The bug is that after shutting down an lxc full-distro container, start always fails. It turns out that shutdown doesn't clean the machine on the machined side. A workaround is to machinectl terminate the container. A proper fix would be that machinectl tracks the state of the init process and cleans the machine if that one disappears. This fix is the reasonable fix we can have on the libvirt side: really try to start the container manually if starting it with systemd failed. It currently never calls the virCgroupNewMachineManual function. --- src/util/vircgroup.c | 3 --- 1 file changed, 3 deletions(-) diff --git a/src/util/vircgroup.c b/src/util/vircgroup.c index e99caf5..466c170 100644 --- a/src/util/vircgroup.c +++ b/src/util/vircgroup.c @@ -1656,9 +1656,6 @@ virCgroupNewMachine(const char *name, group)) == 0) return 0; - if (rv == -1) - return -1; - return virCgroupNewMachineManual(name, drivername, pidleader,