On 1/30/25 14:01, Aaron Conole wrote:
> Dumitru Ceara <[email protected]> writes:
> 
>> On 1/29/25 8:12 PM, Frode Nordahl wrote:
>>> ons. 29. jan. 2025, 15:13 skrev Frode Nordahl <[email protected]>:
>>>
>>>> On Wed, Jan 29, 2025 at 9:20 AM Frode Nordahl <[email protected]> wrote:
>>>>>
>>>>> The commit in the fixes tag introduced support for adding LRPs
>>>>> without networks, leaving them with only the IPv6 Link Local
>>>>> Address (LLA).
>>>>>
>>>>> This is a very useful feature for routing between LRs, avoiding
>>>>> waste of a scarce resource such as IPv4 addresses.
>>>>>
>>>>> It did however miss a couple of direct references to
>>>>> router_port->lrp_networks.ipv4_addrs[0].
>>>>>
>>>>> Check for presence of IPv4 addresses, and use IPv6 LLA when none
>>>>> are present.
>>>>>
>>>>> Expand "ovn -- NAT and Load Balancer flows" test case to include
>>>>> a neighbour LR with LRP without IPv4 address, as that is enough
>>>>> to expose the crash.
>>>>>
>>>>> Note that more elaborate flow checks will be added in a subsequent
>>>>> patch that fixes a logic bug in the routing IPv4 over IPv6 next
>>>>> hop patch.
>>>>>
>>>>> Cc: MJ Ponsonby <[email protected]>
>>>>> Fixes: 0ee90e29eb8b ("Allow creation of a LRP without ipv4.")
>>>>> Signed-off-by: Frode Nordahl <[email protected]>
>>>>> ---
>>>>
>>>> Recheck-request: github-robot
>>>>
>>>
>>> One more time, the test in question does look flaky and does not fail in
>>> GHA on a different  but otherwise equal fork.
>>>
>>> Knowing we might meet this test again in downstream CI infrastructure I'll
>>> make a note for us to revisit it at a later time.
>>>
>>
>> I haven't seen this one recently.  But yes, we should try to figure out
>> why this failed.
>>
>>> Recheck-request: github-robot
>>>
>>
>> Hmm..  For some reason the robot refused to re-run the CI for the first
>> patch of the series and just replied with the same link as in the
>> original CI run:
>> https://mail.openvswitch.org/pipermail/ovs-build/2025-January/044096.html
>>
>> That still points to the same job logs as in the original run:
>> https://mail.openvswitch.org/pipermail/ovs-build/2025-January/044075.html
>>
>> That is, they both point to:
>> Build URL: https://github.com/ovsrobot/ovn/actions/runs/13027499538
>>
>> Aaron, is this something we need to fix on the robot side?
> 
> I applied a change earlier that would try to rerun the failed builds
> only.  However, that makes the run URL not change.  So it becomes
> confusiong to understand "at a glance" - did it really even restart?
> I turned off that config option for now since it has some confusion.
> Supposedly it did restart - but it is difficult to know that.

It is a long standing issue that robot only re-runs one of the checks,
but reports both: https://github.com/ovsrobot/pw-ci/issues/16

> 
>> Thanks,
>> Dumitru
> 

_______________________________________________
dev mailing list
[email protected]
https://mail.openvswitch.org/mailman/listinfo/ovs-dev

Reply via email to