Re: [ovs-dev] [PATCH ovn] ovn.at: Fix flaky test "controller I-P handling with monitoring disabled".

2021-10-06 Thread Numan Siddique
On Wed, Sep 15, 2021 at 4:53 AM Xavier Simonart wrote: > > Test was waiting for port to be up in SBDB before checking number of flows > in OVS. However, there is no guarantee that all flows are installed > in OVS when port is up. Test was randomly failing as some flows were > installed, but not al

Re: [ovs-dev] [PATCH ovn] ovn.at: Fix flaky test "controller I-P handling with monitoring disabled".

2021-10-02 Thread Han Zhou
On Fri, Oct 1, 2021 at 7:53 AM Mark Gray wrote: > > On 27/09/2021 12:13, Xavier Simonart wrote: > > Hi Mark > > Sorry for the late reply. If you cc me, I will usually reply earlier. > > > > > After further investigation, checking that port is up does not fix the > > issue, as port up is reported t

Re: [ovs-dev] [PATCH ovn] ovn.at: Fix flaky test "controller I-P handling with monitoring disabled".

2021-10-01 Thread Mark Gray
On 27/09/2021 12:13, Xavier Simonart wrote: > Hi Mark Sorry for the late reply. If you cc me, I will usually reply earlier. > > After further investigation, checking that port is up does not fix the > issue, as port up is reported too early by the ovn controller (before all > flows are installed

Re: [ovs-dev] [PATCH ovn] ovn.at: Fix flaky test "controller I-P handling with monitoring disabled".

2021-09-27 Thread Xavier Simonart
Hi Mark After further investigation, checking that port is up does not fix the issue, as port up is reported too early by the ovn controller (before all flows are installed), when using conditional monitoring. I proposed different potential solutions on the mailing list and consensus seems to be

Re: [ovs-dev] [PATCH ovn] ovn.at: Fix flaky test "controller I-P handling with monitoring disabled".

2021-09-16 Thread Xavier Simonart
Hi Mark Thanks for looking into this. Yes and no... I do not think that the patch itself will help, but it pointed me to the existence of Logical_Switch_Port.up, which in this case should do the trick. I will update the patch: instead of looking at the flows, I will use something like *wait_colu

Re: [ovs-dev] [PATCH ovn] ovn.at: Fix flaky test "controller I-P handling with monitoring disabled".

2021-09-16 Thread Mark Gray
On 15/09/2021 09:52, Xavier Simonart wrote: > Test was waiting for port to be up in SBDB before checking number of flows Would this feature prevent this? 5c3371922994 ("if-status: Add OVS interface status management module.") > in OVS. However, there is no guarantee that all flows are installed

[ovs-dev] [PATCH ovn] ovn.at: Fix flaky test "controller I-P handling with monitoring disabled".

2021-09-15 Thread Xavier Simonart
Test was waiting for port to be up in SBDB before checking number of flows in OVS. However, there is no guarantee that all flows are installed in OVS when port is up. Test was randomly failing as some flows were installed, but not all. To fix this, we wait until the last flow (with actions=output)