Kubernetes safety fix for Job infinispawn-problem
468
This project is Admission webhook-based fix for Kubernetes 1.10-bug that causes cluster to overuse it's resources if Job fails.
1. Intercept all admission webhooks related to Pods
2. If the request is related to anything else than Job-related pod just accept and return
3. Check the count of pods related to the Job
- If less than Job.Spec.BackoffLimit: accept and return
- Else: Deny and return
This will cause Job to stop respawning new pods but will also freeze the job after spawning X-amount of failed pods. All pods are still there and must be manually deleted.
kubectl apply -f deploy.yml
# When the deployment starts it will automatically register itself to cluster
Content type
Image
Digest
Size
12.1 MB
Last updated
almost 8 years ago
docker pull emblica/kube-safe-maxpods