It seems that the control loop gets stuck in the DELETING state under specific circumstances. I created a vcenter event driver called vxrail. I deleted this event driver and confirmed that it was no longer present when calling dispatch get event-driver. I then immediately made a call to create a new event driver with the same name. This resulted in the newly created event-driver being in an ERROR state. Attempting to delete this event-driver then results in the resource being stuck in the deleting state.
I assume the pod containing the old event driver was not done terminating when the new create call was made. That resulted in the create call erroring out without creating a pod. The delete call is now failing because it can't find the pod to delete. I suspect we might have similar bugs in the other control loop handlers.
Detailed Description
It seems that the control loop gets stuck in the DELETING state under specific circumstances. I created a vcenter event driver called vxrail. I deleted this event driver and confirmed that it was no longer present when calling
dispatch get event-driver
. I then immediately made a call to create a new event driver with the same name. This resulted in the newly created event-driver being in an ERROR state. Attempting to delete this event-driver then results in the resource being stuck in the deleting state.I assume the pod containing the old event driver was not done terminating when the new create call was made. That resulted in the create call erroring out without creating a pod. The delete call is now failing because it can't find the pod to delete. I suspect we might have similar bugs in the other control loop handlers.
Log output
Context
Possible Implementation
Complexity
Impact
Your Environment