Why was DataScript removed? How to run an additional ( init ) script after Cluster provisioning?

[lukasmrtvy]

Why was DataScript ( #7968 ) removed? How to run an additional ( init ) script after Cluster provisioning?

My use-case:
#3665

Thanks

[shanshanying]

Hi @lukasmrtvy

We provide 'OpsDefinition' instead for these customized cases.
User can define an OpsDefinition first and create an Ops of type 'custom'.

[shanshanying]

there are several OpsDefinition in the redis addon. they accept paramters and define the behavior.
e.g. https://github.com/apecloud/kubeblocks-addons/blob/main/addons/redis/templates/opsdefinition-account.yaml

To perform such an operation against the target cluster, you can create an ops of type custom refering to this opsdef:

apiVersion: operations.kubeblocks.io/v1alpha1
kind: OpsRequest
metadata:
  name: xxx
spec:
  clusterName: <clusterName>
  custom:
    components:
    - componentName: <componentName>
       parameters:   # paramters are those defined in opsdef.
      - name: 
        value: 
    opsDefinitionName: <opsDefName>
  type: Custom

[lukasmrtvy]

thanks, seems like this is working:

apiVersion: operations.kubeblocks.io/v1alpha1
kind: OpsRequest
metadata:
  name: postgresql-create-database
  namespace: foobar
spec:
  clusterName: postgresql
  preConditionDeadlineSeconds: 0
  custom:
    components:
    - componentName: postgresql
      parameters:
      - name: DATABASE_NAME
        value: foobar
    opsDefinitionName: postgresql-create-database
  type: Custom
---
apiVersion: operations.kubeblocks.io/v1alpha1
kind: OpsDefinition
metadata:
  name: postgresql-create-database
spec:
  preConditions:
  - rule:
      expression: '{{ eq .component.status.phase "Running" }}'
      message: Component is not in Running status.
  podInfoExtractors:
    - name: master-connection
      podSelector:
        multiPodSelectionPolicy: Any
      env:
        - name: PGUSER
          valueFrom:
            envRef:
              envName: POSTGRES_USER
        - name: PGPASSWORD
          valueFrom:
            envRef:
              envName: POSTGRES_PASSWORD
        - name: PGHOST
          valueFrom:
            envRef:
              envName: POSTGRES_POD_FQDN_LIST
  parametersSchema:
    openAPIV3Schema:
      properties:
        DATABASE_NAME:
          description: |
            DATABASE_NAME is the name of the PostgreSQL database to be created.
          type: string
  actions:
    - name: create-database
      parameters: ["DATABASE_NAME"] 
      workload:
        type: Pod
        podInfoExtractorName: master-connection 
        backoffLimit: 0
        podSpec:
          containers:
            - name: psql-create-database
              image: alpine/psql:18.1
              imagePullPolicy: IfNotPresent
              command: ["/bin/sh", "-c"]
              args:
                - |
                  psql -d postgres -c "CREATE DATABASE $DATABASE_NAME;"

[lukasmrtvy]

@shanshanying can one wait for a healthy Cluster in OpsRequest to trigger the operation, or should I use some wait loop in script?

I want to deploy Cluster and OpsRequest ( create database ) at the same time from IaaC. The thing is, its not possible, as OpsRequest will fail, because Cluster is not healthy in the first seconds. ( failing to podSelector )

Thanks

[wangyelei]

Hi @lukasmrtvy, you can try combining opsrequest.spec.preConditionDeadlineSeconds and opsdefinition.spec.preConditions to implement this capability

[shanshanying]

HI @lukasmrtvy

while setting PG_HOST, POSTGRES_POD_FQDN_LIST is not a good choice if there are multiple replicas in your cluster.

        - name: PGHOST
          valueFrom:
            envRef:
              envName: POSTGRES_POD_FQDN_LIST  # POSTGRES_POD_FQDN_LIST  is a list

Some options are:

Option 1. concat the priamry pod fqdn
set the target pod to prmary

    - name: master-connection
      podSelector:
        multiPodSelectionPolicy: Any
        role: primary  # add the role selecor

and then concat full primary fqdn using target pod name and headless svc name

Option 2. use the svc '-postgresql-postgresql'
this svc is created with role selector: kubeblocks.io/role: primary

[lukasmrtvy]

It just doesn't work, OpsRequest failing to:

    - lastTransitionTime: '2025-12-27T08:55:08Z'
      message: >-
        Failed to process OpsRequest: postgresql-create-database in cluster:
        postgresql, more detailed informations in status.components
      reason: OpsRequestFailed
      status: 'False'
      type: Failed

kubectl get cluster
NAME         CLUSTER-DEFINITION   TERMINATION-POLICY   STATUS    AGE
postgresql   postgresql           Delete               Running   58m

kubectl get  component 
NAME                    DEFINITION            SERVICE-VERSION   STATUS    AGE
postgresql-postgresql   postgresql-17-1.0.1   17.5.0            Running   58m

apiVersion: apps.kubeblocks.io/v1
kind: Cluster
metadata:
  name: postgresql
spec:
  terminationPolicy: Delete
  clusterDef: postgresql
  topology: replication
  componentSpecs:
    - name: postgresql
      serviceVersion: "17.5.0"
      disableExporter: true
      replicas: 1
      volumeClaimTemplates:
        - name: data
          spec:
            storageClassName: ""
            accessModes:
              - ReadWriteOnce
            resources:
              requests:
                storage: 1Gi
---
apiVersion: operations.kubeblocks.io/v1alpha1
kind: OpsRequest
metadata:
  name: postgresql-create-database
  namespace: default
spec:
  clusterName: postgresql
  preConditionDeadlineSeconds: 0
  custom:
    components:
    - componentName: postgresql
      parameters:
      - name: DATABASE_NAME
        value: registry
    opsDefinitionName: postgresql-create-database
  type: Custom
---
apiVersion: operations.kubeblocks.io/v1alpha1
kind: OpsDefinition
metadata:
  name: postgresql-create-database
  # namespace: ""
spec:
  preConditions:
  - rule:
      expression: '{{ eq .component.status.phase "Running" }}'
      message: Component is not in Running status.
  podInfoExtractors:
    - name: master-connection
      podSelector:
        multiPodSelectionPolicy: Any
        role: primary
      env:
        - name: PGUSER
          valueFrom:
            envRef:
              envName: POSTGRES_USER
        - name: PGPASSWORD
          valueFrom:
            envRef:
              envName: POSTGRES_PASSWORD
        - name: PGHOST
          valueFrom:
            envRef:
              envName: POSTGRES_POD_FQDN_LIST
  parametersSchema:
    openAPIV3Schema:
      properties:
        DATABASE_NAME:
          description: |
            DATABASE_NAME is the name of the PostgreSQL database to be created.
          type: string
  actions:
    - name: create-database
      parameters: ["DATABASE_NAME"] 
      workload:
        type: Job
        podInfoExtractorName: master-connection 
        backoffLimit: 10
        podSpec:
          containers:
            - name: psql-create-database
              image: alpine/psql:18.1
              imagePullPolicy: IfNotPresent
              command: ["/bin/sh", "-c"]
              args:
                - |
                  pg_isready -d postgres && psql -a -d postgres -c "CREATE DATABASE $DATABASE_NAME;"

[shanshanying]

HI @lukasmrtvy ,

pls increase preConditionDeadlineSeconds: 0 to some non-zero value in your opsrequest.

[shanshanying]

Hi @lukasmrtvy
I will close the issue for now. if you encounter any issues similar or the same issue again in the future, pls feel free to reopen it.