Closed alculquicondor closed 1 year ago
/assign @trasc
@alculquicondor , this could, theoretically, only happen when the CRDs are v0.4 but the running manager is still 0.3 , do we want to cover this scenario?
Anyway I'll have another look on Tuesday.
In the case of admission, you are right. And no, we don't need to support that.
But I would be worried about cases where a 0.4 controller doesn't populate the field for an old workload, and gets rejected by the API. Although I'm not sure if this is possible. Maybe the controller would still send count: 0
.
Anyway I'll have another look on Tuesday.
Thanks!
It looks like there is another way we can get to that error. While SSA even if only an unrelated condition is changed, the entire sub-resource is re validated. Well heed to make the count optional from an API point point view.
Ah... yes, SSA does checks that other methods don't
Thanks for investigating!
What happened:
What you expected to happen:
The admission to succeed.
How to reproduce it (as minimally and precisely as possible):
This was reproduced by using the helm charts in the main branch, while using a 0.3.1 binary.
Anything else we need to know?:
We probably need to make
count
optional.Environment:
kubectl version
):git describe --tags --dirty --always
):cat /etc/os-release
):uname -a
):