| Message ID | 20220829161320.3126201-1-xsimonar@redhat.com |
|---|---|
| State | Accepted |
| Headers | show |
| Series | [ovs-dev] tests: fix "recompute" test | expand |
| Context | Check | Description |
|---|---|---|
| ovsrobot/apply-robot | success | apply and check: success |
| ovsrobot/github-robot-_Build_and_Test | success | github build: passed |
| ovsrobot/github-robot-_ovn-kubernetes | fail | github build: failed |
On Mon, Aug 29, 2022 at 9:13 AM Xavier Simonart <xsimonar@redhat.com> wrote: > > Recompute test might fail on some slow systems due to sb inactivity probe > causing recomputes. > Disable inactivity probe to avoid this issue. > Hi Xavier, Thanks for making the test case stable, but before that shall we check why inactivity probe triggers recompute? Inactivity probe doesn't change any input, so it is not expected to cause recompute. If that's not the case, I'd fix the I-P engine instead of the test case. Thanks, Han > Fixes: a7c7d451 ("controller: avoid recomputes triggered by SBDB Port_Binding updates") > > Signed-off-by: Xavier Simonart <xsimonar@redhat.com> > --- > tests/ovn.at | 2 ++ > 1 file changed, 2 insertions(+) > > diff --git a/tests/ovn.at b/tests/ovn.at > index 5d73c3379..53519b92b 100644 > --- a/tests/ovn.at > +++ b/tests/ovn.at > @@ -32611,6 +32611,8 @@ for i in $(seq 1 $n_hv); do > ovn_attach n1 br-phys 192.168.0.$i 24 geneve > done > > +check ovn-sbctl set connection . inactivity_probe=0 > + > add_switch_ports() { > start_port=$1 > end_port=$2 > -- > 2.31.1 > > _______________________________________________ > dev mailing list > dev@openvswitch.org > https://mail.openvswitch.org/mailman/listinfo/ovs-dev
Hi Han Thanks for your feedback. Inactivity probe kicks in as the test is heavy (4 hv, each claiming hundreds of ports) + the fact that github might run multiple of those tests in parallel (monitor-all true / false, parallel northd). I saw very long poll interval in the controller (> 5 seconds), due to sb_port_binding handlers. When inactivity probe kicked in, commits to sb from controller failed as connection has dropped, causing recompute in the next run.Isn't this the expected behavior? Thanks Xavier On Mon, Aug 29, 2022 at 9:30 PM Han Zhou <zhouhan@gmail.com> wrote: > > > On Mon, Aug 29, 2022 at 9:13 AM Xavier Simonart <xsimonar@redhat.com> > wrote: > > > > Recompute test might fail on some slow systems due to sb inactivity probe > > causing recomputes. > > Disable inactivity probe to avoid this issue. > > > Hi Xavier, > > Thanks for making the test case stable, but before that shall we check why > inactivity probe triggers recompute? Inactivity probe doesn't change any > input, so it is not expected to cause recompute. If that's not the case, > I'd fix the I-P engine instead of the test case. > > Thanks, > Han > > > Fixes: a7c7d451 ("controller: avoid recomputes triggered by SBDB > Port_Binding updates") > > > > Signed-off-by: Xavier Simonart <xsimonar@redhat.com> > > --- > > tests/ovn.at | 2 ++ > > 1 file changed, 2 insertions(+) > > > > diff --git a/tests/ovn.at b/tests/ovn.at > > index 5d73c3379..53519b92b 100644 > > --- a/tests/ovn.at > > +++ b/tests/ovn.at > > @@ -32611,6 +32611,8 @@ for i in $(seq 1 $n_hv); do > > ovn_attach n1 br-phys 192.168.0.$i 24 geneve > > done > > > > +check ovn-sbctl set connection . inactivity_probe=0 > > + > > add_switch_ports() { > > start_port=$1 > > end_port=$2 > > -- > > 2.31.1 > > > > _______________________________________________ > > dev mailing list > > dev@openvswitch.org > > https://mail.openvswitch.org/mailman/listinfo/ovs-dev >
On Tue, Aug 30, 2022 at 12:21 AM Xavier Simonart <xsimonar@redhat.com> wrote: > > Hi Han > > Thanks for your feedback. > > Inactivity probe kicks in as the test is heavy (4 hv, each claiming hundreds of ports) + the fact that github might run multiple of those tests in parallel (monitor-all true / false, parallel northd). > I saw very long poll interval in the controller (> 5 seconds), due to sb_port_binding handlers. > When inactivity probe kicked in, commits to sb from controller failed as connection has dropped, causing recompute in the next run.Isn't this the expected behavior? > Thanks for clarification! So it is probe failure causing recompute instead of probe itself, which makes sense. I applied it to main with the minor word changes in the commit message. Han > Thanks > Xavier > > On Mon, Aug 29, 2022 at 9:30 PM Han Zhou <zhouhan@gmail.com> wrote: >> >> >> >> On Mon, Aug 29, 2022 at 9:13 AM Xavier Simonart <xsimonar@redhat.com> wrote: >> > >> > Recompute test might fail on some slow systems due to sb inactivity probe >> > causing recomputes. >> > Disable inactivity probe to avoid this issue. >> > >> Hi Xavier, >> >> Thanks for making the test case stable, but before that shall we check why inactivity probe triggers recompute? Inactivity probe doesn't change any input, so it is not expected to cause recompute. If that's not the case, I'd fix the I-P engine instead of the test case. >> >> Thanks, >> Han >> >> > Fixes: a7c7d451 ("controller: avoid recomputes triggered by SBDB Port_Binding updates") >> > >> > Signed-off-by: Xavier Simonart <xsimonar@redhat.com> >> > --- >> > tests/ovn.at | 2 ++ >> > 1 file changed, 2 insertions(+) >> > >> > diff --git a/tests/ovn.at b/tests/ovn.at >> > index 5d73c3379..53519b92b 100644 >> > --- a/tests/ovn.at >> > +++ b/tests/ovn.at >> > @@ -32611,6 +32611,8 @@ for i in $(seq 1 $n_hv); do >> > ovn_attach n1 br-phys 192.168.0.$i 24 geneve >> > done >> > >> > +check ovn-sbctl set connection . inactivity_probe=0 >> > + >> > add_switch_ports() { >> > start_port=$1 >> > end_port=$2 >> > -- >> > 2.31.1 >> > >> > _______________________________________________ >> > dev mailing list >> > dev@openvswitch.org >> > https://mail.openvswitch.org/mailman/listinfo/ovs-dev
diff --git a/tests/ovn.at b/tests/ovn.at index 5d73c3379..53519b92b 100644 --- a/tests/ovn.at +++ b/tests/ovn.at @@ -32611,6 +32611,8 @@ for i in $(seq 1 $n_hv); do ovn_attach n1 br-phys 192.168.0.$i 24 geneve done +check ovn-sbctl set connection . inactivity_probe=0 + add_switch_ports() { start_port=$1 end_port=$2
Recompute test might fail on some slow systems due to sb inactivity probe causing recomputes. Disable inactivity probe to avoid this issue. Fixes: a7c7d451 ("controller: avoid recomputes triggered by SBDB Port_Binding updates") Signed-off-by: Xavier Simonart <xsimonar@redhat.com> --- tests/ovn.at | 2 ++ 1 file changed, 2 insertions(+)