You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Allow to override defaults kubernetes client QPS & Burst options, like we can do with KEDA metrics apiserver.
Use-Case
We have a cluster with ~1600 ScaledObjects and we encounter heavy client-side throttling (more than a minute).
ScaledObjects takes minutes (or hours?) to be reconciled by the manager (a new ScaledObject was still not reconciled after 45min).
Anything else?
I've build an image with new flags to configure these options. We tested with QPS=100 & Burst=150 (max reconciles = 20) and we no longer have the issue. We still have some client-side throttling but they do not exceed 2s and only during the first minutes on startup.
The text was updated successfully, but these errors were encountered:
Proposal
Allow to override defaults kubernetes client QPS & Burst options, like we can do with KEDA metrics apiserver.
Use-Case
We have a cluster with ~1600 ScaledObjects and we encounter heavy client-side throttling (more than a minute).
ScaledObjects takes minutes (or hours?) to be reconciled by the manager (a new ScaledObject was still not reconciled after 45min).
Anything else?
I've build an image with new flags to configure these options. We tested with QPS=100 & Burst=150 (max reconciles = 20) and we no longer have the issue. We still have some client-side throttling but they do not exceed 2s and only during the first minutes on startup.
The text was updated successfully, but these errors were encountered: