Open
Description
Hi all,
My CRD controller manages a variety of child resources. During reconciliation, I compare the current spec against the desired state. However, some fields are stored in a different format by Kubernetes – for example, CPU values. If the user specifies 0.2, Kubernetes stores it as 200m, causing unnecessary event churn.
I could preprocess these values in my spec before applying, but is this a sustainable and reliable approach? This issue isn't limited to CPU – memory limits and potentially other fields may also be affected.
Would using a predicate for spec comparison be more reliable in such cases? How do you handle these formatting discrepancies effectively?
Metadata
Metadata
Assignees
Labels
No labels