Additional test for failures?

[kkaempf]

Test upgrade failures. There are three different stages where it could actually fail:

• Fails at elemental upgrade execution
  • it should not continue with the upgrade of other nodes. Current node could be uncordoned and be functional again.
• Systemd fails to reboot
  • it should reboot to fallback and keep the node cordoned. The node could eventually be uncordoned and be functional again using fallback system?
• Boots on degraded mode (e.g. no network connection)
  • it should reboot to fallback and keep the node cordoned. The node could eventually be uncordoned and be functional again using fallback system?