On Thu, Sep 3, 2020 at 7:44 AM Han Zhou <[email protected]> wrote: > > The checking of lsp "up" status before adding ARP responder flows was > added to avoid confusion in cases when a network admin sees ARP response > for VM/containers that is not up on chassis yet. However, this check > introduces an extra round of flow change in SB and triggers computes > on hypervisors which is unnecessary cost in most cases, especially for > large scale environment. To improve this, this patch provides an option > check_lsp_is_up to disable the check (when setting to "false") for the > use cases when ARP reponse for a port that is "down" isn't considered > harmful. > > Signed-off-by: Han Zhou <[email protected]>
Few minor comments. Acked-by: Numan Siddique <[email protected]> with those addressed. Thanks Numan > --- > northd/ovn-northd.8.xml | 15 ++++++++++----- > northd/ovn-northd.c | 7 ++++++- > ovn-nb.xml | 18 ++++++++++++++++++ > tests/ovn-northd.at | 14 ++++++++++++++ > 4 files changed, 48 insertions(+), 6 deletions(-) > > diff --git a/northd/ovn-northd.8.xml b/northd/ovn-northd.8.xml > index 989e364..df43eb6 100644 > --- a/northd/ovn-northd.8.xml > +++ b/northd/ovn-northd.8.xml > @@ -763,9 +763,11 @@ output; > > <p> > These flows are omitted for logical ports (other than router ports > or > - <code>localport</code> ports) that are down, for logical ports of > - type <code>virtual</code> and for logical ports with 'unknown' > - address set. > + <code>localport</code> ports) that are down (unless <code> > + check_lsp_is_up</code> is configured as false in > <code>options</code> > + column of <code>NB_Global</code> table of the > <code>Northbound</code> > + database), for logical ports of type <code>virtual</code> and for > + logical ports with 'unknown' address set. > </p> > </li> > > @@ -812,8 +814,11 @@ nd_na_router { > > <p> > These flows are omitted for logical ports (other than router ports > or > - <code>localport</code> ports) that are down and for logical ports > of > - type <code>virtual</code>. > + <code>localport</code> ports) that are down (unless <code> > + check_lsp_is_up</code> is configured as false in > <code>options</code> > + column of <code>NB_Global</code> table of the > <code>Northbound</code> > + database), for logical ports of type <code>virtual</code> and for > + logical ports with 'unknown' address set. > </p> > </li> > > diff --git a/northd/ovn-northd.c b/northd/ovn-northd.c > index 7be0e85..291713e 100644 > --- a/northd/ovn-northd.c > +++ b/northd/ovn-northd.c > @@ -86,6 +86,8 @@ static struct eth_addr mac_prefix; > > static bool controller_event_en; > > +static bool check_lsp_is_up; > + > /* MAC allocated for service monitor usage. Just one mac is allocated > * for this purpose and ovn-controller's on each chassis will make use > * of this mac when sending out the packets to monitor the services > @@ -6739,7 +6741,8 @@ build_lswitch_flows(struct hmap *datapaths, struct hmap > *ports, > * - port type is router or > * - port type is localport > */ > - if (!lsp_is_up(op->nbsp) && strcmp(op->nbsp->type, "router") && > + if (check_lsp_is_up && > + !lsp_is_up(op->nbsp) && strcmp(op->nbsp->type, "router") && > strcmp(op->nbsp->type, "localport")) { > continue; > } > @@ -11722,6 +11725,8 @@ ovnnb_db_run(struct northd_context *ctx, > > controller_event_en = smap_get_bool(&nb->options, > "controller_event", false); > + check_lsp_is_up = smap_get_bool(&nb->options, > + "check_lsp_is_up", true); > > build_datapaths(ctx, datapaths, lr_list); > build_ports(ctx, sbrec_chassis_by_name, datapaths, ports); > diff --git a/ovn-nb.xml b/ovn-nb.xml > index 1f2dbb9..4e34518 100644 > --- a/ovn-nb.xml > +++ b/ovn-nb.xml > @@ -147,6 +147,24 @@ > </p> > </column> > > + <column name="options" key="check_lsp_is_up"> How about "ignore_lport_down" ? The default would be false. If you're fine with this, then you need to invert the logic in the code. I don't have a very strong opinion. If you want to stick with check_lsp_is_up, I'm fine with it. > + <p> > + If set to true, ARP/ND reply flows for logical switch ports will be > + installed only if the port is up, i.e. claimed by a Chassis. If set > + to false, these flows are installed regardless of the status of the > + port, which can result in a situation that ARP request to an IP is > + resolved even before the relevant VM/container is running. For > + environments where this is not an issue, setting it to > + <code>false</code> can reduce the load and latency of the control > + plane. The default value is <code>true</code>. > + </p> > + > + <p> > + If the value is nonzero, then it will be forced to a value of > + at least 1000 ms. > + </p> Looks like some copy/paste issue ? Because the value is a boolean :). Thanks Numan > + </column> > + > <group title="Options for configuring interconnection route > advertisement"> > <p> > These options control how routes are advertised between OVN > diff --git a/tests/ovn-northd.at b/tests/ovn-northd.at > index 8344c7f..c98df49 100644 > --- a/tests/ovn-northd.at > +++ b/tests/ovn-northd.at > @@ -1781,3 +1781,17 @@ AT_CHECK([ovn-sbctl lflow-list | grep > "ls_out_pre_lb.*priority=100" | grep reg0 > ]) > > AT_CLEANUP > + > +AT_SETUP([ovn -- disable check_lsp_is_up]) > +ovn_start > + > +ovn-nbctl ls-add sw0 > +ovn-nbctl lsp-add sw0 sw0-p1 -- lsp-set-addresses sw0-p1 "aa:aa:aa:aa:aa:aa > 10.0.0.1" > + > +ovn-nbctl --wait=sb sync > +AT_CHECK([ovn-sbctl lflow-list | grep arp | grep 10\.0\.0\.1], [1], [ignore]) > + > +ovn-nbctl --wait=sb set NB_Global . options:check_lsp_is_up=false > +AT_CHECK([ovn-sbctl lflow-list | grep arp | grep 10\.0\.0\.1], [0], [ignore]) > + > +AT_CLEANUP > -- > 2.1.0 > > _______________________________________________ > dev mailing list > [email protected] > https://mail.openvswitch.org/mailman/listinfo/ovs-dev > _______________________________________________ dev mailing list [email protected] https://mail.openvswitch.org/mailman/listinfo/ovs-dev
