👀 Views: 19 💬 Answers: 1 📅 Created: 2025-06-16

kubernetes jobs custom-resources YAML

I'm reviewing some code and After trying multiple solutions online, I still can't figure this out... I'm sure I'm missing something obvious here, but I'm working with an scenario where a Kubernetes Job that I created to process a batch of tasks is not completing as expected. After about an hour, it fails with a `DeadlineExceeded` behavior. I'm using Kubernetes version 1.24 and I've defined a custom resource definition (CRD) to handle my jobs based on a specific business model. Here’s the relevant snippet of my Job definition: ```yaml apiVersion: batch/v1 kind: Job metadata: name: my-batch-job spec: template: spec: containers: - name: task-processor image: my-docker-repo/task-processor:latest args: - "--process" - "all-tasks" restartPolicy: Never backoffLimit: 4 ttlSecondsAfterFinished: 300 ``` I've also created a CRD that looks like this: ```yaml apiVersion: mycompany.com/v1 kind: MyJob metadata: name: example-job spec: jobName: my-batch-job maxRetries: 3 timeLimit: 3600 ``` The `spec.timeLimit` in my CRD is set to 3600 seconds, but it seems the Job is exceeding this limit. I verified that the Job is meant to process around 100 items from a database, but it's getting exploring on the first few. I've checked the logs of the Job container using `kubectl logs my-batch-job-xxxxx`, and they don't seem to indicate any errors—just that it's processing items. Additionally, I’ve tried increasing the `timeLimit` in the CRD to 7200 seconds, but I still experience the same scenario. I also ensured that the Job's `restartPolicy` is set to `Never`, so it shouldn't be restarting in a loop, which could potentially lead to confusion with the `DeadlineExceeded` behavior. Has anyone encountered this scenario with Kubernetes Jobs and CRDs? Are there any specific configurations I should check or best practices to follow to avoid this question? I'd really appreciate any guidance on this. Am I missing something obvious? I appreciate any insights! How would you solve this?