[Transform] Transform health should show timestamp of problems

[sophiec20]

Elasticsearch Version

Seen in 8.12.0

Installed Plugins

No response

Java Version

bundled

OS Version

n/a

Problem Description

When a transform fails, it goes to a FAILED state. The persistent task remains and the most recent error messages are available via _health. This helps with troubleshooting..

However, in the UI, we do not show any indication of when the failure happened. There are no messages which indicate when the failure happened. There are no timestamps on the health page.

Failures can often be due to cluster instability which cause search failures. This makes investigations more challenging as it could have been 1 hour, or 30 days ago.

Is the transform able to indicate a date for its failure... ?
Is this already in the API response somewhere, and it's just missing from the UI?

Steps to Reproduce

I think you can make a transform fail as follows, but tbc:

• Create a basic transform, configure to run continuously
• Delete/close the source index

Logs (if relevant)

No response

[elasticsearchmachine]

Pinging @elastic/ml-core (Team:ML)

[przemekwitek]

Is the transform able to indicate a date for its failure... ?

TransformHealthIssue class has a field first_occurrence which should contain a timestamp.
Unfortunately, for some reason this field is not set in all cases. More precisely it is currently set in 3 out of 7 cases in the backend code.
So definitely this is something that should be fixed.

Is this already in the API response somewhere, and it's just missing from the UI?

From what I can see, the column was in the UI since 8.7 so most likely this is a backend issue, not UI issue.
See image below: