Anyone else building a Perl wrapper to debounce SDN applies?

[iamgrudge]

The idea is to handle cases where multiple SDN applies are triggered back-to-back (like from automation) while a reload is already ongoing at Time X. Instead of just failing with a "Task Locked" error, the plugin would queue the request. If even newer requests come in while it's still busy, it replaces the existing queued one with the latest version. Once the initial SDN apply finishes, the plugin automatically triggers that last stored request.

Has anyone else implemented this kind of "coalescing" logic to stop API spam from breaking cluster reloads?

 

[GMer]

I haven’t done it myself, but that approach makes sense. Coalescing requests should help avoid task locks and unnecessary reloads.
I’m curious if anyone’s handled this at the API or scheduler level instead of using a wrapper?

 

[shanreich]

Is there a reason why you cannot make the changes to the SDN configuration all at once and then apply when they're done? If there's some issue there then it is probably better to fix that.

 

[iamgrudge]

Lot of testings are going on in my pve-cluster, specifically related to network so people are creating/deleting zones, network etc and applying sdn...
thats why I need to have a solution