When deploying the sonataflow-inference-demo CR, the kubernetes API server starts spawning multiple pods that end up being in terminating state seconds after they're spawned. At the end, only 1 pod remains in Running state as expected, but there are many pods still in Terminating state that are not cleaned up. This issue does not depend on the namespace: it happens in both default and a newly created namespace.

Sonataflow-inference-demo repository:

https://github.com/hbelmiro/sonataflow-inference-pipeline-demo

Steps to reproduce

• Deploy the latest sonataflow operator version (use the code in main and not the latest image)
• Change the base image referenced in the 01-sonataflow-platform.yaml to quay.io/ricardozanini/sonataflow-python-devmode:latest
• Run the deploy.sh command

Expected result:

• One pod is created that eventually reaches the Running/Ready state

Actual result:

• Multiple pods are created that reach the Terminating state shortly, while only 1 reaches the Running,Ready state.

$>oc get pod
NAME                        READY   STATUS        RESTARTS   AGE
pipeline-55bb5b889d-lz22x   0/1     Terminating   0          17s
pipeline-5c8bd7db86-dvqlw   0/1     Terminating   0          17s
pipeline-fc9b44cdb-jdwsv    1/1     Running       0          4m17s