[ovs-dev] [PATCH] ovs-save: Don't always include the default flow during restore

[Guru Shetty]

On Thu, 13 Sep 2018 at 10:38, Timothy Redaelli <tredaelli at redhat.com> wrote:

> On Wed, 12 Sep 2018 10:36:22 -0700
> Guru Shetty <guru at ovn.org> wrote:
>
> > On Tue, 11 Sep 2018 at 09:07, Flavio Leitner <fbl at sysclose.org> wrote:
> >
> > > On Sun, Sep 09, 2018 at 02:20:02PM +0200, Timothy Redaelli wrote:
> > > > Currently the default flow (actions=NORMAL) is present in the
> > > > flow table
> > > after
> > > > the flow table is restored also when the default flow is removed.
> > > >
> > > > This commit changes the behaviour of the "ovs-save save-flows"
> > > > command
> > > to use
> > > > "replace-flows" instead of "add-flows" to restore the flows. This
> > > > is
> > > needed in
> > > > order to always have the new flow table as it was before
> > > > restoring it.
> > > >
> > > > Reported-by: Flavio Leitner <fbl at sysclose.org>
> > > > Reported-at: https://bugzilla.redhat.com/show_bug.cgi?id=1626096
> > > > Signed-off-by: Timothy Redaelli <tredaelli at redhat.com>
> > > > ---
> > >
> > > From the performance point of view, I thought we could do a
> > > del-flows first followed by add-flows, but you mentioned off
> > > list that the performance with replace-flows is still good.
> > >
> >
> > Can you please provide some data. For e.g., with 200,000 flows.
>
> Hi,
> my usual setup to test this scenario is:
>
> VM1 (tap0) <-> OVS (ovsbr0) <-> VM2 (tap1).
>
> In this case I'm testing it by using kernel datapath.
>
> I add 1 milion of rules like
> "table=0, dl_src=$MAC, actions=drop" with different mac addresses,
> using a script and the 2 rules in order to make the VM pings:
> table=0,in_port=1,action=output:2
> table=0,in_port=2,action=output:1
>
> I launch a ping (1 second for each ping) from VM1 to VM2 IP, then I
> reload openvswitch on the host and I have the following downtime (lost
> ICMP packets):
>
> with add-flows (aka without the patch):  11 seconds
> with replace-flows (aka with the patch): 11 seconds too
>
Thats good enough for me. I wanted to make sure that there is no penalty
here for large flows.