Use spring cloud data flow to pull up a batch pod. In some cases, the pod is always pending. At this point, the developer thinks that the container failed to pull up and re-pull a pod, but the original pending pod's conditions are re-satisfied, The status changes to running, then my batch program is run twice.

The problem detail can be list below: the pod scheduled by scdf(spring cloud data flow) needed to mount glusterfs volume, in some time, the glusterfs server was down, and the pod status Backoff sync for timeout to mount. With glusterfs server running, then the backoff sync pod cloud be running also, and the problem appear.