On Wed, Dec 16, 2015 at 12:07:23PM +0300, Ilya Maximets wrote: > On 16.12.2015 11:27, Ben Pfaff wrote: > > On Wed, Dec 16, 2015 at 09:58:56AM +0300, Ilya Maximets wrote: > >> > >> > >> On 16.12.2015 09:53, Ben Pfaff wrote: > >>> On Wed, Dec 16, 2015 at 09:50:56AM +0300, Ilya Maximets wrote: > >>>> While killing OVS may not free all allocated resources. > >>>> > >>>> Eample: > >>>> Socket for vhost-user port will stay in a system > >>>> after 'systemctl stop openvswitch' and opening > >>>> that port after restart will fail. > >>>> > >>>> Signed-off-by: Ilya Maximets <i.maxim...@samsung.com> > >>>> --- > >>>> > >>>> version 2: > >>>> * added '-T 1' > >>>> * '-t $1' --> '-t $rundir/$1.$pid.ctl' > >>> > >>> Thanks. > >>> > >>> Have you tested it? > >>> > >> > >> Yes. On my setup it works. > >> Only reason why it put RFC to that patch: > >> Is all daemons have unixctl exit command? Or we can just let ovs-appctl to > >> fail here > >> in that case? > > > > I think that all of our current daemons do have an "exit" command, but > > it might be worth removing the "2" after EXIT. The purpose of the > > delays in this chain is to wait a bit for a given signal to be processed > > through the operating system and the daemon, but EXIT should wait until > > the process has actually exited, or at least get very close to that. > > Maybe a short delay like .1 would be reasonable, to allow for a brief > > race window. > > I don't think so, because ovs-vswitchd replies on connection to ovs-appctl > immediately, but actual exiting performed while stopping execution in > main loop.
Then that means that this patch will generally make it take about 4 longer to stop OVS, since killing each daemon will sleep for 2 seconds. _______________________________________________ dev mailing list dev@openvswitch.org http://openvswitch.org/mailman/listinfo/dev
