CODEWITHSUNDEEP
kubernetesgoogle-cloud-platformgoogle-kubernetes-engine
Sanish Joseph
Sanish Joseph

Reputation: 2256

Container CPU request value for Kubernetes

I have several microservices and I am deploying them to GCP Kubernetes. I am using the free credits and trying out my deployments. My question is when we define CPU requests, based on what we define it? I have set it to 250mCPU but that fills my cluster nodes which are small in with CPU.

Currently, I have 3 nodes with 940mCPU allocable CPU and 3 nodes of the same kind. Now I deployed one API with 3 replicas and assigned 250mCPU for each. With all Kubernetes internal items, all nodes are almost full.

So my question is, based on what we can assign a value for CPU for a service. 250mCPU was a random value. What are others doing to find a minimum CPU for Kubernetes? I have one ASP.NET Core API and 8 NodeJS APIs. If it's based on usage, what's' the best values to start with for a new product?

Upvotes: 2

Views: 319

Answers (1)

paltaa
paltaa

Reputation: 3244

Mostly by doing stress testing on your application, also, you could use a vertical pod autoscaler in "recomendation" mode, which will monitor your application for some time and then make a recomendation for you to set limits.

Documentation: https://cloud.google.com/kubernetes-engine/docs/concepts/verticalpodautoscaler

Remember you cannot use a vertical pod autoscaler and horizontal pod autoscaler at the same time, unless the vertical pod autoscaler its in recomendation mode.

Upvotes: 4

Related Questions