HostPreflight hostOS collector does not run on Amazon Linux 2023

[xavpaice]

Bug Description

When running host preflights, the hostOS collector does not respond correctly when run on Amazon Linux 2023

./preflight https://raw.githubusercontent.com/replicatedhq/kURL/main/addons/containerd/1.6.9/host-preflight.yaml
 * [Host OS Info] Running collector...
Error: no data has been collected

Expected Behavior

The hostOS collector should run successfully.

Steps To Reproduce

Create a VM running Amazon Linux 2023.
Obtain the latest preflight binary (by whichever means is convenient)

Use the following as a preflight spec:

apiVersion: troubleshoot.sh/v1beta2
kind: HostPreflight
metadata:
  name: hostOS
spec:
  collectors:
    - hostOS: {}
  analyzers:
    - hostOS:
        outcomes:
          - pass:
              when: "centos = 7"
              message: "CentOS 7 is supported"
          - pass:
              when: "centos = 8"
              message: "CentOS 8 is supported"
          - fail:
              when: "ubuntu = 16.04"
              message: "Ubuntu 16.04 is not supported"
          - pass:
              when: "ubuntu = 18.04"
              message: "Ubuntu 18.04 is supported"
          - pass:
              when: "ubuntu = 20.04"
              message: "Ubuntu 20.04 is supported"

Additional Context

Include the following information.

• Troubleshoot version: 0.76.2
• Operating system: Amazon Linux
• Operating system version: 2023

[banjoh]

This is the intended behaviour, but the error message is misleading. It ought to be something like completed with no analysis results. The spec has no default outcome so not analysis results are reported. If you are to have it look like below, it will work. Note the last message: "No supported OS found" outcome which is the default/catch-all since it has no when condition

apiVersion: troubleshoot.sh/v1beta2
kind: HostPreflight
metadata:
  name: hostOS
spec:
  collectors:
    - hostOS: {}
  analyzers:
    - hostOS:
        outcomes:
          - pass:
              when: "centos = 7"
              message: "CentOS 7 is supported"
          - pass:
              when: "centos = 8"
              message: "CentOS 8 is supported"
          - fail:
              when: "ubuntu = 16.04"
              message: "Ubuntu 16.04 is not supported"
          - pass:
              when: "ubuntu = 18.04"
              message: "Ubuntu 18.04 is supported"
          - pass:
              when: "ubuntu = 20.04"
              message: "Ubuntu 20.04 is supported"
          - fail:
              message: "No supported OS found"

There has to be an explicit default outcome cause we cannot assume pass, fail or warn.

I wonder whether preflight should succeed in such cases for consistency, but not have any results in the analysis.json file. I say so cause if there was another analyser that generated an outcome, this preflight run would succeed but have 1 outcome, not 2 as would be expected.

I wonder if we should log a warning if an analysis ends up with no outcome. This would help troubleshoot specs. In future we can validate specs.