URL:
Reporter RHNID: mikephillips1973
Section: - Guided Exercise: Controlling Pod Scheduling
Language: en-US (English)
Workaround: Prior to step 5, students should identify which node the docker-registry pod is running on:
$ oc get pods -o wide --all-namespaces

All of step 5 should make changes to the node which the docker-registry is not running on. So if the docker-registry is running on node2, students should make node1 unschedulable and then drain node1. If the docker-registry is running on node1, then students should make node2 unschedulable and then drain node2.

Description: After node2 has been drained (Step 5.2), step 5.3 reads, "Ensure that all the application pods were re-created on node1." There seems to be a 50/50 chance that this works. The core problem involves which node the docker-registry pod happens to be running on. If it is running on node1, then the lab works as expected. If the docker-registry is running on node2 though, the lab fails. This is because the docker-registry pod wants to be scheduled on a node with the region=infra label. However, neither node1 nor node2 have the infra label anymore. This results in the docker-registry pod not being scheduled to either node. This has a second impact in that the hello pod cannot start up because the image can't be pulled from the docker-registry (as the docker-registry pod isn't running).