On 8/8/24 15:03, Ales Musil wrote:
> The test was checking if there was any recompute after all operations.
> This proven to be flaky because there might have been "random"
> recompute happening in the middle due to the following events:
> 
> 1) SB becomes read only
> 2) non_vif_data handler for OVS_interface fails
> 3) Recompute of non_vif_data is not allowed because SB is read only
> 4) non_vif_data is marked as cancelled
> 5) Next engine run will do force recompute of all nodes
> 
> To prevent that check specifically for the handler that is supposed
> to gracefully handle the interface changes.
> 
> Fixes: 603f38da062b ("controller: Remove OvS iface type check in I-P 
> processing.")
> Signed-off-by: Ales Musil <[email protected]>
> ---
> v2: Address comment from Ilya:
>     - Add recompute case to the test so if the debug format changes the test 
> fails.
>     - Add missing checks for ovs-vsctl.
> ---
>  tests/ovn-controller.at | 43 +++++++++++++++++++++++++----------------
>  1 file changed, 26 insertions(+), 17 deletions(-)
> 

Seems reasonable.

Acked-by: Ilya Maximets <[email protected]>
_______________________________________________
dev mailing list
[email protected]
https://mail.openvswitch.org/mailman/listinfo/ovs-dev

Reply via email to