[ovs-dev] [PATCH] ofproto-dpif-upcall: Log the value of flow limit.

Flavio Leitner fbl at sysclose.org
Tue Sep 29 20:09:01 UTC 2020


On Tue, Sep 29, 2020 at 04:53:22PM +0200, Ilya Maximets wrote:
> On 9/15/20 9:20 PM, Flavio Leitner wrote:
> > The datapath flow limit is calculated by revalidators so
> > log the value as well.
> > 
> > Signed-off-by: Flavio Leitner <fbl at sysclose.org>
> > ---
> >  ofproto/ofproto-dpif-upcall.c | 3 ++-
> >  1 file changed, 2 insertions(+), 1 deletion(-)
> > 
> > diff --git a/ofproto/ofproto-dpif-upcall.c b/ofproto/ofproto-dpif-upcall.c
> > index 72a5b4d73..05a912f57 100644
> > --- a/ofproto/ofproto-dpif-upcall.c
> > +++ b/ofproto/ofproto-dpif-upcall.c
> > @@ -1283,7 +1283,8 @@ should_install_flow(struct udpif *udpif, struct upcall *upcall)
> >      atomic_read_relaxed(&udpif->flow_limit, &flow_limit);
> >      if (udpif_get_n_flows(udpif) >= flow_limit) {
> >          COVERAGE_INC(upcall_flow_limit_hit);
> > -        VLOG_WARN_RL(&rl, "upcall: datapath flow limit reached");
> > +        VLOG_WARN_RL(&rl, "upcall: datapath flow limit reached (max: %u)",
> 
> I think that we're not printing this value just because it's dynamically
> adjustable by revalidators.  From that perspective, I don't think that
> 'max' is a good word to describe the value, because users might think
> that they have a direct control over it and might think that it's the
> value set by the 'flow_limit' configuration knob.
> 'current dynamic limit' or something similar might be better choice, but
> I'm not sure.

Sent v2:
https://mail.openvswitch.org/pipermail/ovs-dev/2020-September/375646.html

-- 
fbl


More information about the dev mailing list