-
Bug
-
Resolution: Done
-
Major
-
None
-
3
-
False
-
-
False
-
Release Notes
-
RHOAISTRAT-93 - Model serving 1H24 enhancements
-
-
Bug Fix
-
Done
-
-
-
Dashboard - General 2.7, Dashboard - General-1
-
Moderate
Is there an existing issue for this?
- I have searched the existing issues
Deploy type
Downstream version (eg. OpenShift AI 2.4)
Version
2.5 RC1
Current Behavior
When trying to deploy a Kserve model, some error might be shown to the user after clicking on 'Deploy'. In this specific case, the error was about a missing webhook (which I believe was triggered by trying to enabled modelmesh in the same cluster, FWIW).
Even though it appears as if the error prevented the successful deployment of the model, you can see from the first screenshot that the headers for the deployed models table are already loaded into the DS Project page. After a few seconds, the actual InferenceService CR is created in the Project namespace, which is then shown as a row in the table - although with unknown status.
Given the fact that there was an error during the model's deployment, the status will never be updated and keeps showing as unknown in the table - which I believe is already covered by #2172 for this use case and others.
Expected Behavior
No InferenceService is created if an error happened, the model is not shown in the table if it has no chance of actually becoming available
Steps To Reproduce
Install RHODS 2.5 RC1
Create DS Project
Create Data Connection
Try to deploy a kserve/caikit model while triggering an error (e.g. enable/disable modelmesh from the DSC CR before the attempt)
See the error