Generate Deployment, StatefulSet, Service, Ingress, HPA, RBAC and sixteen other Kubernetes resource kinds, with safe defaults for requests and probes.
Deployment · apps/v1
3 decisions this cannot make for you.
Stateless pods, replaced freely and rolled out gradually.
The selector is immutable after creation. Getting the labels wrong means deleting and recreating the Deployment, which is downtime.
# Deployment — Stateless pods, replaced freely and rolled out gradually.
# The selector is immutable after creation. Getting the labels wrong means deleting and recreating the Deployment, which is downtime.
# Generated as a starting point. Nothing here was checked against a cluster.
apiVersion: apps/v1
kind: Deployment
metadata:
name: web
labels:
app: web
spec:
replicas: 2
selector:
matchLabels:
app: web
strategy:
type: RollingUpdate
rollingUpdate:
maxSurge: 1
maxUnavailable: 0
template:
metadata:
labels:
app: web
spec:
containers:
- name: web
image: nginx:1.27.3
ports:
- containerPort: 8080
name: http
resources:
requests:
cpu: 100m
memory: 128Mi
limits:
cpu: 500m
memory: 256Mi
readinessProbe:
httpGet:
path: /healthz
port: 8080
initialDelaySeconds: 5
periodSeconds: 10
livenessProbe:
httpGet:
path: /healthz
port: 8080
initialDelaySeconds: 15
periodSeconds: 20
securityContext:
allowPrivilegeEscalation: false
runAsNonRoot: true
readOnlyRootFilesystem: true
capabilities:
drop:
- ALL
securityContext:
runAsNonRoot: true
runAsUser: 10001
fsGroup: 10001
This produces a manifest, not a validated one. There is no Kubernetes schema here and no cluster has been contacted, so a field your cluster’s admission controllers will reject — a required label, a policy on image registries — will be rejected on apply and not here. Run kubectl apply --dry-run=server to find those: it runs the real validation and the real admission controllers without persisting anything.
Kubernetes Resource Generator runs entirely on your device — nothing is applied, no cluster is contacted, and a generated Secret carries placeholders rather than values.
You’re on the Free plan. Compare plans
The defaults lean safe on purpose. Requests and limits are on, probes are on, and the security context drops every capability and refuses privilege escalation. Not because those are always right — a batch job needs no readiness probe — but because the cost of deleting a line you did not need is nothing, and the cost of never adding one you did need is an incident.
A generator that emits the minimum valid manifest teaches the minimum, and the minimum is how pods end up BestEffort and first in line for eviction.
Each kind carries its own caution, shown beside the output, and these are specific rather than generic. A Deployment's selector is immutable after creation, so getting the labels wrong means deleting and recreating it, which is downtime. A StatefulSet does not delete its PersistentVolumeClaims when you delete it — deliberate, and also how clusters accumulate orphaned disks. A Secret's `data` is base64, which is encoding and not encryption.
A generated Secret emits `stringData` with replace-me placeholders and never a `data` block, because a base64 blob sitting in a committed file reads as protected and is not.
Nothing is applied and no cluster is contacted. A field your cluster's admission controllers will reject — a required label, a policy on image registries — will be rejected on apply and not here, so `kubectl apply --dry-run=server` is still worth running.
Because a pod with no requests is in the BestEffort QoS class, which makes it the first thing evicted when a node runs short of memory — and the pod that disappears is often not the one that caused the shortage. You can turn them off in one click; the default is the one that fails less badly.
No, and that is deliberate. It emits `stringData` with placeholder values rather than a base64 `data` block, so you have to put the real value in yourself through whatever your secret management is. A base64 blob in a committed manifest looks encrypted and is not.
It is valid YAML in the right shape, but nothing has validated it against your cluster — no schema, no admission controllers, no cluster contacted. Run `kubectl apply --dry-run=server`, which runs the real validation without persisting anything.
Because `spec.selector` cannot be changed after creation. If the labels are wrong you cannot patch them — you delete the Deployment and create it again, and between those two things nothing is serving. It is worth getting right the first time, which is why the caution sits next to the output.