kubernetes
/
autoscaler
mirror of https://github.com/kubernetes/autoscaler.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

Merge pull request #8307 from ialidzhikov/fix/header 

docs: Unify the header for a section
This commit is contained in:
Kubernetes Prow Robot 2025-07-09 05:25:26 -07:00 committed by GitHub
parent cc01c8756f 2abd557186
commit 8f9a24aafe
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 2 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
vertical-pod-autoscaler/docs/examples.md
View File

@ -110,9 +110,9 @@ It is possible to set the failurePolicy of the webhook to `Fail` by passing `--w
Please use this option with caution as it may be possible to break Pod creation if there is a failure with the VPA.
Using it in conjunction with `--ignored-vpa-object-namespaces=kube-system` or `--vpa-object-namespace` to reduce risk.
### Specifying global maximum allowed resources to prevent pods from being unschedulable
## Specifying global maximum allowed resources to prevent pods from being unschedulable
The [Known limitations dcoument](./known-limitations.md) outlines that VPA (vpa-recommender in particular) is not aware of the cluster's maximum allocatable and can recommend resources which will not fit even the largest node in the cluster. This issue occurs even when the cluster uses the [Cluster Autoscaler](https://github.com/kubernetes/autoscaler/blob/master/cluster-autoscaler/FAQ.md#basics). The vpa-recommender's resource recommendation can exceed the allocatable of the largest node in the cluster. Hence, pod's will be unschedulable (in `Pending` state) and the pod wouldn't fit the cluster even if a new node is added by the Cluster Autoscaler.
The [Known limitations document](./known-limitations.md) outlines that VPA (vpa-recommender in particular) is not aware of the cluster's maximum allocatable and can recommend resources which will not fit even the largest node in the cluster. This issue occurs even when the cluster uses the [Cluster Autoscaler](https://github.com/kubernetes/autoscaler/blob/master/cluster-autoscaler/FAQ.md#basics). The vpa-recommender's resource recommendation can exceed the allocatable of the largest node in the cluster. Hence, pod's will be unschedulable (in `Pending` state) and the pod wouldn't fit the cluster even if a new node is added by the Cluster Autoscaler.
It is possible to mitigate this issue by specifying the `--container-recommendation-max-allowed-cpu` and `--container-recommendation-max-allowed-memory` flags of the vpa-recommender. These flags represent the global maximum amount of cpu/memory that will be recommended **for a container**. If the VerticalPodAutoscaler already defines a max allowed (`.spec.resourcePolicy.containerPolicies.maxAllowed`) then it takes precedence over the global max allowed. The global max allowed is merged to the VerticalPodAutoscaler's max allowed if VerticalPodAutoscaler's max allowed is specified only for cpu or memory. If the VerticalPodAutoscaler does not specify a max allowed and a global max allowed is specified, then the global max allowed is being used.
The recommendation for computing the `--container-recommendation-max-allowed-cpu` and `--container-recommendation-max-allowed-memory` values for your cluster is to use the largest node's allocatable (`.status.allocatable` field of a node) minus the DaemonSet pods resource requests minus a safety margin: