During online migration of a virtual machine, its
prl_vm_app process on the target node consumes more memory than it is assigned to this VM, and after some time the migration operation is canceled due to out-of-memory (OOM) situation.
For example, migrating a VM with 24 GB RAM, it can be like the following before migration interrupting:
[root@node ~]$ free -m total used free shared buffers cached Mem: 64228 64060 167 0 2 151 -/+ buffers/cache: 63906 321 Swap: 16383 16383 0
[node@node ~]$ free -m total used free shared buffers cached Mem: 64228 4778 59449 0 39 1347 -/+ buffers/cache: 3391 60836 Swap: 16383 413 15970
The following error could be founded in the system log
Sep 23 13:13:18 test.vm.com kernel: [336719.262854] Out of memory: OOM killed process 1374 (prl_vm_app) score 0 vm:215154152kB, rss:62621660kB, swap:16469268kB
This case was recognized as a product issue with the ID #PSBM-29109
The fix will be included in the future updates of the Parallels Cloud Server.