FWIW I think this is a real bug where the subsequent reconfigs lose the br0 association on each of the host-side veths. Adding the following after each reconfiguration shows this (cannot restart due to a tmp unmount failure?)

      $client->succeed("systemctl stop container\@webserver");
      $client->succeed("systemctl start container\@webserver");

The comment on the disabled rstp test seems to imply that any restart of br0 will cause the container to lose connectivity. This seems wrong to me?

cc @kampfschlaefer

FWIW I think this is a real bug where the subsequent reconfigs lose the br0

I agree. I don't think it's a problem with the test. But still no point running it on Hydra while we know it fails...

@xeji We obviously failed to act on this when it happened, probably because the test has not been in the blocking set. Is there an associated ticket? I don't believe the correct response is to disable it when finally someone has noticed that we actually broke container networking. :)

git bisect reveals that the breaking commit is 07e0c0e

@srhb thank you!