On 09/06/2025 1:14, Ilya Maximets wrote:
> On 6/8/25 10:45 AM, Roi Dayan via dev wrote:
>> From: Eli Britstein <[email protected]>
>>
>> The following messages are reduced to DBG.
>> 2025-05-11T10:33:26.778Z|07801|connmgr|INFO|br-sfc<->unix#24021: 2
>> flow_mods in the last 0 s (2 adds)
>> 2025-05-11T10:33:30.582Z|07802|connmgr|INFO|br-sfc<->unix#24024: 2
>> flow_mods in the last 0 s (2 adds)
>> 2025-05-11T10:33:30.648Z|07803|connmgr|INFO|br-sfc<->unix#24027: 2
>> flow_mods in the last 0 s (2 adds)
>
> Any justification for this? These logs are normally pretty useful to see
> the change rate. If you see the constant stream of these messages, then
> you probably need to fix the controller that sends so many flow updates.
> Flow modifications cause revalidation, which is heavy, and so keeping OVS
> in a constant revalidation loop is unlikely to be healthy for the system.
>
> You may also consider having a stable connection instead of reconnecting
> for every change, e.g. have a real OF controller instead of invoking
> ovs-ofctl for every small change. This way the messages will be printed
> less frequently.
>
> Alternative might be to aggregate across multiple connections and then log
> once at lower frequency. Showing in the log all the accumulated connection
> handles might be cumbersome though.
>
> Either way, just demoting the fairly useful log doesn't seem reasonable.
>
> Best regards, Ilya Maximets.
Thanks for the response. I'll check your suggestions.
_______________________________________________
dev mailing list
[email protected]
https://mail.openvswitch.org/mailman/listinfo/ovs-dev
