A simple tool built on top of jq that summarizes the CPU and memory
requests/limits in Kubernetes manifests. It takes into consideration HorizontalPodAutoscaler
objects that may affect the number of replicas.
The output is a JSON array containing an object for each workload resource (Pod, Deployment, StatefulSet...)
found in the input. The output is consistent regardless of the order of the inputs files or the keys in the
YAML, so it can be used with diff to identify changes exclusively related to the definition of CPU and
memory.
The kuberes script needs:
Choose an operation mode and pass the names of the files to parse as parameters or pipe the YAML text.
The latter option is useful in combination with helm template.
./kuberes sum example/*
# or
cat example/deployment.yaml | ./kuberes sumTake into consideration that YAML documents need to be explicitly separated by ---, so doing something
like this will not work as expected:
cat example/* | ./kuberes sum # Documents are appended one after the other without using ---A Dockerfile is included in case you prefer not to install locally the dependencies.
docker build -t kuberes:latest .Mount the folder with the Kubernetes resources into /data and pass the usual CLI parameters:
docker run --rm -v $PWD/example:/data:ro kuberes sum '/data/*' The tool provides different working modes depending on the desired level of information aggregation.
The following outputs correspond to running the tool on the files of the examples folder.
Raw information from the YAML files. Also add minReplicas and maxReplicas values based on Horizontal Pod Autoscalers.
[
{
"apiVersion": "apps/v1",
"containers": [
{
"image": "httpd",
"name": "httpd",
"resources": {
"limits": {
"cpu": "500m",
"memory": "1024Mi"
},
"requests": {
"cpu": "250m",
"memory": "256Mi"
}
}
},
{
"image": "fluentd",
"name": "fluentd",
"resources": {
"limits": {
"cpu": "200m",
"memory": "256Mi"
},
"requests": {
"cpu": "50m",
"memory": "128Mi"
}
}
}
],
"kind": "Deployment",
"maxReplicas": 5,
"minReplicas": 1,
"name": "my-deplo"
},
{
"apiVersion": "apps/v1",
"containers": [
{
"image": "httpd",
"name": "httpd",
"resources": {
"limits": {
"cpu": "500m",
"memory": "1024Mi"
},
"requests": {
"cpu": "250m",
"memory": "256Mi"
}
}
},
{
"image": "fluentd",
"name": "fluentd",
"resources": {
"limits": {
"cpu": "200m",
"memory": "256Mi"
},
"requests": {
"cpu": "50m",
"memory": "128Mi"
}
}
}
],
"kind": "Deployment",
"maxReplicas": 3,
"minReplicas": 3,
"name": "my-other-deplo"
},
{
"apiVersion": "v1",
"containers": [
{
"image": "nginx",
"name": "my-pod",
"resources": {
"limits": {
"cpu": "200m",
"memory": "512Mi"
},
"requests": {
"cpu": "100m",
"memory": "256Mi"
}
}
},
{
"image": "fluentd",
"name": "sidecar",
"resources": {
"limits": {
"cpu": "50m",
"memory": "50Mi"
},
"requests": {
"cpu": "10m",
"memory": "25Mi"
}
}
}
],
"kind": "Pod",
"maxReplicas": 1,
"minReplicas": 1,
"name": "my-pod"
}
]Top level objects with lists of the underlying container requests and limits.
[
{
"apiVersion": "apps/v1",
"kind": "Deployment",
"limits": {
"cpu": [
"500m",
"200m"
],
"memory": [
"1024Mi",
"256Mi"
]
},
"maxReplicas": 5,
"minReplicas": 1,
"name": "my-deplo",
"requests": {
"cpu": [
"250m",
"50m"
],
"memory": [
"256Mi",
"128Mi"
]
}
},
{
"apiVersion": "apps/v1",
"kind": "Deployment",
"limits": {
"cpu": [
"500m",
"200m"
],
"memory": [
"1024Mi",
"256Mi"
]
},
"maxReplicas": 3,
"minReplicas": 3,
"name": "my-other-deplo",
"requests": {
"cpu": [
"250m",
"50m"
],
"memory": [
"256Mi",
"128Mi"
]
}
},
{
"apiVersion": "v1",
"kind": "Pod",
"limits": {
"cpu": [
"200m",
"50m"
],
"memory": [
"512Mi",
"50Mi"
]
},
"maxReplicas": 1,
"minReplicas": 1,
"name": "my-pod",
"requests": {
"cpu": [
"100m",
"10m"
],
"memory": [
"256Mi",
"25Mi"
]
}
}
]Top level objects with the sum of the underlying container requests and limits.
