expectedMetrics: expectedMetrics{1, 0},
		},
		{
			name:   "no-such-pod",
			podKey: podKey(testPodWithEphemeral),
		},
		{
			name: "pod-deleted",
			pods: func() []*v1.Pod {
				deleted := metav1.Now()
				pods := []*v1.Pod{testPodWithEphemeral.DeepCopy()}
				pods[0].DeletionTimestamp = &deleted
				return pods
			}(),
			podKey: podKey(testPodWithEphemeral),
		},
		{
			name:   "no-volumes",

			return nil
		}
		logger.V(5).Info("Error getting pod from informer", "pod", klog.KObj(pod), "podUID", pod.UID, "err", err)
		return err
	}

	// Ignore pods which are already getting deleted.
	if pod.DeletionTimestamp != nil {
		logger.V(5).Info("Ephemeral: nothing to do for pod, it is marked for deletion", "podKey", key)
		return nil
	}

The user of your API (an external developer) will create an invoice in your API with a POST request.

Then your API will (let's imagine):

* Send the invoice to some customer of the external developer.
* Collect the money.
* Send a notification back to the API user (the external developer).
    * This will be done by sending a POST request (from *your API*) to some *external API* provided by that external developer (this is the "callback").

You can also measure the cumulative time your build agents spent building a changeset, which will give you a sense of the amount of work the CI infrastructure has to exert. The cache's effect here is less money spent on CI resources, as you don't need as many CI agents to maintain the same number of changes built.