Config script does not exit or restart at errors

[solsson]

Based on a remark in #167 (comment).

For example if a kubectl command fails to look up zone identifier, due to a transient error, we'd get one pod without the rack awareness support.