-
SIMONIN Matthieu authored
`qemu_sink` example has been renamed `qemus` and modified as follows: - `boot_and_log.sh` starts 2 VMs - The script is instantiated in the deployment file and - at start-up each VMs will start: - an UDP server listening on 192.168.0.53:1234 (resp. 192.168.0.54:1235) - an UDP client connecting to 192.168.0.53:1235 (resp. 192.168.0.53:1234) - We use netcat in the VMs (using cloud-init) to achieve the above, and it's usable through a tmux inside each VM (root account) This example validates the initial implementation of Tansiv using Slirp and UDP traffic between two VMs. Lessons learnt: - How VM are identified by the coordinator ? Until now, we use a logical name (sets in the deployment.xml) which must be built by the information in the VSG packet. We use the IPs of the source and destination as logical name. Unfortunately here the traffic originating from both VMs uses 127.0.0.1 as ip source. Out of the box this causes some troubles in the coordinator. I could have used the ip+port for this logical name but I've followed another path: I gave the knowledge of the logical identifier to each qemu process and slirp is using it to set the source of the VSG packet accordingly. This way we ensure that this logical name is unique in the coordinator and refers to the right VM.