In the values.yaml file, there is an entry
resources: {}
This allows you specify values for CPU and memory:
https://kubernetes.io/docs/tasks/configure-pod-container/assign-memory-resource/
https://kubernetes.io/docs/tasks/configure-pod-container/assign-cpu-resource/
Memory is specified in bytes (can be M or Mi, or K or Ki etc.). CPU is specified in a fraction of a “CPU Unit” where 1 CPU unit = 1 AWS vCPU. They can additionally be specified using m as a unit to represent a “milliCPU” (1/1000 CPU) so 100m = 0.1 CPU.
The resources have 2 values that can be configued, a request and a limit.
When provisioning your pods, Kubernetes will look at the total CPU units and memory it has available for the cluster/namespace and if your request exceeds the available resource, your pod will not be provisioned.
When pods are running, Kubernetes will monitor the limit. The container’s memory can exceed the request, but it cannot exceed the limit. Once the limit is exceed, the container will become a candidate for termination and will restart. The CPU limit acts as a throttle. The container will never be allowed to exceed the CPU limit.
resources:
limits:
cpu: 150m
memory: 1024Mi
requests:
cpu: 100m
memory: 512Mi
In the above example, when provisioning the pod, Kubernetes will only create the pod if it has 100m CPU and 512Mi memory available. However, it will allow the CPU time to grow to 150m before throttling and the memory to grow to 1024Mi before killing the container.
So, you should set the requests values to be the amount you expect the container to use under normal circumstances. The limits should be set to allow your container to have short bursts of activity where they exceed the normal resources without termination. Of course, if you are going to exceed the requests values for any length of time, then you should consider horizontal scaling. However the higher limits will give extra capacity for a short period whilst your autoscaling fires up.
kubernetes helm docker