[ovs-dev] [PATCHv3] upcall: Allow max_idle be configured through ovs-vsctl.

[Ben Pfaff]

Thanks.

On Wed, Feb 19, 2014 at 02:34:40PM -0800, Joe Stringer wrote:
> Ah, I see now. That would be my misunderstanding, from following the
> "flow-limit" and "flow_limit" configuration options and equating them with
> each other. I plan to send a fresh version.
> 
> 
> On 19 February 2014 14:20, Ben Pfaff <blp at nicira.com> wrote:
> 
> > bridge_reconfigure() reads other-config:max-idle from the Open_vSwitch
> > table.  That part is fine.  I'm complaining about max_idle in the
> > Flow_Table table.  Why do both exist?
> >
> > On Wed, Feb 19, 2014 at 02:00:44PM -0800, Joe Stringer wrote:
> > > Hmm. The way I follow it is that bridge_reconfigure() will read it from
> > the
> > > table, and call ofproto_set_max_idle() to set the ofproto_flow_idle
> > > variable. Later, in revalidate_udumps(), we take a local copy of
> > > ofproto_flow_idle, and this is used later in the function (by code that
> > > already exists).
> > >
> > > I've just tested it, and it works as expected - e.g., setting a value of
> > 1
> > > makes us delete flows from the datapath incredibly quickly, as observed
> > via
> > > ovs-appctl upcall/show. Is the "minInteger"/"maxInteger" meant to prevent
> > > the user from setting values outside that range, or are they merely
> > > suggestions for the manual?
> > >
> > >
> > > On 19 February 2014 13:45, Ben Pfaff <blp at nicira.com> wrote:
> > >
> > > > On Tue, Feb 18, 2014 at 11:01:39AM -0800, Joe Stringer wrote:
> > > > > Signed-off-by: Joe Stringer <joestringer at nicira.com>
> > > >
> > > > This patch adds a new "max_idle" column to the Flow_Table table, and
> > > > documents it, but doesn't appear to use it anywhere in actual code.
> > > >
> >