Well, I think I have found the flaw, but it is very strange...
A while ago, I moved the vmdk image on my disk and changed the path to it inside the lab, but only at each node level ("Configure" menu), instead of changing the path at the "Edit -> Preferences -> Qemu -> Qemu Guest" level: my mistake.
However, when I look at the topology.net file, the image path under [QemuDevice] reflects the correct current path.
This should still work, right? But it doesn't and creates a "Zombie process":
Zombie Qemu Process.jpg [ 18.27 KiB | Viewed 3226 times ]
Now, if I change the path under the Qemu guest definition with "Edit -> Preferences -> Qemu -> Qemu Guest", save it, and insert a new Qemu guest on the topology, start it, now the CPU reflects the load, and the node really starts.
It seems that what really matters is the general Qemu guest image path, instead of the specific node configuration...And hovering the mouse above the node says "Running" when it's not the case with the specific node configuration. I think that state probing should be changed to reflect the real CPU usage of that particular node; there is one "qemu-system-x86_64" (64 bits) per running node with the CPU available
. For example, with 2 nodes (the first process below has nothing to do with our discussion):
Qemu Guests Processes.jpg [ 134.72 KiB | Viewed 3228 times ]