[
{
"apiVersion": "apps/v1",
"kind": "Deployment",
"limits": {
"cpu": "700ms",
"memory": "1280MiB"
},
"maxReplicas": 5,
"minReplicas": 1,
"name": "my-deplo",
"requests": {
"cpu": "300ms",
"memory": "384MiB"
}
},
{
"apiVersion": "apps/v1",
"kind": "Deployment",
"limits": {
"cpu": "700ms",
"memory": "1280MiB"
},
"maxReplicas": 3,
"minReplicas": 3,
"name": "my-other-deplo",
"requests": {
"cpu": "300ms",
"memory": "384MiB"
}
},
{
"apiVersion": "v1",
"kind": "Pod",
"limits": {
"cpu": "250ms",
"memory": "562MiB"
},
"maxReplicas": 1,
"minReplicas": 1,
"name": "my-pod",
"requests": {
"cpu": "110ms",
"memory": "281MiB"
}
}
]Top level objects with minRequests, minLimits, maxRequests and maxLimits, taking into consideration minReplicas and maxReplicas.
[
{
"apiVersion": "apps/v1",
"kind": "Deployment",
"limits": {
"cpu": {
"max": "3500ms",
"min": "700ms"
},
"memory": {
"max": "6400MiB",
"min": "1280MiB"
}
},
"maxReplicas": 5,
"minReplicas": 1,
"name": "my-deplo",
"requests": {
"cpu": {
"max": "1500ms",
"min": "300ms"
},
"memory": {
"max": "1920MiB",
"min": "384MiB"
}
}
},
{
"apiVersion": "apps/v1",
"kind": "Deployment",
"limits": {
"cpu": {
"max": "2100ms",
"min": "2100ms"
},
"memory": {
"max": "3840MiB",
"min": "3840MiB"
}
},
"maxReplicas": 3,
"minReplicas": 3,
"name": "my-other-deplo",
"requests": {
"cpu": {
"max": "900ms",
"min": "900ms"
},
"memory": {
"max": "1152MiB",
"min": "1152MiB"
}
}
},
{
"apiVersion": "v1",
"kind": "Pod",
"limits": {
"cpu": {
"max": "250ms",
"min": "250ms"
},
"memory": {
"max": "562MiB",
"min": "562MiB"
}
},
"maxReplicas": 1,
"minReplicas": 1,
"name": "my-pod",
"requests": {
"cpu": {
"max": "110ms",
"min": "110ms"
},
"memory": {
"max": "281MiB",
"min": "281MiB"
}
}
}
]Summary of the total minRequests, minLimits, maxRequests and maxLimits across all objects.
This information means that your pods:
- To just exist without activity, need
.requests.memory.minMiB and.requests.cpu.minmilicores available in the cluster. - To fully scale out to the max number of replicas, need
.requests.memory.maxMiB and.requests.cpu.maxmilicores available in the cluster. - If fully scaled out, will try to phagocyte up to
.limits.memory.maxMiB and.requests.cpu.minmilicores.
[
{
"limits": {
"cpu": {
"max": "5850ms",
"min": "3050ms"
},
"memory": {
"max": "10802MiB",
"min": "5682MiB"
}
},
"requests": {
"cpu": {
"max": "2510ms",
"min": "1310ms"
},
"memory": {
"max": "3353MiB",
"min": "1817MiB"
}
}
}
]The output follows a set of rules to make it deterministic:
- Resources are sorted by name.
- If there are resources that share name, they are ordered by kind (Pod, Deployment...).
- Object keys are sorted alphabetically
However, the tool, does not take into consideration namespace.