could you please help, iam getting this error Status: Conditions: Last Transition Time: 2022-11-06T22:13:41Z Message: Cannot read targetRef. Reason: Deployment kube-system/my-app does not exist Status: True Type: ConfigUnsupported Last Transition Time: 2022-11-06T22:13:41Z Message: No pods match this VPA object Reason: NoPodsMatched Status: True Type: NoPodsMatched Last Transition Time: 2022-11-06T22:13:41Z Message: No pods match this VPA object Reason: NoPodsMatched Status: False Type: RecommendationProvided Recommendation: Events: trying on k8s 1.25.3
Hi Mohan, thanks for watching. The error message says that it couldn't find a deployment named my-app. VPA resource act on the deployment that you specify but it couldn't find the deployment. Are you sure the deployment has been created and pods running?
Hi. Really nice explanation. But I am having problems right after running ./vpa-up.sh. My vpa-recommender pod is always in pending state. Like this: kube-system vpa-admission-controller-6d449d5876-sld4l 1/1 Running 0 19m 192.168.32.5 kworker kube-system vpa-recommender-6cb878fdc7-s6px8 0/1 Pending 0 19m kube-system vpa-updater-6c9d65dd6f-9dt4z 1/1 Running 0 19m 192.168.32.4 kworker Thanks !
Hi Abhinav, thanks for watching. Is there anything that you are doing differently to the steps in this video? If you are following the exact steps and seeing this issue, I can re-test this video. Otherwise your environment might be different to mine trying to reproduce. More information would be helpful. Cheers.
Its really a good video... I have been searching for vpa but did not fine any good one and finally got insight from this video ... Keep up the good work venkat 😀.
Hi Venkat. Thanks for the tutorial. I tried using the metrics-server in KIND (1master, 2 worker) cluster. kubectl top nodes (shows below errors) Default mode (no edit) "Error : metrics not available yet" with insecure-tls : Error from server (ServiceUnavailable): the server is currently unable to handle the request (get nodes.metrics.k8s.io) Tried with helm charts too. Not able to fix.
Hi Srini, I had the same problem recently. Neither direct deployment nor helm deployment worked for metrics-server. I think its because the nodes are containers rather than complete machines, the metrics are not being collected. I then had to use vagrant environment to use metrics server. I will try it again this weekend and let you know if I succeed. Thanks.
I am doing a similar approach. I created a nodejs app that receives requests from a benchmark tool and generate cpu and mem utilization to trigger the vpa. Additionally I created a service manifest to expose my app using load balacer, also I pointed to the image of my app and removed "command" and "args" of the deployment file. When the load is generated I can see through the minikube dashboard that the cpu and mem usage is increasing and reaching the limit but the vpa doesn't recreate the conteiner with more resources. Any ideas?
Hi Kewyn, thanks for watching. When you create the VPA resource, have you changed the updatepolicy to either Auto or Recreate? In this video, I showed with UpdatePolicy set to Off which will show the recommendations but won't take any action on increasing the resource limits. Then changed the updatepolicy to auto to actually increase the resource limits under load.
@@justmeandopensourceI am testing it using minikube. Metrics-server is running fine. It is possible to be something related to the service load balancer?
@@justmeandopensource no they have that repo helm repo add cowboysysop cowboysysop.github.io/charts/ helm install my-vertical-pod-autoscaler cowboysysop/vertical-pod-autoscaler --version 3.0.0 artifacthub.io/packages/helm/cowboysysop/vertical-pod-autoscaler
@@justmeandopensource when load goes down and container are not in much use ,does hpa recommendation get back to CPU 100m and memory 50mi and new pod if created has this configuration?
@@vishalvarshney961 I only tested hpa. In case of vpa, the controller will get a cpu/mem recommendation based on the current load and will modify the pods accordingly. You could test it yourself.
Venkat, awesome video! question about the autoscaler manifest itself, instead of creating individual VPA for each deployment, can you define multiple targetRef in the spec stanza for each additional deployment in that same VPA? Thanks in advance, keep up the good work!
Hi, I don't think you can have multiple targetRef sections in the VPA specification. I read through some documentations, and there were no mentions about multiple targets for autoscaling. Best to test it. Sure it will throw error. Thanks.
Hi Leonardo, thanks for watching this video. I haven't looked at GitOps yet but many viewers asked for it. So soon I will learn and upload video. Cheers.
It is possible to test it using minikube? Because when I try to do the gcp tutorial for VPA I got an error: error: unable to recognize "my-rec-vpa.yaml": no matches for kind "VerticalPodAutoscaler" in version "autoscaling.k8s.io/v1"
Hi Kewyn, did you managed to get metrics-server working in your minikube environment? May I know how you installed metrics-server? Did you follow the step in this video or did you do "minikube addons enable metrics-server"? Is "kubectl top nodes" showing metrics? Thanks.
Hi Kewyn, I just tested it on my Minikube environment and its working perfectly fine. However you need to do one slight change in the vpa api version. When you deploy my-rec-vpa, use v1beta2 instead of v1. You can see more details of my testing in the below pastebin link. pastebin.com/DLMwNidf Please let me know if it worked. Thanks.
Hi Venkat, awesome video! my question is that , actually approx i have done but one error showing .... Conditions: Last Transition Time: 2021-11-22T10:16:39Z Message: Cannot read targetRef. Reason: Unhandled targetRef apps/v1 / Rollout / my-rec-deployment, last error no matches for kind "Rollout" in group "apps" Status: True Type: ConfigUnsupported Last Transition Time: 2021-11-22T10:16:39Z Message: No pods match this VPA object Reason: NoPodsMatched Status: True Type: NoPodsMatched Last Transition Time: 2021-11-22T10:16:39Z Message: No pods match this VPA object Reason: NoPodsMatched Status: False Type: RecommendationProvided Recommendation:
I ran into a problem, VPA keeps terminating old pods and creating a new pod with specified resource request instead of recommended request, it seems that vpa detected a pod doesn't have recommended resource request, but failed to transfer the recommended request to new pod. Any idea about the problem?
Hi Jiaxing, thanks for watching. I haven't encountered that issue yet unfortunately to be able to assist you. It must be your configurations thats causing this. Have you tried that with any other deployments? Is it happening to all deployments using VPA?
![[ Kube 58 ] Using Vertical Pod Autoscaling in Kubernetes](/img/logo.svg){kind=logo}
