Closed moralismercatus closed 8 years ago
The problem is that we are using a script, called by vm-node, to execute QEMU instead of executing it directly. The PID obtained and used to kill QEMU is actually that of the script, and not of the QEMU instance.
The solution is to abolish the script mechanism. Instead, if we don't want the executables in $PATH, we can specify them in a configuration file given to vm-node.
Now using a configuration file to specify the path of QEMU. A script is no longer used.
The behavior has only been observed on certain machines with qemu-2.3.
The correct behavior for is for vm-node to signal the VM instance to terminate, wait for the termination to complete, then start the next. The behavior observed is that the previous VM instance either doesn't terminate at all, or does so only after the next started.