Flag to disable caching on ApiServerSource adapter

[rh-hemartin]

Problem
We do have an ApiServerSource tracking 5 resources across 1000+ namespaces which results into a lot of:

I0718 14:02:19.224595 1 request.go:697] Waited for 17m12.738947279s due to client-side throttling, not priority and fairness, request: GET:https://172.30.0.1:443/apis/tekton.dev/v1/namespaces/.../taskruns?limit=500&resourceVersion=0

due to the number of requests (5 resources * 1000+ namespaces = ~5000).

We would like to have a flag in the ApiServerSource propagated to the Adapter Deployment so it does not cache.

Persona:

System Operator: wants to reduce load in Kubernetes.

Exit Criteria

There are no caching calls from the adapter container.

Time Estimate (optional):

I would say that pure code would be ~24 hours of good quality coding time.

Additional context (optional)
I need to develop a reproducer.

[Cali0707]

/triage accepted

[Kunal1522]

hey @Cali0707 i would like to work on this !! can i take this up ?

[Cali0707]

Sure @Kunal1522 feel free to work on this

/assign @Kunal1522

[Kunal1522]

i did some research and i think one solution could be when disableCache is true, internally force cluster-scoped watches (like namespaceSelector: {}). this way 5 resources × 1000 namespaces = 5000 api connections gets reduced to just 5 connections (one per resource type). then we do client-side filtering in the adapter to drop events from namespaces that don't match the original selector, so same events reach the sink. this should fix the throttling issue since we're making way fewer api calls, and it's backward compatible because the flag is optional.
should i go ahead and open pr ?