
1) Start a cluster (e.g. 5 nodes).
2) On Standby controller, add a sleep (e.g. 5 seconds) to main.cc:
3) Reboot standby controller.
4) Use the script scale_opensaf in python/samples directory to add a node (e.g. PL-6) while standby is rebooting.
5) Observe a coredump on standby node.
Diff:
With the patch of #2112, this issue does not occur.