Why a build is Superseded by another one?

Hi, friends:

I created a new job recently. It is a centralized cleanup job that means other job trigger it when the main execution is done. So, this cleanup job should work parallelly.

This cleanup job is working , but it is not perfect. Sometimes , a build would be terminated by others. The error is sporadic.

[2025-04-23T19:51:33.151Z] 23/04/2025 19:51:30 - INFO - [ccadl-4f7b][MainThread][GardenerDriver.py:shoot_deletion_succeeded:1144][WAIT_FOR_DELETION] ccadl-4f7b-orc - [69%] Deleting Kubernetes API server, Destroying shoot infrastructure
[2025-04-23T19:51:36.434Z] 23/04/2025 19:51:36 - INFO - [ccadl-4f7b][MainThread][GardenerDriver.py:shoot_deletion_succeeded:1144][WAIT_FOR_DELETION] ccadl-4f7b-orc - [83%] Destroying external domain DNS record, Waiting until all extension resources have been deleted, Waiting until extension resources that should be handled after kube-apiserver have been deleted, Waiting until shoot infrastructure has been deleted
[2025-04-23T19:51:41.699Z] 23/04/2025 19:51:41 - INFO - [ccadl-4f7b][MainThread][GardenerDriver.py:shoot_deletion_succeeded:1144][WAIT_FOR_DELETION] ccadl-4f7b-orc - [87%] Waiting until shoot infrastructure has been deleted
[2025-04-23T19:51:49.067Z] Superseded by landscapes-cleanup#451
[2025-04-23T19:51:49.070Z] Sending interrupt signal to process
[2025-04-23T19:51:49.070Z] Killing processes
[2025-04-23T19:51:49.345Z] kill finished with exit code 0
[2025-04-23T19:51:49.808Z] Terminated
[2025-04-23T19:51:49.996Z] shoot.core.gardener.cloud/ccadl-4f7b-orc annotated
[2025-04-23T19:51:49.996Z] configmap "auditpolicy-ccadl-4f7b-orc" deleted
[2025-04-23T19:51:49.996Z] script returned exit code 143

Does some one know what is happening ?

Thank you very much.

Best Regards
Jin

And I found that the build #451 finished at exact 19:51:49.

I guess a build sends out killing signal when it finishes. And if a heading build takes long time and the later one finishes earlier, the error happens.

Do you know why it is ? And is there any solution to prevent sending the killing signal ?

Thank you very much.

Jin

This happens when you have configured a pipeline job to not allow concurrent builds and also checked the option to abort previous builds

Upon termination of the job, the process tree killer tries to kill all running process that have been started by the job.

Solutions:

  • Disable the option to abort previous builds, means that the next build will stay in queue until the current finished.
  • allow concurrent execution but you have to check if this makes sense in your scenario or could have unwanted side effects

Hi, Markus:

Thank you for your help very much.

But, I do not activate the settings you mentioned.

This can also happen if a job uses the milestone step, then a successful build will lead to cancellation of all previous still running builds