The issue manifests as the prometheus counter ansible_puller_play_summary containing all zeros even with successful runs. This is due to two problems:
1) It is assumed in the counter logic that the hostname was the host alias used during the ansible playbook. This is not always correct.
2) During failures (and when "debug" is on), the ansible runner is exiting before parsing the metric result
This was manually tested in a "scratch" environment
The issue manifests as the prometheus counter ansible_puller_play_summary containing all zeros even with successful runs. This is due to two problems: 1) It is assumed in the counter logic that the hostname was the host alias used during the ansible playbook. This is not always correct. 2) During failures (and when "debug" is on), the ansible runner is exiting before parsing the metric result
This was manually tested in a "scratch" environment