Misconfiguration Handling in HBN DPUService Will Revert to Last Working Config

[tobiasgiese]

Description

Deploying the HBN DPUService with an "illegal" configuration results in inconsistent and unexpected behavior, depending on the nature of the configuration issue:

Scenarios:

1. Invalid YAML Configuration:

   • If the configuration YAML is syntactically invalid, the hbn-init container validates it, crashes, and logs the error. The hbn-doca service does not start.
   • Expected Behavior: This scenario is handled correctly.

2. Valid YAML, But Illegal Configuration:

   • If the YAML configuration is valid but contains values that are illegal from an NVUE perspective:

     • The doca-hbn service starts (unexpected behavior: it should not start or should start with an empty configuration).
     • The doca-hbn service applies the "last known good" configuration (unexpected behavior: it should not retain any previous configuration).

   • Any existing startup.yaml file on the filesystem is not removed by the hbn-init container. As a result:

     • The service improperly "remembers" prior configurations stored on the filesystem.
     • The service does not attempt to re-apply the new configuration correctly during subsequent restarts.

Users deploying misconfigured HBN DPUService may experience a misleading or outdated service state.

Resolution

This issue is under investigation.

Note: There is currently no workaround for this issue.