Skip to content

Spark Operator (Kubeflow)

External Operator

The Spark Operator is a Helm/kustomize-managed operator from kubeflow/spark-operator. ODH maintains a fork at opendatahub-io/spark-operator. It manages SparkApplication lifecycle on Kubernetes.

Overview

Property Value
Operator spark-operator
Namespace spark-operator
Repository kubeflow/spark-operator
Components 2
Reconcile Timeout 300s
Max Reconcile Cycles 10

Components

controller

Namespace: spark-operator

Managed Resources

API Version Kind Name Namespace
apps/v1 Deployment spark-operator-controller spark-operator
v1 ServiceAccount spark-operator-controller spark-operator
coordination.k8s.io/v1 Lease spark-operator-controller spark-operator

Replicas: 1

Steady-State Checks

Type Kind Name Namespace Condition
conditionTrue Deployment spark-operator-controller spark-operator Available

Timeout: 60s


webhook

Namespace: spark-operator

Managed Resources

API Version Kind Name Namespace
apps/v1 Deployment spark-operator-webhook spark-operator
v1 ServiceAccount spark-operator-webhook spark-operator
v1 Service spark-operator-webhook spark-operator
coordination.k8s.io/v1 Lease spark-operator-webhook spark-operator

Replicas: 1

Webhooks

Name Type Path
spark-operator-webhook validating /validate
spark-operator-webhook mutating /mutate

Steady-State Checks

Type Kind Name Namespace Condition
conditionTrue Deployment spark-operator-webhook spark-operator Available

Timeout: 60s

Results Summary

12 experiments tested. Verified findings: NetworkPartition Failed (controller permanently non-functional), DeploymentScaleZero Degraded (both components). Other experiments Resilient.

Verdict Count Notes
Resilient 9 Core failure modes recover automatically
Degraded 2 DeploymentScaleZero for both controller and webhook
Failed 1 NetworkPartition leaves controller permanently non-functional

Pages