[BUG] No resource quota support in Spark Connect

[Vadim-elo]

What happened?

• When a server pod fails to run due to quota, the quota error can only be seen in the spark controller operator logs. There are no events or errors in the sparkconnect crd, as is done for sparkapplication
• Additionally, the Spark Connect server pod can start if it has enough resources, but the executor pods will not start if they have no more resources left.

Reproduction Code

Add -enable-resource-quota-enforcement=true to spark operator webhook and configure SparkConnect so that it requests more resources than are available

Expected behavior

The error should be displayed in the Spark Connect crd

Actual behavior

Spark Connect's CRD is stuck in an empty state, pods aren't starting, the status doesn't change, and the Spark operator is cyclically trying to start Spark Connect.

Environment & Versions

• Kubernetes Version: 1.25.11
• Spark Operator Version: 2.3.0
• Apache Spark Version: >= 3.5.0

Additional context

Full quota support is needed in Spark Connect, as is done for Spark application.

Impacted by this bug?

Give it a 👍 We prioritize the issues with most 👍

[Denyme24]

hey team @Vadim-elo @ChenYi015 @ImpSy @nabuskey, i would like to take this up.

Approach

1. Add validating webhook for SparkConnect (similar to SparkApplication)

   • Validate resource quotas before creation
   • Calculate total resources: server + (executors × replicas)

2. Enhance controller to update SparkConnect status with quota errors

   • Set state to Failed with clear error messages
   • Prevent silent failures

3. Register webhook in operator startup

This will provide the same quota enforcement that SparkApplication currently has, with clear error messages visible to users via kubectl describe.

[Denyme24]

/assign

[github-actions]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.

[tariq-hasan]

/remove-lifecycle stale