Urbackup of LXC containers : stability issues

Trying to deploy urbackup in our infrastructure, all goes well (almost) on standard standalone debian hosts. Now trying to generalize in an LXC containerized environment, and…

What we know before starting :

  • the environment runs 6 LXC containers, each has a local Inet interface bound on private vRack (OVH)
  • the urbackup is installed on a standalone backup host in the same vRack, so we have private network without filtering between.
  • hosts and container run debian version from 9 (server) to 11. LXC container is running debian 10.

How it was done :

  • Installing a client in each container and running it with no snapshot strategy (let try it simple for files backup)
  • No installation on container host system

what we saw :

  • first container starts ok, and gets up in server
  • adding container instances makes the server registering completely unstable. clients stay up a while and finally go back definitively down. Some need discovery explicit request, some other (same setup) never discovered,
  • Some containers discovered after a while with strange recording “apl1-web3-str Yes NaN/NaN/NaN NaN:NaN NaN/NaN/NaN NaN:NaN”. Those seem online while their IP status in discovery request say the opposite…

Seems LXC environment is not immediately workable without some precautions…

Is there some previous experiences to learn from?