Here is a basic outline of my configuration that I have made specifically for this case:
I have reset everything on the host and created two new guest machines running off two COW containers derived from one baseline freshly made qcow2 container. Networking is following: Tap interface called "tap2" is used to connect VDE2 switch to it. Than both guest being connected via VDE switch while WAN routing is performed by the host. For simplicity I have called one of the guests "client-system" another one is a "server-system". I can't seem to find any error reports/logs on my host or any of the guests. If you could please advice me on any further debugging I'll try to provide as much data as possible. How did I test connection between my guests? I've simply created a 4 GB file on the "server-system" guest than I would try to perform shhfs, http or NFS file transfer from the "client-system" guest. I have created a dummy file filled up with zeroes: user@server-system$ dd if=/dev/zero of=/home/test_big_file_4GB bs=1M count=4K At this stage the most reliable connection I could make is via sshfs mount. During one of my test I've managed to squeeze 1.6 GB transfer on it... though it still failed on me. It seems that by removing the "vitio_net" module (and reloading it again) I can remedy this issue till the next transfer... Than it fails in the same manner. Once again I do; user@client-system$ modprobe -r vitio_net && modprobe vitio_net Interesting thing is that it fails only on the guest that is doing transfer TO. So the system that is being transferred FROM doesn't loos its connection at all. Please find attached Host and guests Network config files. I don't know if they are of any use to you, but you never know I guess... As I have previously mentioned I don't get any issues if I have e1000 nic model enabled instead of virtio. Is there anything else I can do here to help with this? Thank you.
KVM-Virtio-Bug-669184.tar.bz
Description: application/bzip-compressed-tar
