On Wed, Sep 27, 2023 at 10:55:04AM +0100, Daniel P. Berrangé wrote: > On Mon, Sep 25, 2023 at 08:58:38PM +0200, Andrea Bolognani wrote: > > They are unnecessary, since all sockets for a service are now > > enabled as soon as one of them is and each service has a very > > strong dependency on all of its sockets. > > You earlier modified the .service units to have BindsTo= for > each of the sockets it depends to. > > Thus if any one of the .sockets is stopped, this means the > .service is stopped too. > > The logic removed here though was doing a different job. That > said that that if $FOO.socket is stopped, it would force stop > the $FOO-admin.socket and $FOO-ro.socket too. > > IOW, it prevented having only the RO/admin sockets running, > without the primary socket. > > I believe that's still needed > > Also, you didn't add BindsTo on the libvirtd.service, because > that has to be able to run without socket activation for > upgrade scenarios. So we shouldn't be modifying the libvirtd > sockets anyway. I'll perform some testing just to make sure, but I think you're right and I will most likely drop this patch in v2. -- Andrea Bolognani / Red Hat / Virtualization
