I am upgrading K8s from 1.15 to 1.16. Before I do it, I must migrate my statefulset yaml API to the apps/v1 version. But K8s doesn't allow me to do it.
The previous version of yamls is here (the variables are stored in another file):
apiVersion: apps/v1beta1
kind: StatefulSet
metadata:
name: {{ .NAME_KAFKA }}
namespace: {{ .NS }}
spec:
serviceName: {{ .NAME_KAFKA }}-service
replicas: {{ .CLUSTER_SIZE_KAFKA }}
updateStrategy:
type: RollingUpdate
template:
metadata:
labels:
app: {{ .NAME_KAFKA }}
annotations:
prometheus.io/scrape: "true"
prometheus.io/port: "9009"
spec:
priorityClassName: {{ .PRIORITY_HIGHEST }}
nodeSelector:
lifecycle: OnDemand
terminationGracePeriodSeconds: 301
affinity:
podAntiAffinity:
requiredDuringSchedulingIgnoredDuringExecution:
- labelSelector:
matchExpressions:
- key: "app"
operator: In
values:
- {{ .NAME_KAFKA }}
topologyKey: "kubernetes.io/hostname"
containers:
- name: local-{{ .NAME_KAFKA }}
imagePullPolicy: Always
image: {{ .REPO }}/{{ .IMAGE_KAFKA }}:{{ .VERSION_KAFKA }}
resources:
requests:
memory: 768Mi
cpu: 500m
limits:
memory: 768Mi
cpu: 500m
ports:
- containerPort: 9092
name: server
- containerPort: 9009
name: prometheus
volumeMounts:
- name: datadir
mountPath: /var/lib/kafka
env:
- name: KAFKA_HEAP_OPTS
value : "-Xmx512M -Xms512M"
- name: KAFKA_OPTS
value: "-Dlogging.level=INFO"
readinessProbe:
tcpSocket:
port: 9092
initialDelaySeconds: 60
timeoutSeconds: 5
periodSeconds: 5
successThreshold: 1
failureThreshold: 5
securityContext:
runAsUser: 1000
fsGroup: 1000
volumeClaimTemplates:
- metadata:
name: datadir
spec:
accessModes:
- ReadWriteOnce
storageClassName: gp2
resources:
requests:
storage: 20Gi
I change the api-version from apps/v1beta1 to apps/v1 in the yaml file and try to apply. Predictably, I receive this error.
error: error validating "STDIN": error validating data: ValidationError(StatefulSet.spec): missing required field "selector" in io.k8s.api.apps.v1.StatefulSetSpec; if you choose to ignore these errors, turn validation off with --validate=false
So I add the field StatefulSet.spec.selector
apiVersion: apps/v1
kind: StatefulSet
metadata:
name: {{ .NAME_KAFKA }}
namespace: {{ .NS }}
spec:
selector:
matchExpressions:
- key: "app"
operator: In
values:
- {{ .NAME_KAFKA }}
serviceName: {{ .NAME_KAFKA }}-service
replicas: {{ .CLUSTER_SIZE_KAFKA }}
updateStrategy:
type: RollingUpdate
template:
metadata:
labels:
app: {{ .NAME_KAFKA }}
annotations:
prometheus.io/scrape: "true"
prometheus.io/port: "9009"
spec:
priorityClassName: {{ .PRIORITY_HIGHEST }}
nodeSelector:
lifecycle: OnDemand
terminationGracePeriodSeconds: 301
affinity:
podAntiAffinity:
requiredDuringSchedulingIgnoredDuringExecution:
- labelSelector:
matchExpressions:
- key: "app"
operator: In
values:
- {{ .NAME_KAFKA }}
topologyKey: "kubernetes.io/hostname"
containers:
- name: local-{{ .NAME_KAFKA }}
imagePullPolicy: Always
image: {{ .REPO }}/{{ .IMAGE_KAFKA }}:{{ .VERSION_KAFKA }}
resources:
requests:
memory: 768Mi
cpu: 500m
limits:
memory: 768Mi
cpu: 500m
ports:
- containerPort: 9092
name: server
- containerPort: 9009
name: prometheus
volumeMounts:
- name: datadir
mountPath: /var/lib/kafka
env:
- name: KAFKA_HEAP_OPTS
value : "-Xmx512M -Xms512M"
- name: KAFKA_OPTS
value: "-Dlogging.level=INFO"
readinessProbe:
tcpSocket:
port: 9092
initialDelaySeconds: 60
timeoutSeconds: 5
periodSeconds: 5
successThreshold: 1
failureThreshold: 5
securityContext:
runAsUser: 1000
fsGroup: 1000
volumeClaimTemplates:
- metadata:
name: datadir
spec:
accessModes:
- ReadWriteOnce
storageClassName: gp2
resources:
requests:
storage: 20Gi
But when I try to apply it, I receive this error:
The StatefulSet "kafka-name" is invalid: spec: Forbidden: updates to statefulset spec for fields other than 'replicas', 'template', and 'updateStrategy' are forbidden
According to K8s documentation, there should be a way to update the API in this yaml without rebuilding all the stateful sets in K8s. But how can I do it?
Thanks in advance.
When upgraded to 1.16, you can convert your resources.
As written in the blog you can use kubectl convert. The migration recommend convert. See kubernetes.io/blog/2019/07/18/api-deprecations-in-1-16
kubectl convert -f ./my-statefulset.yaml --output-version apps/v1
Another option is to delete the statefulset with cascade=false.
It deletes the statefulset and keeps the pods up and running. Now you should be able again to apply the new statefulset file.
kubectl delete statefulset/<your-statefulset> --cascade=false