limit on total number of concurrency shares for API Priority & Fairness?

[David Oppenheimer]

The total number of concurrency shares allocated to the priority levels you get in a default cluster ("kubectl get prioritylevelconfigurations.flowcontrol.apiserver.k8s.io") adds up to 245. We tried adding an additional priority level with 32 concurrency shares on GKE, but it was rejected with the following (GKE-specific) admission control error:

admission webhook "flowcontrol-guardrails.common-webhooks.networking.gke.io" denied the request: priority level configuration <name of priority level we were trying to add> cannot be added as requested shares 32 exceed available shares by 2; re-allocate shares from other non-cluster-critical priority level configurations and retry

It seems like GKE is enforcing a limit of 275 shares, but I could not find any documentation about such a limit in the open-source Kubernetes documentation nor the GKE documentation.

Does anyone know what rule is being enforced here?

Thanks!

[David Oppenheimer]

Just pinging this thread in case it got lost in the shuffle over the Thanksgiving holiday.

Thanks in advance for any suggestions!

[Federico Bongiovanni]

Hi David, thanks for your questions. 

In respect to GKE, probably this is not the right Channel to ask what a particular Cloud Provider does or how it configures the limits.

On Upstream K8s, you can see the default configurations using kubectl get prioritylevelconfigurations.flowcontrol.apiserver.k8s.io or kubectl get prioritylevelconfigurations. 

--
You received this message because you are subscribed to the Google Groups "K8s API Machinery SIG" group.
To unsubscribe from this group and stop receiving emails from it, send an email to kubernetes-sig-api-m...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/kubernetes-sig-api-machinery/CACygFFT6Gg3vPD4nVtQEH-yqR2pHXBBsaKfeTSzJd-0SuOh9ng%40mail.gmail.com.