On Mon, May 01, 2017 at 05:31:54PM -0700, Han Zhou wrote:
> On Mon, May 1, 2017 at 11:19 AM, Ben Pfaff <[email protected]> wrote:
> >
> > On Tue, Apr 18, 2017 at 11:12:14AM -0700, Han Zhou wrote:
> > > In commit c1bfdd9d it disables probe when not needed, but commit
> > > 715038b6 updates ovn-controller probe interval for OVNSB DB
> > > periodically according to ovn-remote-probe-interval config, and sets
> > > it to DEFAULT_PROBE_INTERVAL_MSEC if not configured, even if the
> > > connection type is unix socket which doesn't need probe.
> > >
> > > This fix avoids probe interval update if not needed (always set to 0).
> > >
> > > Signed-off-by: Han Zhou <[email protected]>
> > > ---
> > >
> > > Notes:
> > >     v1->v2: fix commit id mentioned in commit message.
> >
> > Thanks for reporting and fixing this bug.  I agree there's a bug here,
> > but I'd rather honor the user's explicit request to set a nonzero probe
> > interval, if there is such a request in ovsdb.
> >
> > How about this?  I have not tested it, beyond verifying that it
> > compiles.
> >
> > Thanks,
> >
> > Ben.
> >
> 
> Yes, this idea is better. But the code has a problem.
> 
> > @@ -933,6 +934,8 @@ update_probe_interval(struct controller_ctx *ctx)
> >                      ? smap_get_int(&cfg->external_ids,
> >                                     "ovn-remote-probe-interval",
> >                                     DEFAULT_PROBE_INTERVAL_MSEC)
> > -                    : DEFAULT_PROBE_INTERVAL_MSEC);
> > +                    : stream_or_pstream_needs_probes(ovnsb_remote)
> > +                    ? DEFAULT_PROBE_INTERVAL_MSEC
> > +                    : 0);
> 
> Here is the fix (tested):

Thanks for the testing and the fix.  Can you please re-send this as a
full patch?  The version that made it to the list got wordwrapped.

Thanks,

Ben.
_______________________________________________
dev mailing list
[email protected]
https://mail.openvswitch.org/mailman/listinfo/ovs-dev

Reply via email to