We run our cluster with two nodes of type standard 2: 7.5Gb and 2vCPU
are there any recommended minimum size for a cluster on GKE. I assume there is no real master as this is a managed "service"?
I'm struggling to deal with resource limits.
We run our cluster with two nodes of type standard 2: 7.5Gb and 2vCPU
are there any recommended minimum size for a cluster on GKE. I assume there is no real master as this is a managed "service"?
I'm struggling to deal with resource limits.
There is no recommended minimum size. However, pods have both CPU and memory requests and limits.
Requests define how much free CPU/memory there must be on a node so a pod can be scheduled there; that amount is then reserved for that pod and won't be considered 'free' for scheduling of a next pod.
On the other hand, limits define the maximum amount a pod can ask for - these can be overcommitted.
Try looking at your
kubectl describe nodes
output, which lists all pods and their requests and limits. By default, the requests are 100m (10% of a core) - if you know that some of your pods don't need that much, set this lower. Then you will be able to schedule more pods on a node or at least work out the number of nodes you need.