Handling Process
1. Check the size of the remaining memory of the CNA node where the VM resides.
2. Migrate some VMs that occupy little memory in online mode or disable
some VMs, ensuring that the remaining memory of the CNA node is
sufficient. Then migrate the VMs that occupy much memory in online mode.
Root Cause
1. View the /var/log/libvirt/libvirtd.log file. The failure cause is found that the balloon operation of the VM times out.
Figure 1 linvirtd log
2. View the /var/log/memoc/memoc.log file. Errors are reported all the time for the balloon operation of the VM.
Figure 2 memoc log
3. The total physical memory of the CNA node is 32 GB. Check the memory
of the CNA node. It is found that the memory is insufficient and the VM
fails to obtain all the memory, resulting in the balloon operation
timeout.
Figure 3 Insufficient memory
In the memory reuse scenario, all the memory needs to be obtained before
VM hot migration is started. In this case, the memory of the CNA node
is insufficient, causing the hot migration failure.
Suggestions
1.
Before migrating a VM in online mode in the memory reuse scenario,
ensure that the available memory of the source node is sufficient for
the VM to balloon to all specified memory.
2. Avoid performing hot migration for the VMs that occupy much memory in the memory reuse scenario.